Nina and Sam Data

Accenture Technology Vision 2023 - Page 05

In healthcare, this shared reality is pushing us towards personalized care. With funding from the European Union, a consortium of hospitals, researchers, and startups have joined the Neurotwin project, an effort to build digital twins of individual human brains.1, 2 Each twin would be used to help healthcare providers understand and predict triggers for neurological diseases and improve preventative interventions. The project is set to launch two studies on individuals suffering from Alzheimer's and epilepsy in 2023.3

Accenture Technology Vision 2023 - Page 06

For example, the very first 3D-printed steel bridge in the world was built in Amsterdam and was designed with this kind of digital-physical hybridization in mind.12 Not only was the bridge created with 3D-printing and robotic arms, but it was also designed specifically to coexist with a digital twin. It was embedded with a cutting-edge sensor network that now feeds a digital twin with real-time data on vibration, strain, weather conditions, and more

Accenture Technology Vision 2023 - Page 07

Other companies are endeavoring to thread digital back into physical. Google has been slowly integrating a new wayfinding feature into Google Maps called Live View. It creates an overlay of details about, or directions around, users' environments via geolocators and smartphone cameras.13

Accenture Technology Vision 2023 - Page 07

And Snap has been expanding its AR technology far beyond filters, partnering with companies like Amazon, Puma, and Ralph Lauren to integrate 3D product modeling and augmented reality try-on features into the shopping experience.14,15 This blurring of the digital-physical boundary is now exposing new modes of commerce: the company is rolling out a new "Dress Up" feature where users can discover, try on, and shop for new products directly in the app, and the company reported 250 million people had used its AR shopping lenses over five billion times in just one year alone

Accenture Technology Vision 2023 - Page 07

OpenAI open-sourced one of the most powerful Automatic Speech Recognition (ASR) neural networks, called Whisper, which was trained on nearly 700,000 hours of speech-related data and approaches humanlike levels of accuracy.16 ASR and natural language processing are fusing the digital and physical by doing away with today's abstractions like keyboards and gestures and allowing people to engage the digital world in one of the most human ways possible: speech.

Accenture Technology Vision 2023 - Page 07

Or consider the growing trend of cobots (collaborative robots), a market that is set to expand to $16.3 billion by 2028.17 These machines let the power of machine intelligence and automation bleed into the physical world, allowing people to work more naturally and unearth new efficiencies. In one example, Moxi, a cobot designed for hospitals, was able to save healthcare workers at one hospital 3,200 hours by taking care of routine tasks like deliveries and allowing the staff to spend more time giving care to patients.1

Accenture Technology Vision 2023 - Page 11

Take AlphaFold, a recent innovation out of DeepMind. While other efforts like AlphaGo and AlphaStar explored AI's ability to win different types of games, AlphaFold turned to a far more practical and significant application: protein folding.19 Proteins are the building blocks of biology, and their unique threedimensional structures determine what function they will perform. So, for instance, if a pharmaceutical company were developing a new drug, understanding various protein shapes would be critical to understanding how the drug will interact with the person it is being administered to. The problem is, historically, these are extremely labor intensive to study, or computationally intensive to simulate. But in the summer of 2022, DeepMind publicly released a database of 200 million different protein structures -which covers nearly every protein known to human beings-now available to all companies and researchers around the world.20

Accenture Technology Vision 2023 - Page 14-15

Take Mango Materials. The renewable bioproducts startup collaborated with researchers around the world to develop a bacteria-produced bioplastic designed to decompose in the ocean.30 So far, five oceanographic equipment manufacturing partners have committed to replacing all their traditional plastic with this new degradable bioplastic once it's ready, with commercialization and other product applications being explored as the collaboration evolves.31

Accenture Technology Vision 2023 - Page 15

When it comes to using data to improve treatment, hospitals are typically limited to their own datasets for patient privacy reasons. But in a recent study published in Nature Medicine, 20 hospitals from around the world participated in training a federated learning model called EXAM, which predicted future oxygen requirements of COVID-19 patients.32 The model leveraged data, including patient vital signs, laboratory data, and chest X-rays, from all of the participating hospitals-but each hospital trained their own copy of the AI model and periodically shared updates with a centralized server, which then aggregated them to train the global model. In this case, the hospitals created a shared reality that they couldn't have before, using federated learning to safely share digital data on physical information to improve AI predictions.

Accenture Technology Vision 2023 - Page 25

In August 2022, Microsoft launched Microsoft Entra Verified ID, a new product based on decentralized identity standards. The company's vision is to build a new future for digital identity, where instead of widely spreading identity data across countless apps and platforms, people and organizations can have greater trust in and control over what information is accessed, by whom, and for how long.45,46 And already, proof of concepts and pilot programs are underway with two universities, a healthcare system, and a government service.47

Accenture Technology Vision 2023 - Page 25

And other companies are also innovating in this space. Early Warning Services and seven banks including Capital One, Wells Fargo, Chase, and Bank of America launched Authentify -an identity verification product that lets users log into their online banks from participating websites and apps, streamlining the process for sharing sensitive banking data.48

Accenture Technology Vision 2023 - Page 25

Apple has expanded its Wallet app, moving beyond payment information to let users store and share government-issued IDs like driver's licenses.49 And while not an example of innovation itself, the Bill & Melinda Gates Foundation committed $200 million to digital public infrastructure projects, including digital ID.50 These are only a fraction of the initiatives underway, and the market for digital identity is expected to swell from $27.9 billion in 2022 to $70.7 billion by 2027.5

Accenture Technology Vision 2023 - Page 26

b.well saw this as an identity challenge, and partnered with Mastercard to build a new intelligent identity platform for healthcare systems. Their platform replaces emails and passwords with mobile-native biometrics and provides secure document scanning and storage, giving users greater control over their medical information and letting them seamlessly share it with providers and payers.53 Moreover, it opens the possibility of integrating other data, like proof of income, to ease the approval process for insurance policies or f inancial relief programs. And on the healthcare side, it gives providers access to standardized and high-fidelity information about those in their care, all while reducing fraud and risk.5

Accenture Technology Vision 2023 - Page 28

Look at how Wien Energie, an Austrian energy provider, recently tokenized one of the largest photovoltaic (PV) solar plants in the country.55 The company created unique identifying tokens for each individual PV module in the plant, which were then sold to customers. Token holders receive an annual payment relative to the amount of energy produced by the plant. Right now, it's a closed system, so payment is made in the form of discounts on electric bills. But the company envisions a future where these tokens can be used to fund new plants, used as proof of origin for energy sources, or traded to fuel things like electric vehicles. By giving the plant and solar cells a unique digital ID, the company was able to expose innovative new business models, ones that turned customers into stakeholders.

Accenture Technology Vision 2023 - Page 28

The UN calls for the creation of a legal ID for all people by 2030 as part of its Sustainable Development Goals, so it's no surprise governments around the world are beginning to make digital identity a top priority.56 As one of the first digital identity adopters, Estonia has 98% of its citizens registered with its electronic identity system (e-ID) that is used for public services like voting, health insurance, taxes, and more.57 India's Aadhaar, with a 93% adoption rate, ties biometric information to a 12-digit pin and has been used for offline and online purposes like COVID-19 vaccinations and contact tracing, as well as banking and financial services-and it is now being adopted by Sri Lanka too.58, 59, 60, 61, 62 And the European Union plans to roll out a digital identity system in 2023.63 Soon enough, it won't just be early adopters creating digital identities. Businesses may soon be required to do so to integrate with state-level programs.

Accenture Technology Vision 2023 - Page 32

The Modular Open Source Identity Platform (MOSIP) is an example of a centralized effort. It is based off India's Aadhaar system, but is completely open source and modular, meant to fit various countries' requirements. Governments can leverage the code as a starting point to build digital identity systems which they then have complete control over. MOSIP is intended to help countries-particularly in regions where physical institutions are weak (or lacking entirely)-fast track into the digital era by implementing identity capabilities built digitally from the ground up. It is currently being used in Ethiopia and the Philippines, and there is a planned inter-country effort across West Africa too.68,69

Accenture Technology Vision 2023 - Page 32-33

The other, increasingly popular, approach to core identity has been to create decentralized, or distributed, platforms. These efforts, sometimes known as Self-Sovereign Identity (SSI), are based on blockchain and distributed ledger technology, so rather than one central authority, they rely on consensus mechanisms from multiple parties to validate identity. In late 2022, the World Wide Web Consortium (W3C) published a specification for Decentralized ID (DID), signifying an important step forward for decentralized identity efforts.70

Accenture Technology Vision 2023 - Page 33

Early efforts are already demonstrating how decentralized core identities could radically transform our experiences on the web. For example, Microsoft's ION is an identification service that runs on the Bitcoin network. Users of ION can replace email and password logins with a unique identifier authenticated by blockchain.72 Any other personal data exists off-chain, making this purely a core identity solution. And Web3 wallets, like MetaMask and Trust Wallet, are increasingly being positioned as core IDs as well. They provide users with a unique non-custodial ID (the individual holds the private key to the ID), which can be used to access a whole slew of next-generation digital services, like storing NFTs, transferring cryptocurrency, or interacting with distributed apps (dApps).73

Accenture Technology Vision 2023 - Page 33

Importantly though, enterprises don't have to wait for tomorrow's standards or apps to start benefiting from this technology. The government of British Columbia (BC) built OrgBook BC, a blockchainbased publicly searchable directory for the 1.4 million organizations registered in the province.74 Countrywide every year, Canadian companies waste an estimated C$10 billion on administrative red tape, like vetting permits.75 Now, with OrgBook, a unique digital identity is created for every registered business in BC, which is then associated with the various licenses and permits the business has qualified for. This allows OrgBook BC's publicfacing website to automatically validate a company's ID and credentials against the blockchain every time its name appears in a search, providing users a trusted record of companies' registration status and selected licenses and permits.

Accenture Technology Vision 2023 - Page 35

Take Starbucks, which is exploring how digital wallets can transform their loyalty program with a new NFT effort called Starbucks Odyssey. Based on a proof-of-stake blockchain built by Polygon, Starbucks customers will be able to collect unique tokens, store them in digital wallets, and exchange them for benefits like virtual classes, unique merchandise, or a trip to one of the company's coffee farms.76 The customer owns their tokens outright, so if they don't want one of the benefits, they are free to sell it on Odyssey's open marketplace to another customer. It transforms the standard loyalty program into a community in which the customer is also a stakeholder and has much more control over what they own. Starbucks knows its pilot is experimental but believes it's an important first step to "[potentially] create an expanded, shared-ownership model for loyalty"-and to build upon in future collaborations as well.

Accenture Technology Vision 2023 - Page 36

Permission.io and Starbucks are leveraging tokenization, one of the leading ways enterprises are starting to innovate around identity. Tokenization is the process of taking a "thing" and creating an associated digital asset, often stored on a blockchain. Importantly, these examples also show how identity innovation isn't limited to people. Starbucks and Permission.io are tokenizing loyalty and attention. The "thing" can be any asset from a shipping container to a piece of art or f inancial securities.

Accenture Technology Vision 2023 - Page 36

Unilever and SAP are using a blockchain system where tokens are created at the source for batches of palm oil.80 The tokens allow Unilever to trace specific crops through their supply chain and connect them directly back to the farms they were sourced from. The Industrial and Commercial Bank of China (ICBC) created Icago, a tokenization system that allows public transit operators to issue NFT "carbon credits" to riders, who can then exchange those tokens for China's central bank digital currency.81 And Goldman Sachs is building an end-to-end tokenization solution for assets in its portfolio.

Accenture Technology Vision 2023 - Page 37-38

Onyx, an identity solution being developed by JPMorgan, for instance, could be used as a way to credential people across the metaverse, Web3, and decentralized finance.92 The solution will allow people to bind digital assets to a decentralized ID and then pick and choose which data they want to share to access services across the web-such as opting to use one's credit score to leverage a "buy now, pay later" option.93

Accenture Technology Vision 2023 - Page 44

As businesses and governments around the world continue to invest in digital transformation, these windows are popping up everywhere. Look at the airline industry. For years, industry-wide data on flying was scattered. But Cirium, an aviation-data marketplace, figured transparency could improve the way the industry is run. The company now quantifies and aggregates aviation data from 97% of scheduled flights worldwide, so customers can track an aircraft's usage and part wear, see real-time f light schedules of about 880 airlines, and even find a specific flight's CO2 emissions.111, 112, 113

Accenture Technology Vision 2023 - Page 44-45

Similarly, due to countless shippers, forwarding agents, and transportation companies behind the scenes, the freight industry has long struggled with black-box costs and price volatility. But the company Xeneta tackled this problem by aggregating millions of data points on ocean and air freight rates as well as shipping lanes. Xeneta today analyzes more than 300 million freight rates, 160,000 port-to-port lanes, and 40,000 airport-to-airport connections.114 It uses that data to define benchmark rates at market value and shares those rates on its platform. Overcharging is easier to spot and stop as a result. Associated British Foods, for example, used Xeneta's benchmark rates to negotiate a 30% decrease in their shipping spending.115 With a clear window into the industry- one that anyone can see through-stakeholders can leverage this information to make their operations more efficient.

Accenture Technology Vision 2023 - Page 47

Spotify demonstrates another way to attract customers through information sharing. First introduced in 2016, Spotify Wrapped is a personalized summary of the music, artists, podcasts, and genres that a user listened to in the past year.124 By packaging and presenting personalized data from an individual user back to them, it not only allows users to reflect on the year and learn about their listening habits, but it is also designed to be shared on social media, sparking engagement and surfacing content-with Spotify at the center. In 2021, Spotify Wrapped received mentions in 1.2 million Twitter posts, and during the week of its campaign, downloads of Spotify's mobile app increased 21%.

Accenture Technology Vision 2023 - Page 49

From our health to our home, sensing technologies increasingly play an integral role in our daily lives. The new Apple Watch can calculate sleep cycles down to the minute and track heartbeat irregularities

Accenture Technology Vision 2023 - Page 49

And in Sweden, researchers built a low-cost nano-sensor that can detect pesticides on fruit in a matter of minutes.12

Accenture Technology Vision 2023 - Page 49

While each of these new sensors operates on a small scale, they give clear, quantifiable lines of sight into increasingly larger systems. One research effort demonstrated how a low-cost microsensor can recognize and locate the distinct sound of a leaky water pipe with 100% accuracy.129

Accenture Technology Vision 2023 - Page 50

University of Washington created a self-powering sensor that can float in the air like a dandelion seed and measure the temperature, humidity, and other details of the surrounding environment.130

Accenture Technology Vision 2023 - Page 50

Take Planet Labs. Every day, satellites from Planet Labs capture 300 million square kilometers of images from space at a 3.7-meter resolution-a clarity so high that the naked eye can see individual animals.132,133 The company partnered with King Abdullah University in Saudi Arabia to capture daily, cloud-free images from space of a Nebraska corn field. The images showed the crops' health, which helped farms tailor field irrigation, and better predict crop yields.134,135

Accenture Technology Vision 2023 - Page 50

For example, T-Mobile has been building out a 5G network and they are thinking bigger than connectivity just for phones. The company partnered with Pano AI and Portland General Electric to deploy 5G-connected cameras in remote, wildfire vulnerable areas. Pano AI placed its artificial intelligence enabled cameras in rural locations, then used T-Mobile's 5G network to quickly transmit the data.136 This lets users access real-time observations and data-and has resulted in users reporting fires when they start and enabling authorities to respond more quickly.

Accenture Technology Vision 2023 - Page 51

Recently, for example, the restaurant reservation company OpenTable used Clarifai's auto-tagging feature to help employees moderate content in images uploaded to OpenTable's platform. After integrating this auto-tagging, OpenTable noticed its content moderation became 16 times more productive; on average, each moderator jumped from moderating 300 to 5,000 photos a day.140

Accenture Technology Vision 2023 - Page 52

Another technology that can bolster your data strategy is a knowledge graph, which links related data points and visualizes those links. These graphs often include semantic layers, or text descriptions of the data, which make it easier to find what you need. Stardog partnered with NASA to build a knowledge graph out of various data sources about the manufacturing of NASA's Space Launch System rocket. So, instead of NASA engineers manually pulling in data, they could query a knowledge graph to get the data they sought, resulting in a ten-fold increase in their productivity.141

Accenture Technology Vision 2023 - Page 52

Importantly, emerging tools are allowing non engineers to tap into technologies in the data management toolbox as well, allowing your enterprise to build a more data-forward company culture. For instance, self-service analytics can let employees, regardless of their technical background, explore your enterprise's data to find insights. This can both empower workers and help solve business problems.

Accenture Technology Vision 2023 - Page 52

Take the Seattle Seahawks, a professional American football team.142 For years, the National Football League (NFL) Voice of the Fan survey found that some Seahawks supporters had a hard time hearing games in the team's stadium. But the team's management did not know where those fans sat to fix the game's speakers there. So, the Seahawks turned to Tableau's heat mapping tool. After inputting their Voice of the Fan survey data, four corners of the team's stadium lit up in Tableau's visualization, showing where those fans who complained of poor sound quality sat. The Seahawks could reconfigure the stadium's audio system without buying a new one, saving money and improving fan experiences

Accenture Technology Vision 2023 - Page 52-53

With data fabrics, the goal is to defragment an enterprise's data architecture by leveraging automation and metadata to build a single source of truth across disparate data sources. It is a top down approach that installs a virtual layer on top of an organization's various data sources for tightly controlled, unified, data management. Metaphorically speaking, it is like building a zoo around your data. You aren't changing the underlying data, and in fact you're trying to preserve the environments-but at the same time you're building consistency across the enclosures, better descriptions about what's inside, and a map to where everything is so it's all readily accessible by everyone.

Accenture Technology Vision 2023 - Page 53

Data mesh is another emerging concept which tackles the problem in a different way. If data fabrics try to assert top-down management over data, data mesh treats each data source as an independent product. Coined by Zhamak Dehghani in 2019, data mesh stresses domain-specific expertise, management, and governance.145 The owners of data sources are distributed and entrusted to manage data themselves, with the expectation that they are trying to reduce the friction to access that data (typically through the use of APIs).

Accenture Technology Vision 2023 - Page 55

For instance, Bentley Systems' platform for building digital twins of critical infrastructure systems, iTwin, now includes iTwin Experience, which acts as a "single pane of glass" to integrate and overlay engineering, operations, and information technology data.146 This allows owners and operators to more easily visualize, query, and analyze digital twins at different levels of granularity and scale, effectively empowering them to make better informed, more actionable decisions.

Accenture Technology Vision 2023 - Page 55

To help building owners, Measurabl and Singularity Energy launched a tool that quantifies and monitors buildings' CO2 emissions.148 This tool can determine a building's compliance with local regulations, give building owners a financial risk report based on their carbon emissions, and create a decarbonization plan per building. Making carbon emissions visible will not only ensure a company's compliance and commitment to sustainability, but overall, it will take a city one step closer to being cleaner and healthier

Accenture Technology Vision 2023 - Page 56

For example, The Washington Post launched its personalized annual year-in-review in 2022, summarizing a subscriber's reading habits to reveal their unique reading "personality."150 It reminds readers of all the content they've enjoyed and the value they got from The Post-and the reader's unique "Newsprint" looks like a fingerprint, ready to be shared on social media. And other companies provide year-end data back to their customers in aggregate. For instance, according to Strava's 2022 Year In Sport, cyclists riding in groups rode both farther and faster, on average, than those who rode solo-hard proof that it really is beneficial to have a training partner.151

Accenture Technology Vision 2023 - Page 57

For example, as more enterprises begin to look beyond the limits of our planet and up towards space, the world is quickly running into the problem of "space junk." Debris from satellites, rocket launches, and more are starting to crowd the areas around our planet creating all manner of difficulty, from obstructing astronomers to interfering with future rocket launches. Normally, no enterprise would treat this as "their" problem, but Privateer, a space startup, decided to shed light on it.154 The company built a publicly accessible object tracking platform to trace the orbits and velocity of all the objects crowding our planet. The hope is that by quantifying this problem, others can start working on a solution-or at least stop making the problem worse in the near term.

Accenture Technology Vision 2023 - Page 64

One company taking advantage of this is CarMax. CarMax is using GPT-3, a large language model and predecessor to GPT-4, to improve the car-buying experience.174 Knowing that there's a massive amount of information potential car buyers may want to read through before making a purchase decision, CarMax used Microsoft's Azure OpenAI Service to access a pretrained GPT-3 model to quickly read and synthesize over 100,000 customer reviews for every vehicle make, model, and year that they sell. From these reviews, the model generated 5,000 easy to-read summaries-a task that CarMax said would have taken its editorial team 11 years to complete.

Accenture Technology Vision 2023 - Page 64

Other organizations are also experimenting with foundation models, adapting them for tasks ranging from powering customer service bots, to generative product design, to automated coding. And the models are advancing fast, with companies across industries quickly discovering new ways to use them. As foundation models broaden and extend what we can do with AI, they are letting companies transform human-AI interaction and build an entirely new generation of AI applications and services.

Accenture Technology Vision 2023 - Page 66

DeepMind's Gato is one of the most exciting examples to date. The company calls Gato a "generalist agent" because it is multimodal and can complete over 600 different tasks.181,182 Using a single AI model with f ixed weights, it can chat, caption images, play Atari video games, stack blocks with a robotic arm, and more. Additionally, it can learn these various tasks simultaneously and switch between them without having to forget previous skills. For context, AlphaZero- an older DeepMind model known for playing chess, Go, and shogi-had to unlearn how to play chess in order to play Go.

Accenture Technology Vision 2023 - Page 67

transformer models, introduced by Google researchers in 2017.183 One of the newest classes of AI models, transformers are neural networks that identify and track relationships in sequential data (like the words in a sentence), to learn how they depend on and influence each other. They are typically trained via self-supervised learning, which for a large language model could mean pouring through billions of blocks of text, hiding words from itself, guessing what they are based on surrounding context, and repeating until it can predict those words with high accuracy.184 This technique works well for other types of sequential data too: some multimodal text-to-image generators work by predicting clusters of pixels based on their surroundings

Accenture Technology Vision 2023 - Page 68

And a research team from the University of Texas at Austin, the Indian Institute of Technology Madras, and Google Research, proposed Generalizable NeRF Transformer (GNT), a transformer-based architecture for NeRF reconstruction.191, 192 A NeRF (Neural Radiance Field) is a neural network that can generate 3D scenes based on only partial 2D views-and experimenting with transformers to generate 3D data like this could have big metaverse implications

Accenture Technology Vision 2023 - Page 72

On the flip side, as companies develop new foundation models, they may also run out of training data. Most language and image models rely on huge amounts of web data, and organizations' ability to increase model size depends on finding new data sources. Additionally, developing foundation models for new modalities may require huge amounts of hard-to-obtain data. Synthetic data, perhaps even from early foundation models, could be a powerful tool for building these new models

Accenture Technology Vision 2023 - Page 73

Multimodal foundation models' ability to recognize multiple data types and identify the relationships between them is also pushing the envelope of what AI is capable of, enabling powerful new systems. GPT-4, for example, is multimodal and accepts both image and text inputs, meaning that if someone were to show it a picture of the inside of their refrigerator, it could correctly identify the items inside, suggest meals that can be made with those ingredients, and then provide step-by-step cooking instructions.206

Accenture Technology Vision 2023 - Page 73

And Meta has long seen the value of an AI system that can interpret content on its platform-especially when it comes to detecting hate speech. But this is a task that's historically been difficult for machines because people tend to communicate in multimodal ways on these platforms (using text and image together to tell a joke, for instance). So Meta has launched a series of foundation and multimodal AI projects to help them analyze different types of communication-like text, image, and video-in conjunction. The company created the Hateful Memes dataset to address the shortage of publicly available training data for classifying memes; it developed FLAVA, a multimodal foundation model that works across dozens of tasks; and it built Omnivore, a model that can operate across images, video, and 3D data, doing things like detecting content in both videos and images.207, 208

Accenture Technology Vision 2023 - Page 90

Quantum computers have a "natural" advantage (over classical computers) in simulating quantum mechanics, which governs the behavior of molecules, atoms, and electrons. As such, it is contributing to the field of chemistry, in perhaps the nearest-term application of quantum computers.248 Though computational speedups may come with quantum computers, what they provide in this case is an advantageous level of accuracy in modeling distinct parts of a chemical reaction- and this enhanced understanding has many potential applications.

Accenture Technology Vision 2023 - Page 90

For instance, Hyundai is partnering with the quantum computing startup IonQ to analyze and simulate battery materials-in this case lithium oxide, contained in lithium-air batteries.249 Using hybrid algorithms that leverage computing by both classical and quantum computers, they can improve the chemical makeup for greater efficiency and eliminate possible sources of waste.

Accenture Technology Vision 2023 - Page 90

Building supercomputers and quantum computers is not cheap, but experimenting with them is significantly cheaper than before, thanks to cloud platforms. Firefly Aerospace, for instance, is a startup that relies on cloud supercomputing to do advanced simulations to save massive amounts of money on prototyping, enabling them to build a rocket to go to the moon.250 And while quantum computing still needs time to mature, it's clear that these computers have a major role to play helping companies effectively drive science technology feedback loops and accelerate materials and energy innovation.

Accenture Technology Vision 2023 - Page 93

Lastly, this "big bang" of computing and science is giving life to an entirely new field: synthetic biology. Driven forward by technological advances in DNA sequencing and synthesis and technology inspired best practices, synthetic biology combines engineering principles with biology to create new organisms or enhance existing ones. Its promise lies in what those organisms are then able to do or produce, from new foods to pharmaceuticals, fuels, and more-thereby changing manufacturing processes and products as we know them today.

Accenture Technology Vision 2023 - Page 94

Though a relatively new field, synthetic biology is being used already. It's well-known for aiding in the development of the COVID-19 vaccine, but it's also used by Bolt Threads to make bio-based leather and Upside Foods to produce synthetic meat like duck, beef, and chicken.275, 276, 277 Synthetic biology techniques can also be used to make dyes, cosmetics, and spices. Not only are these new and innovative products, but they are more environmentally friendly and sustainable than their alternatives

Accenture Technology Vision 2023 - Page 96

Moving on to climate change, of all the challenges the world faces today, this may be the biggest. Sustainability is certainly on most businesses' radars, and all of the areas we've discussed, from materials and energy to space innovations and synthetic biology, have the potential to contribute to sustainability in a significant way. For example, Accenture, the Irish Centre for High-End Computing ... of all the challenges the world faces today, climate change may be the biggest. (ICHEC), and IonQ collaborated to create a scalable software platform for chemistry simulation on quantum computers. It is used to calculate the energy needed to break chemical bonds in molecules like PFAS-helping to find mechanisms to destroy PFAS, which are human-made carcinogenic "forever chemicals" that pollute the environment.281

Accenture Technology Vision 2023 - Page 96

And LanzaTech is using synthetic biology to make microbes that take carbon emissions and turn them into valuable raw materials like fuels and chemicals.282

Accenture Technology Vision 2024 - Page 06

Consider the impact generative AI and transformer models are having on the world around us. What began as chatbots like ChatGPT and Bard has become a driving force in making technology more intuitive, intelligent, and accessible to all. One example is Adobe Photoshop's Generative Fill and Generative Expand features, powered by Adobe Firefly.2 These innovations let anyone add, expand, or remove content from images non-destructively, using simple text prompts. Users can now experiment with their ideas, ideate around different concepts, and produce dozens of variations faster than ever before. Where AI once focused on automation and routine tasks, it's now shifting to augmentation, changing how people approach work, and is rapidly democratizing the technologies and specialized knowledge work that were once reserved for the highly trained or deep-pocketed.

Accenture Technology Vision 2024 - Page 07

And generative AI has the potential to impact much more than just the task at hand. It's also starting to profoundly reshape organizations and markets. Google Cloud, for instance, recently announced a generative AI search tool meant to help doctors and nurses rapidly find patient information that may be stored across multiple systems and in different formats - a major challenge that has plagued healthcare systems for years.3 FrameDiff, a generative AI computational tool created by MIT CSAIL researchers, is crafting synthetic protein structures that go beyond what exists in nature to open new possibilities in drug development.4 And for software companies, tools like GitHub's Copilot (a generative AI assistant that helps write code) are demonstrating potential to make software engineers more satisfied with their jobs.5 In fact, in many cases generative AI tools are so intuitive to use and employees are adopting them so rapidly, they are permeating workplaces from the bottom up - faster than organizations can create formal programs.

Accenture Technology Vision 2024 - Page 08

To solve challenges innate to digital work, like video fatigue, Microsoft made major updates to Microsoft Mesh, their platform for creating immersive spaces that blend digital and physical.6 The company is trying to use immersion to solve pain points today, as well as drive new collaborative ways of working. Recognizing the importance social media plays in many people's lives, but also the friction it can create, social media newcomers

Accenture Technology Vision 2024 - Page 08

Boston Dynamics has long been at the forefront of making robotics more human, promising a smoother integration between robotics and the world around us. For instance, their bipedal robot Atlas has been trained on diverse tasks, allowing it to mimic human movement and physical intuition.7 What's more is these robots no longer just mirror humans physically, but socially too. Humans will usually interact with robots through complex lines of code and puzzling machine logic, leaving an impasse between people who don't speak that language and the robots next to them. However, researchers found a way to put ChatGPT onboard a Boston Dynamics robot, allowing people to use natural language to command the robot or ask it about its previous tasks and receive a clear response in plain English.

Accenture Technology Vision 2024 - Page 16

Data is one of the most important factors shaping today's digital businesses. And these chatbots - that can synthesize vast amounts of information to provide answers and advice, use different data modalities, remember prior conversations, and even suggest what to ask next - are disrupting that undercurrent. Many long-standing enterprise functions like digital marketing, advertising, and product discovery all stand to change.

Accenture Technology Vision 2024 - Page 17

It started in November 2022 when OpenAI launched ChatGPT, which became the fastest growing app of all time.20 Large language models (LLMs) had been around for years, but ChatGPT's ability to answer questions in a direct and conversational manner made the difference. Microsoft, OpenAI's partner, quickly recognized what was happening and released Bing Chat in February 2023.21 They positioned it not just as a search tool but as a "copilot for the web." The next month, Salesforce announced Einstein GPT - a generative AI CRM technology that leverages all the customer data a company stores in Salesforce applications to generate content like personalized marketing materials or knowledge articles.22 By June, the electronic health record software company Epic was integrating GPT-4 into its products to allow clinicians to speedily generate summaries of patient charts.23

Accenture Technology Vision 2024 - Page 17

How people access and interact with data is changing. Fast. And this change is a code red - not just for traditional search companies, but for every company. It's not just web search engines but search in the very broadest sense that is being impacted by generative AI - everything from searching for an email or a file to looking up customer details in a CRM database.

Accenture Technology Vision 2024 - Page 18

Every enterprise has an information strategy. It's defined through the software they use, how they market information externally, and how they use it internally. Search, broadly speaking, has underpinned these strategies for decades. But every day, consciously or not, more people shift from search to asking, and more companies look to meet them where they are, using generative AI chatbots as the UI for enterprise platforms and a new access point for customers.

Accenture Technology Vision 2024 - Page 19

It makes sense that the shift from search to asking is so alluring. It's shaking things up in a way enterprises desperately need. Putting an LLM-advisor with the breadth of enterprise knowledge at every employee's fingertips could unlock the latent value of data and finally let enterprises tap into the promise of data-driven business. Of course, it's not as simple as turning the knob to something new. Generative AI will be the interface that rests on top of enterprises' vast data architectures, but if businesses want to capture its many benefits and transform how people access their information, they need to finally get that foundation right.

Accenture Technology Vision 2024 - Page 20

To demonstrate just how powerful a knowledge graph and semantic search can be, consider a use case from Cisco Systems. Like many large global enterprises, Cisco's sales team had tons of content to leverage. But they struggled to find relevant documents through their index-driven search due to a lack of metadata. So, they turned to Neo4j to help create a metadata knowledge graph. While they did not use LLMs, they relied on natural language processing to create an ontology and a machine tagging service to assign document metadata, which was then stored in a graph database. Now, finding information takes half the time, and Cisco saves its salespeople over four million

Accenture Technology Vision 2024 - Page 21

First, companies can train their own LLM from scratch, though this approach is rare given the significant resources required. Some of the leaders here are AI powerhouses, including Amazon, OpenAI, Google, Meta, AI21, and Anthropic. Bloomberg also took this approach, using its own massive knowledge base of financial data, along with a public dataset, to train a 50-billion parameter LLM for the financial industry called BloombergGPT.30 It will be made available to customers on the Bloomberg Terminal.31 For large companies with vast resources, however, self-training an LLM from scratch may be an appealing approach to secure a competitive advantage.

Accenture Technology Vision 2024 - Page 21

A slight variation on this is also gaining traction. Enterprises are beginning to fine-tune smaller language models (SLMs) for specialized use cases. SLMs like DeepMind's Chinchilla and Stanford's Alpaca have started to rival larger models while requiring only a fraction of the computing resources.36 These SLMs are not only more efficient, running at lower cost with smaller carbon footprints, but they can be trained more quickly and used on smaller, edge devices.

Accenture Technology Vision 2024 - Page 22

Lastly, one of the most popular approaches to building an LLM-advisor has been to "ground" pre-trained LLMs by providing them with more relevant, use case-specific information, typically through retrieval augmented generation (RAG). As suggested by the name, this combines an information retrieval system with a generative model, which can be either self-trained or used out-of-the-box and accessed through an API. At a high level, RAG works something like this: First, a user will type in their request. Next, that input is used to search for and retrieve relevant documents - whether unstructured data like text from Word documents, chats, or PDFs, or structured data like CSVs or database tables - as vector embeddings. Then, these documents, along with a prompt, are sent to the LLM. The LLM is, of course, trained on a huge amount of data initially, but only uses the specific information it receives to generate its response to the user.

Accenture Technology Vision 2024 - Page 22

Grounding an LLM through in-context learning and RAG takes much less time and compute power, and furthermore, requires far less expertise than training LLMs from scratch or fine-tuning. In fact, this approach is built into Microsoft 365 Copilot, an AI assistant for Microsoft 365 applications and services.37 And Salesforce's Einstein GPT uses this approach to ground generative AI chatbot responses too, when connected to one of OpenAI's LLMs or any other external LLM.38 This option works best for use cases that require up-to-date information, though verifying for accuracy may still be necessary.

Accenture Technology Vision 2024 - Page 23 - 24

First and most importantly, as businesses begin to explore the new possibilities LLM advisors bring, they need to understand the associated risks. In March 2023, a lawyer submitted a brief to a New York judge. In it, he cited multiple prior court decisions, indicating that his client's case should not be dismissed.39 But there was a problem. None of those court decisions, or related quotations and citations, could be found - ChatGPT had created fictitious cases. According to the lawyer, he was not aware that ChatGPT could fabricate information, but the judge was not pleased. Not only did the judge fine that lawyer and others involved, but he required them to notify the real judges who were identified as having written the fake cases.40 Most embarrassing: the colossal mistake wound up on the front page of The New York Times.41 What this lawyer experienced was an almost intrinsic characteristic of LLMs: "hallucinations." Because LLMs are trained to deliver probabilistic answers with a high degree of certainty, there are times when these advisors confidently relay incorrect information. And as LLM applications start to take a bigger share of how we access and relay information, or interact with and integrate software, there can be serious consequences. Any way you slice it, when you don't know if what you're reading is true, that's a major issue

Accenture Technology Vision 2024 - Page 25

Morgan Stanley, for instance, has a vast internal knowledge library, including hundreds of thousands of documents ranging from investment strategies to market research and other insights.44 These documents can be found across various internal sites, mostly in PDF form, so it can require significant time and energy for advisors to scan those documents and find the answers they're looking for. Now, however, with the help of GPT-4, Morgan Stanley has created a generative AI chatbot that can harness this wealth of internal knowledge and help advisors get the insights they need, instantly

Accenture Technology Vision 2024 - Page 25

Toyota is another company leveraging generative AI to scour vehicle owner's manuals and provide more direct answers to people's car questions.45 Currently, they have a proof-of-concept that allows a driver to ask aloud a question like, "How do I disable the VSC?" The Toyota AI will respond with clear instructions, as well as with the pages in the manual where the driver can find the answer. In addition, research shows that customer service workers can benefit from generative AI chatbot assistance too, likely through the dissemination of more tacit knowledge.46 The findings showed that the AI assistant not only helped workers become more productive, resolving 14% more issues per hour on average, but it also improved customer sentiment and employee retention, while reducing requests for the manager

Accenture Technology Vision 2024 - Page 26

Some companies are looking at plugins to give explicit access to external data and improve the outputs of generative AI chatbots like ChatGPT and Bard. For instance, Edmunds. com and CarGurus.com - websites providing car inventory, pricing, and reviews - both launched ChatGPT plugins to help prospective car-buyers.51 This way, customers can get up-to-date information and explore cars in their own terms and language, without being constrained by the limited search fields. Today, companies with Bard plugins include Redfin, Instacart, and Spotify.52

Accenture Technology Vision 2024 - Page 31

It might sound futuristic, but it could happen sooner than you think. Already, enterprises are embedding AI across business operations. Generative AI has transformed industry leading creative tools at Adobe and propelled product ideation at Volkswagen.16,17 Siemens and Fanuc have reimagined manufacturing by embedding AI across robotics and industrial processes.18,19

Accenture Technology Vision 2024 - Page 31

A useful analogy for the progression of AI agents is the advancement of self-driving cars. For many years, drivers were entirely responsible for the operation of the vehicle (no AI). But then semi-automated systems like cruise control or lane assist came into play (AI that assists). After that, automated driving became available to drivers in limited conditions, and then fully self-driving cars requiring no driver at all (agents with increasing action). And if you extrapolate this trend, we can imagine a future with self-driving cars that all work together on the road (an ecosystem of agents). For cars, these advances have not come as precise step changes but as progress on a continuum. The evolution of AI agents will be the same

Accenture Technology Vision 2024 - Page 31

Look at DoNotPay, a company designed to help consumers save money - from contesting parking tickets to identifying unused subscriptions. Until recently, DoNotPay identified these issues and prompted customers to take action - but then the company integrated GPT-4 and AutoGPT into its software.21 The first user of these new features was DoNotPay's CEO. He gave the agent access to his financial accounts and prompted it with a concise yet complex task: f ind me money. The agent identified $81 in unnecessary subscriptions and an unusual $37 in-flight Wi-Fi fee. Then, it offered to automatically send cancellations to the subscription providers, drafted a letter to contest the Wi-Fi charges, and checked in with the CEO for review. As icing on the cake, it even drafted and sent emails that negotiated a 20% reduction in the CEO's cable and internet bill

Accenture Technology Vision 2024 - Page 32

AI assistants are maturing into proxies that can act on our behalf. As these agents emerge, the resulting business opportunities will depend on three core capabilities: access to real time data and services; reasoning through complex chains of thought; and the creation of tools - not for human use, but for the use of the agents themselves. Along with humans to guide and oversee them, these advancements will allow agent ecosystems to complete tasks in both the physical and digital worlds, generating immense value for every enterprise that takes part - and risking obsolescence for those who don't.

Accenture Technology Vision 2024 - Page 34

In March 2023, OpenAI announced the first plugins for ChatGPT. "Plugins" allow LLMs to look up information, use digital software, execute code, call APIs, and generate outputs beyond text by allowing the model to access the internet. Instead of relying solely on the weights and tokens that make up the model's intelligence, ChatGPT can now search Expedia to get travel information, access Instacart for ordering groceries, and engage Wolfram (a computational intelligence platform) to perform complex mathematical calculations.23 After just a few months, ChatGPT had access to hundreds of plugins.24 By the time you read this, those numbers may be higher.

Accenture Technology Vision 2024 - Page 35

AutoGPT and BabyAGI are two open-source applications that leverage LLMs and automate chain-of-thought prompting. These applications will take broad queries or instructions, and then prompt themselves to think through the steps and ways to accomplish their goals, articulating for themselves a detailed set of instructions that they will then use to accomplish the original ask.28,2

Accenture Technology Vision 2024 - Page 35

Take Nvidia, which along with researchers from several universities, explored the possibility of developing an "embodied agent." They built their agent - Voyager - in Minecraft, a popular game about survival and exploration that takes place in a 3D world.30 To navigate this world, players acquire resources that allow them to forge new tools, like a pickaxe or lantern, which let them further traverse and shape their environment. Voyager was given the instruction to explore, and it was equipped with a skill library it could add to over time. As Voyager met new barriers, it would learn which tools were needed to overcome the obstacle, then store that information in its library. When encountering further obstacles, it would increasingly draw from its skill

Accenture Technology Vision 2024 - Page 36

Google's PaLM-E can take a command in natural language, break it down into a series of subtasks, then generate and execute commands to control physical robots.36 It's not difficult to imagine such an agent leading an entire manufacturing plant. And MetaGPT can automate an entire software development stream by acting as a product manager, architect, project manager, and engineer all rolled into one, delegating tasks to its array of GPTs. From one line of text, MetaGPT can generate user stories, competitive analyses, requirements, data structures, APIs, documents, and beyond.37 The agent ecosystem may seem overwhelming. After all, beyond the three core capabilities of autonomous agents, we're also talking about an incredibly complex orchestration challenge, and a massive reinvention of your human workforce to make it all possible. It's enough to leave leaders wondering where to start.

Accenture Technology Vision 2024 - Page 39

Investment research company Morningstar has successfully focused its GPT-3.5-embedded chatbot "Mo" on relevant proprietary information by providing such a support system.39 Prompt-tuned on more than 10,000 pieces of proprietary research, Mo serves as an advisor to Morningstar's financial advisors and customers - and it's able to do this because Morningstar's human workforce set the stage in the background. Specifically, one Morningstar team created rules for what Mo can and can't answer, and Morningstar's lawyers ensured that none of Mo's capabilities violated ethical or regulatory bounds.40 Morningstar proactively deployed their humans to put bounds on Mo because reactive trial-and-error isn't an option when you're dispensing financial advice.

Accenture Technology Vision 2024 - Page 47

Meta has been rapidly developing its Reality Labs VR and AR products, and introduced Codex Avatars, which use AI and smartphone cameras to create photorealistic avatars.23,24 Epic's RealityScan App lets people scan 3D objects in the physical world with just their phone and turn them into 3D virtual assets.25 Underlying it all, advancing technologies like generative AI continue to make it faster and cheaper to build spatial environments and experiences.

Accenture Technology Vision 2024 - Page 48

Google met the moment by unveiling a mobile Google Maps, which drew on real-time user data to refine its accuracy at staggering speed. Now, nearly wherever you and your phone go, you can get from Point A to Point B. More than one billion people today use mobile Google Maps.

Accenture Technology Vision 2024 - Page 49

Successful spatial computing deployments in industrial settings have shown it can be used to better convey massive amounts of complex information by tapping into multiple senses and communication avenues at once. Other experiments have found that when we see applications as "spaces," we can mold experiences to the individual's environment and gestures, or give them freedom to self-direct. Mobile and desktop users, in contrast, could only click or swipe where the design let them. And with spatial computing able to augment our physical

Accenture Technology Vision 2024 - Page 51

Enter Universal Scene Description (USD), or what can best be described as a file format for 3D spaces. Developed by Pixar, USD is a framework that lets creators map out aspects of a scene, including specific assets and backgrounds, lighting, characters, and more. Since USD is designed around bringing these assets together in a scene, different software can be used across each one, enabling collaborative content building and nondestructive editing.27 USD might sound like its main use is in entertainment applications, but it is quickly becoming central to the most impactful spatial applications, notably within industrial digital twins

Accenture Technology Vision 2024 - Page 51

With some companies working on making spatial worlds look and feel similar, others are focused on standardizing how we access those experiences. OpenXR is an open API standard that has now been adopted by most major device manufacturers.31 It ensures applications can guarantee uniformity in head and hand position, controls, visual display, and more across most devices by using a single API (rather than testing and designing for all of them independently).

Accenture Technology Vision 2024 - Page 52

Some are already taking this into account. The Open Metaverse Alliance for Web3 (OMA3) is building a standard for how we move across experiences. Today, if you want to move from one metaverse world to another, you need to quit one application and move to the next, almost like quitting and relaunching your browser every time you want to go to a new website. In 2023, OMA3 launched a project called the Inter-World Portaling System, an effort to develop a protocol that would let developers move users from one space to another without breaking the immersion, like how an address bar sits at the top of whatever website you visit.32

Accenture Technology Vision 2024 - Page 52

Generative AI is speeding up the creation of 3D digital content. Take Nvidia's Neuralangelo, a generative neural network. It can rapidly turn 2D video clips into 3D digital objects or structures.34 Those assets can be imported into VR or AR spaces, digital twins, or video games. And Intel's LDM3D AI model can churn out 360-degree 3D images from simple text prompts.35

Accenture Technology Vision 2024 - Page 52

Other examples are cutting out the need to create a 3D model in the first place. Qualcomm's Snapdragon Spaces is leaning into realism by letting people blend digital content with their physical environment. The AR SDK is allowing developers to rapidly create new applications that blend the digital and physical by using semantic scene understanding, hand tracking, object detection, and more.36 And Google's TryOnDiffusion uses generative AI to let online shoppers see how clothes would realistically drape or fold over their unique bodies.37

Accenture Technology Vision 2024 - Page 52

Some tools are making it simple enough for anyone to build spatial environments. Google's Geospatial Creator, powered by the Google Maps Platform and ARCore (Google's AR Software Development Kit), lets you create an immersive experience in just minutes. You can build 3D digital content to layer onto real-world locations and it integrates with design engines like Unity and Adobe Aero.38 The best part? Little to no coding skills are needed.

Accenture Technology Vision 2024 - Page 53

This is why part of Meta's training and guides for developers building 3D spaces includes proxemics as a topic.39 Proxemics is the study of how we use space, including how population density or physical proximity impacts how we act, communicate, and relate to each other. Meta recognizes that spatial awareness is key to making these experiences work. A user's experience in a crowded space may well be an important factor in figuring out when an enterprise needs another instance of a digital store or world to combat overcrowding.

Accenture Technology Vision 2024 - Page 53

Inworld AI points to a new direction. It creates AI characters with personalities, that can communicate verbally as well as non-verbally.41 These characters are context-aware, so they don't hallucinate or refer to content outside their set world, mitigating misinformation risks.42 And they offer a sense of interpersonal realism that will ground the spatial apps of tomorrow

Accenture Technology Vision 2024 - Page 54

In past VR attempts, adding haptics, or touch, could be bulky or underwhelming. But University of Chicago researchers recently proposed using electrodes to better mimic touch.43 They built an electrode system with 11 controllable tactile zones on a person's fingers, so they could "feel" digital content. Imagine a meditation spatial app that took you to a virtual beach where you could "feel" the grains of sand.

Accenture Technology Vision 2024 - Page 54

Scents can make digital spaces lifelike, too, by evoking memories or triggering the allimportant fight-or-flight response. Scentient, a company trying to bring olfactory senses to the metaverse, believes scent can be the key to overcoming the "uncanny valley" - when a scene feels almost believable, but not quite enough, turning users away. Scentient's leaders think scent brings realism and dimension to a digital space, and they have been experimenting with the technology for training firefighters and emergency responders, where smells, like the presence of natural gas, can be critical for evaluating an emergency.44 At first glance, the idea of scent may seem gimmicky, but it can have real impact, and be a key part of a successful space.

Accenture Technology Vision 2024 - Page 55

When it comes to conveying complex information, the advantage of the spatial medium over the alternatives is probably clearest. Since a space can let users move and act naturally, information can be conveyed in more dynamic, immersive ways. We've already seen it in action. Some of the earliest examples of successful spatial apps were industrial digital twins, virtual training scenarios, or real-time remote assistance - all use cases where lots of information may need to be shared, and where conventional methods can cause information overload, leading to confused or ill-trained employees.

Accenture Technology Vision 2024 - Page 56

Surgical training, for instance, is an extremely information-heavy task. Research from the University of California Los Angeles found that medical students performed two and a half times better on surgical skills assessments when they learned new procedures in virtual reality versus the standard guides.49 And importantly, including more spatial information can improve results even further. Haptic feedback - the sensation of touch - helped surgical trainees achieve proficiency in cortical bone drilling roughly seven times more often than those operating without it.50

Accenture Technology Vision 2024 - Page 56

Paris's Centre Pompidou capitalized on this ability to personalize a museum experience through a collaboration with Snapchat and artist Christian Marclay.53 Marclay overlayed Centre Pompidou's fa�ade with a colorful digital instrument that users could play in many ways through Snapchat AR. Visitors could also record and share how they "played" the museum. Rather than an experience exclusively shaped by the museum's curators, visitors were able to infuse their own creativity and discovery into the space.

Accenture Technology Vision 2024 - Page 56

Teleoperating through a spatial environment, however, could help by displaying information more clearly, conveying it in a realistic environment when and where it's needed. Already, a surgeon in China has successfully removed a patient's gallbladder from 4,600 KM away, operating a surgical robot in the distant operating room.51

Accenture Technology Vision 2024 - Page 56-57

And Fiat, the automotive brand, is showing how user self-driven discovery can help with sales conversion.54 Typically, when people are buying a car they are able to test drive one model, but then have to look at a different model or a bunch of advertising pamphlets to see the customization options. They don't necessarily get to drive the exact car they want to buy. Fiat built the Fiat Metaverse Store to challenge this paradigm. Within the virtual store, users are able to customize their car model with all the various body types, colors, interiors, and infotainment options.55 Users can then take the car out on a virtual test drive to experience what it would be like and see the features up close. Throughout the experience they are accompanied by "Product Genius" - a connection to a live expert who can answer any questions

Accenture Technology Vision 2024 - Page 57

Lastly, spatial applications bring advantages to physical spaces; they can augment, enhance, and extend physical places without materially changing them. Imagine a future office where physical monitors, projectors, and displays are replaced by spatial computers and apps. People will have the flexibility to design simpler spaces, lowering overhead costs, and to change their surroundings more easily.

Accenture Technology Vision 2024 - Page 57

Gap and Mattel leveraged this when launching their Barbie clothing line.57 Using Google's Geospatial Creator, they transformed New York City's Times Square with lifelike 3D Barbie dolls, massive digital billboards, and floating neon-pink signs. This launch event showed the spatial medium as a captivating, scalable alternative to building expensive new infrastructure.

Accenture Technology Vision 2024 - Page 58

Accenture, Google, and Telstra, Australia's leading telecommunications firm, also harnessed spatial computing in a pilot app to improve the stadium experience - without significant physical changes.58 In the app, fans could find their seats with immersive wayfinding and enjoy augmented reality experiences.

Accenture Technology Vision 2024 - Page 58

In one cross-industry collaboration at CHRISTUS MUGUERZA Hospital Conchita in Mexico, a renowned orthopedic surgeon was equipped with a Microsoft HoloLens during surgery.59 The technology transformed the surgical environment, allowing the doctor to access patient records, X-rays, scans, and 3D models during a procedure. Importantly, digitally accessing this information mid-surgery let the operating team preserve a sterile field for the patient. When the team needed to refer to patient information, they could do so and overlay it while still in the operating room - without stepping out, removing scrubs, finding the record in question, recalling what they needed to know, changing back into scrubs, and re-sterilizing upon return - so there was no need to delay

Accenture Technology Vision 2024 - Page 63

Look at how neurotech is beginning to connect with people's minds. Recently, two separate studies from researchers at the University of California San Francisco and Stanford University demonstrated using neural prostheses - like brain-computer interfaces (BCI) - to decode speech from neural data.19,20 This could help patients with verbal disabilities "talk" by translating attempted speech into text or generated voices.

Accenture Technology Vision 2024 - Page 63

Or look at technologies that read body movement, like eye and hand tracking. In 2023, Apple's Vision Pro introduced visionOS, which lets users navigate and click with just their gaze and a simple gesture, bypassing the need for a handheld controller.22 The highly precise eye tracking acts as a targeting system; users can pinch their index finger and thumb together to click on what they're looking at. Even more impressive eye tracking could also be coming. Apple has filed a patent that describes using pupil dilation to predict if a user intends to do something - like select a button - even before they do it.

Accenture Technology Vision 2024 - Page 67

Some are using wearable devices to track biosignals that can help predict what people want or understand their cognitive state. For instance, researchers at the University of Washington Seattle explored ways gaze centered eye tracking in VR can grant users more control over a prosthesis.26 Here, following a person's eye movements could help predict intent, like if they wanted to move in a certain direction or pick up a particular object.

Accenture Technology Vision 2024 - Page 67

As another example, Immersion Neuroscience is a company using smartwatches and fitness sensors to measure subtle changes in people's heart rate to predict their cognitive state.27

Accenture Technology Vision 2024 - Page 67

Others are building more detailed ways to understand people's intent in relation to their environments. Researchers at Tongji University's School of Automotive Studies, for instance, wanted to find out how to reduce human-vehicle collisions.28 Where most crash prevention efforts focus simply on detecting pedestrians, these researchers conducted a study that went further. They captured details including the distance between the vehicle and the pedestrian, the speed of the vehicle, and the pedestrian's physical posture - like the angle between their thighs and calves or between their calves and the ground. Grasping a person's posture while walking on a street can be a clue to discern where they are likely to move next, potentially making the roads safer for everyone.

Accenture Technology Vision 2024 - Page 67

Waymo, the self-driving car subsidiary of Alphabet, is doing something similar. In addition to watching everything happening around the car - like traffic signs and signals, pedestrians, construction, cyclists, and other Our bodies electronic cars - the Waymo Driver autonomous system also makes predictions about other road users' intent.29 Understanding that pedestrians, cyclists, and others all move differently, it predicts the many possible paths each might take, all in real-time.

Accenture Technology Vision 2024 - Page 68

Another approach to human intent is through AI. Consider human-robot collaborations. People's state of mind, like if they're feeling ambitious or tired, can impact how they approach a task. But while humans tend to be good at understanding these states of mind, robots aren't. So, researchers at the University of Southern California tried to teach robots to identify these states to help them better assist people.30 Typically, training a robot to fit an individual's work style takes a lot of time. But the researchers proposed a transfer learning system, where the robot observes someone performing a small, canonical task, then creates a preference model that it continuously updates as the robot and person interact.31

Accenture Technology Vision 2024 - Page 68

Blackrock Neurotech and Phantom Neuro, for instance, have partnered to build advanced prosthetic limbs and exoskeletons.33 By implanting a prosthetic at the amputated site, they can detect neural signals through muscle activity, and use that to control the robotic arm. The companies hope to build devices that can move in response to a person's intentions much like an intact limb would.

Accenture Technology Vision 2024 - Page 68

These are just a few of the "human interface" technologies starting to emerge. In coming years, we expect to see a wide range of devices and systems that can better understand human intent, ranging all the way from fully external, to skin-touching, to invasive implants in our bodies. While we focus primarily on external and skin-touching technologies here, it's worth noting that invasive devices are also advancing. Across the board, what's clear is that we're already starting to understand human intent to a previously impossible degree, and we're only getting better

Accenture Technology Vision 2024 - Page 69

In 2022, Galvani Bioelectronics, a joint venture from GlaxoSmithKline and Google's Verily, implanted its first neurostimulator in a patient with rheumatoid arthritis.34 And in early trials, Synchron's partially-invasive brain implant hopes to restore communication and other functions to severely paralyzed individuals.35

Accenture Technology Vision 2024 - Page 69

Two key advances are driving this. The first is decoding brain signals. For decades, it's been possible to sense brain signals, yet the leap to commercial products is a giant one.36 It's very difficult to identify common signals and patterns across different people's brains. But advances in AI pattern detection, as well as greater availability of brain data, is making a big difference.

Accenture Technology Vision 2024 - Page 69

at Accenture Labs, researchers tested how AI can efficiently generate synthetic brain signals to train detection models, without needing to rely on people's original brain signals.39 This capability demonstrates the potential to develop novel AI health solutions that make use of synthetic proxies, rather than depending on sensitive patient bio-signals.

Accenture Technology Vision 2024 - Page 70

new devices, like Wearable Sensing's DSI-24 headset that uses a dry electrode EEG system, are more resilient to motion and noise.42 And while fMRI is likely to remain in medical settings, a newer technology called fNIRS (functional nearinfrared spectroscopy) is making it possible to measure blood flow in the brain without people needing to be in a tube in a lab.43

Accenture Technology Vision 2024 - Page 71

Portability is critical to commercialization, and some companies are already building products that depend on it. EMOTIV, a bioinformatics startup, and X-trodes, which builds wireless wearable tech, are collaborating on a wearable at-home solution for brain and physiological measurement.45 They're building sticker electrodes that conform to people's skin and enable use cases where larger EEG devices are not ideal - like sleep studies. And Apple has submitted a patent application for measuring biosignals and electrical activity from a user's brain through an AirPods Sensor System.46

Accenture Technology Vision 2024 - Page 72

Accenture, for instance, is working with neurotech startup Mendi, to study how applications of neurofeedback can improve learning and training programs.50 Still others are thinking about consumer products. In the video game industry, leaders are exploring everything from how eye tracking can control or influence VR experiences to how neurotech can help them better understand how players react to games.51,52

Accenture Technology Vision 2024 - Page 73

Other industries see the promise, too. Cornell University researchers are experimenting with algorithms that can predict athletes' moves during a sports game with about 80% accuracy.53 Their algorithms use computer vision to interpret visual information about the player in real time, like their position on the court and their body posture, then combine that information with contextual data like the team's strategy or the player's role.

Accenture Technology Vision 2024 - Page 74

Researchers at the University of Technology Sydney, for instance, have developed a BCI headset that uses a biosensor to pick up brainwaves and translate them into commands.54 In a test with the Australian Army, soldiers were able to use the device to control a four-legged robot dog with just their minds - with up to 94% accuracy. And Snap has acquired NextMind, the maker of a mindcontrolled headband that lets people interact with and command digital objects with their brain signals.

Accenture Technology Vision 2024 - Page 75

And InnerEye, a neurotech startup, is transforming human-machine collaboration in yet another way. People's brains can process visual images very quickly but are slowed down by the cognitive and motor processes needed to decide and execute a response. InnerEye demonstrated that a BCI headset can take advantage of our brain's rapid response to patterns to enhance productivity.57 The company had a test subject watch airport security X-ray scans rapidly appear and disappear on a computer screen - at about three images per second. This pace is normally way too fast for a person to properly look for things like hidden firearms. But after the stream of scans ended, almost all the images flagged for firearms were indeed correct.

Accenture Technology Vision 2024 - Page 75

Still others are thinking about the "human interface" as a safety measure. Meili Technologies is a startup working to improve vehicle safety. It uses deep learning, visual inputs, and in-cabin sensors to detect if a driver has been incapacitated by a heart attack, seizure, stroke, or other emergency.59,60 And SmartCap is a fatigue awareness product built into a hard hat.61 It measures alertness Our bodies electronic and fatigue by analyzing EEG brain signals and provides early warning alerts. SmartCap Technologies first developed the product for the Australian mining industry to prevent safety hazards. The company was later purchased by Wenco International Mining Systems, a division of Hitachi Construction Machinery, which is expanding it into North America.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 46

Supercapacitors or ultracapacitors, which do not rely on electrochemical storage, provide another option for energy storage, with significant advantages in weight, safety, life, energy density, charging speed and toxicity [210].

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 53

4D Arrays: Manipulation of antenna arrays whereby individual elements are selectively turned off and on, creating unique and useful sidebands. These sidebands are suitable for multi-channel transmission and smart beamforming.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 54

Adaptive Spectrum Management: Cognitive radios or sensors employ AI/ML algorithms to adapt to a changing EM environment, including jamming and increased commercial use. They support more effective and agile use of the available spectrum. Applications extend from radar, signals intelligence, electronic warfare, and communications.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 58

Quantum computing: The use of superposition and entanglement to create qubits capable of being used for computation. The term quantum information science may also be used, although this includes not only quantum computers but the development of new specialised quantum-based algorithms, programming languages, interfaces, etc. Quantum computers are best seen as employing specialised processors suitable for a very limited (but important) class of problems in optimisation and simulation.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 60

Of all the quantum technologies, sensors are the most well-developed, enabling precision measurements of physical quantities such as atomic energy levels, photonic states, and spins.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 72

First, 2-D materials host considerable promise for truly disruptive effects. Graphene, the quintessential 2-D material, was initially unequivocally isolated in 2004. Graphene is one of the thinnest, strongest, stiffest, and more stretchable crystal materials [335]. As a result, 2D materials are an area of intense research and development, focused on exploring the unique properties of graphene and similar 2-D materials, such as another form of carbon called graphyne [336, 337, 338]. Because of these properties, there is intense interest in identifying alternative 2-D materials. Recently eight new materials have been identified that have a structure similar to graphene [100]: antimonene [339, 340, 341], arsenene (a single-layer buckled honeycomb structure of arsenic) [342, 343], bismuthine [344], borophane [345, 346], borophene [347], phagraphene [348], phosphorene[349] and stanene [350, 351, 352]. Other similar 2-D materials such as Cr2ge2te6 [353], Rhenim Disulfide [354], Titanium Carbides [355] and Molybdenum ditelluride [356] have unique optical, ferromagnetic, physical or electrical properties. A search for unique electronic, optical and physical properties and applications in areas such as electronics and energy storage drives further exploration.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 73

3D Printing and the additive manufacturing process it supports entered the mainstream of manufacturing and public consciousness in early 2000. While the terms are often used interchangeably, 3D-print adds materials in an iterative process to build up objects from digital models. At the same time, additive manufacturing uses 3D printing at an industrial scale to manufacture products [363]. Over the last few years, significant research advances have been made [364] in printing methods, devices, materials development, printing process and post-process modifications [365]. Novel applications and new methods continue to develop rapidly and are areas of intense research. One promising application area is based on biomimetics, e.g. replicating biological structures through 3D printing [366].

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 73

The increased use of 3D printing can potentially be highly disruptive in a defence context. For example, production lines of equipment and vehicles are currently closed down after production ceases. This means all spare parts must be produced before the line is closed. Consequently, military equipment is often retired and made surplus once the ability to find spares ceases to be cost-effective. 3D printing theoretically would be able to recreate new parts as long as the digital models are available, thereby extending the life of major pieces of equipment. Similarly, production lines could be re-established quickly and effectively.

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 73

First developed in 2013, 3D printed materials that transform under changing environmental stimuli such as pressure, heat, pH, light, humidity, or temperature are called 4D printed materials [100, 367, 368]. Such materials hold promise for new designs or sensors, especially in biomedical applications such as biomedical robots, tissue engineering or bio-scaffolds. [369, 370, 371]

NATO Science & Technology Trends 2023-2043 Volume 1: Overview - Page 74

On a larger scale, 3D-printed extruded concrete provides a very high degree of design flexibility and structural options for large buildings, especially those with complex geometries [100, 378, 379]. This may significantly reduce construction costs, increase deployment options for buildings in operational areas, and potentially reduce the costs for garrison locales. This technology is particularly disruptive in emergency shelters or extra-terrestrial habitation using locally sourced materials, allowing reduced costs and viable habitats for human habitation on the moon or other such bodies [380, 381]. 3D printing of nano-ceramics [382] is also being explored and is a very challenging process.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 33

AI techniques may be used to digest large datasets of textual data to conduct sentiment and social media analysis. Unsurprisingly, such large data collection exercises can be leveraged politically and militarily to influence storytelling, narrative creation and present emerging threats to pol itics [145, 146].

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 33

Some AI techniques, such as behavioural research advances in natural language processing (NLP) and AI-enabled synthetic text, video and audio (e.g. fake social media posts) in autonomous agents (e.g. bots) use disruptive deep Figure A.8: AI Cyber operations fake information as a statecraft tool to shape public perception to convenient geopolitical or military significant storytelling. Hence, demonstrating how content-based AI-enabled information is weaponised and presents a threat [145, 146, 149].

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 33

Further applications of embedded AI in autonomous vehicles (AVs) provide the intelligence behind the decision-making process, planning and execution of tasks [135]. For instance, vehicle control and human assistance depend on the level of entrusted autonomy. Standard or partial AVs perceive the context relying on advanced sensor networks and interact within it, signalling the need for human intervention. Whereas, highly automated vehicles (HAVs) depend on AI to operate complex data-intensive decision making functions required to understand the surroundings, plan for subsequent actions, and manoeuvre [135]. For instance, Project Maven embodies and weaponises the convergence of AI algorithms and drones in support of mission intervention [134].

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 33

In addition, breakthroughs in high-resolution image recognition, spanning from facial recognition to more subtle facial expression, enable biometric identification through facial features and emotion analysis [145]. Furthermore, use cases in the medical domain point out AI's high-resolution image recognition capabilities, which allow it to identify tumours and cancer cells [143, 150]. Thereon, raising expectations for military or counter-terrorism applications [145]

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 34

Complementary, advanced NLP progress on language processing and Learning enables system development to process complex inputs and engage in human-like conversation while developing a collaborative intelligence among cohorts of agents. Still, XAI, trust, and testing, validation, verification & assurance (TVVA) remain critical challenges for AI in human-machine symbiosis as for any defence and security environment.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 38

AI systems (especially when paired with digital twins) have the potential to minimise equipment downtime, minimise system failures, improve inventory and repairs management etc. Problems of these sorts are similar to those encountered in the commercial world and are therefore primed for early adoption by NATO.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 38

For example, Intelligence analysts can leverage trusted systems capable of tasking, collecting, processing, exploiting, disseminating (TCPED), and retrieving information across the spectrum of available sensors and relevant archival data.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 39

In such an interoperable system, interfaces will need to be understood, and heterogeneous systems, present or future, will need to work together with no or few restrictions. Therefore, common and standardised procedures must be established by NATO to define and conduct verification, validation and accreditation (VV&A) of AI-enabled operational decision support AI systems before they are used in Alliance military operations and all AI systems at large. Organisations will harness the full potential of interoperable and interconnected AI systems in the first instance through process building and policy writing on differing VV&A [104]. Secondly, addressing the limitations and pitfalls around AI-Black box and explainability [170, 42] and, simultaneously, investing heavily in M-M teaming and symbiosis, thus increasing trust across all components at all levels of the organisation.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 46

However, deployed unmanned weapon systems are remotely operated by a warfighter who augments the system's guidance, situational assessment, and decision-making - a supervisory role. These systems have demonstrated unquestioned value, playing vital roles such as Improvised Explosive Device (IED) interrogation, aerial surveillance, checkpoint inspection, and land or sea mine clearance. Although these systems help keep warfighters safe in 4D operational environments (Dull, Dirty, Dangerous and Dear) and have improved surveillance.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 47

RAS will further underpin operational success, as reduced costs, miniaturisation, broader employment, superior sensors, and sophisticated AI [175, 198, 199] will drive growth and more general usage. In particular, the growing importance of land and maritime Autonomy and the increased use of RAS for logistics, sustainment and manufacture is hard to overstate [200, 201].

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 47

Advanced RAS collaborative systems build upon (a federation of) individual agents (whether human, kinetic or non-kinetic) collective capability of adaptable RAS defining, distributing, and executing the tasks and sub-tasks necessary to achieve a common mission within a dynamic and uncertain environment.

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 49

The LONGSHOT program aims to develop a missile-firing missile [211], or more accurately, an air-launched missile-armed UAV, similar to what the European Defence Agency (EDA) proposes as high-performance gun launch and missile propulsion systems. A system with an extended engagement range increased mission effectiveness and reduced risk compared to crewed aircraft [211].

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 61

biotechnology is increasingly replacing oil refining to make commodity chemicals, including energetics and propellants. It has been reported that two-thirds of the world's most used chemicals could be synthesised from renewable raw materials, reducing reliance on fossil fuels. Microbes grown in fermentation vats are already being trained to produce these chemicals, and synthetic biology allows for making previously unattainable chemicals. Biomanufacturing is attractive to industry because of the low capital expenditure to build a facility, and bio-foundries can fairly easily switch from one product to another with little investment aside from seed stock since the same equipment can be used to produce a variety of chemicals even though they are sourced from different organisms

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 62

Researchers at Binghamton University have merged biotechnology with structural engineering. They have incorporated dormant fungal spores into a concrete matrix, which revive when exposed to water and oxygen permeating through the cracks [239]

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 64

Not only can DNA store information regarding living organisms, but scientists have also utilised this media to store digital information. Researchers have encoded gift cards into DNA and reconstituted them to buy books, while scientists from Harvard have even stored a short movie clip in living bacterial DNA and were able to recall the information [253].

NATO Science & Technology Trends 2023-2043 Volume 2: Analysis - Page 83

Probabilistic Programming Languages generalise programming languages by "two added constructs: (1) the ability to draw values at random from distributions, and (2) the ability to condition values of variables in a program via observations." [354]. Increased computational resources have enabled improvements in PPL but are still limited by computational effectiveness and ease of development. They are, however, applicable to a wide variety of analytic and modelling problems [354, 355, 356]