Generative AI | Accenture
Innovate securely, responsibly and sustainably with Large Language Models (LLMs) and Generative AI.
AI
GEN AI LLM
generative AI
technology
tech
2023
tech vision
tecnology vision 2023
Accenture
A new era of generative AI for everyone The technology underpinning ChatGPT will transform work and reinvent business
Table of 03 Welcome to AI’s new inflection point Contents 04 How did we get here? | Milestones in the journey to generative AI 05 Consume or customize: Generative AI for everyone 08 A look ahead at the fast-paced evolution of technology, regulation and business 12 Embrace the generative AI era: Six adoption essentials 19 The future of AI is accelerating 21 Glossary and References 22 Authors A new era of generative AI for everyone | 2
Introduction Welcome to AI’s new inflection point ChatGPT has woken up the world to A foundation model is a generic term for Business leaders recognize the significance the transformative potential of artificial large models with billions of parameters. With of this moment. They can see how LLMs intelligence (AI), capturing global attention recent advances, companies can now build and generative AI will fundamentally and sparking a wave of creativity rarely seen specialized image- and language-generating transform everything from business, to before. Its ability to mimic human dialogue models on top of these foundation models. science, to society itself—unlocking new and decision-making has given us AI’s first Large language models (LLMs) are both performance frontiers. The positive impact true inflection point in public adoption. a type of generative AI and a type of on human creativity and productivity will be Finally, everyone, everywhere can see the foundation model. massive. Consider that, across all industries, technology’s true disruptive potential for Accenture found 40% of all working hours themselves. The LLMs behind ChatGPT mark a significant can be impacted by LLMs like GPT-4. This turning point and milestone in artificial is because language tasks account for 62% intelligence. Two things make LLMs game of the total time employees work, and 65% changing. First, they’ve cracked the code on of that time can be transformed into more language complexity. Now, for the first time, productive activity through augmentation machines can learn language, context and and automation (see Figure 3). intent and be independently generative and creative. Second, after being pre-trained on vast quantities of data (text, images or audio), these models can be adapted or fine- tuned for a wide range of tasks. This allows them to be reused or repurposed in many different ways. A new era of generative AI for everyone | 3
How did we Machine learning: Analysis and prediction phase The first decade of the 2000s marked the rapid advance viewed machine learning as an incredibly powerful field of various machine learning techniques that could analyze of AI for analyzing data, finding patterns, generating get here? massive amounts of online data to draw conclusions – insights, making predictions and automating tasks at a or “learn” – from the results. Since then, companies have pace and on a scale that was previously impossible. Milestones in the journey Deep learning: Vision and speech phase to generative AI The 2010s produced advances in AI’s that search engines and self-driving cars use perception capabilities in the field of machine to classify and detect objects, as well as the learning called deep learning. Breakthroughs voice recognition that allows popular AI speech in deep learning enable the computer vision assistants to respond to users in a natural way. Generative AI: Enter the language-mastery phase Building on exponential increases in the size and phase in the abilities of language-based AI applications. Models capabilities of deep learning models, the 2020s will be such as this will have far-reaching consequences for business, about language mastery. The GPT-4 language model, since language permeates everything an organization does day to developed by OpenAI, marks the beginning of a new 2 day—its institutional knowledge, communication and processes. A new era of generative AI for everyone | 4
Consume or customize: Generative AI for everyone A new era of generative AI for everyone | 5
Consume or customize: Generative AI for everyone Consume or customize: Generative AI for everyone Easy-to-consume generative AI applications like We’re at a phase in the adoption cycle when ChatGPT, DALL-E, Stable Diffusion and others are most organizations are starting to experiment rapidly democratizing the technology in business by consuming foundation models “off the shelf.” and society. The effect on organizations will be However, the biggest value for many will come profound. The ability of LLMs to process massive when they customize or fine tune models using data sets allows them to potentially “know” their own data to address their unique needs: everything an organization has ever known—the Consume entire history, context, nuance and intent of a business, and its products, markets and customers. Generative AI and LLM applications are ready to Anything conveyed through language (applications, consume and easy to access. Companies can systems, documents, emails, chats, video and audio consume them through APIs and tailor them, to recordings) can be harnessed to drive next-level a small degree, for their own use cases through innovation, optimization and reinvention. prompt engineering techniques such as prompt tuning and prefix learning. Customize But most companies will need to customize models, by fine-tuning them with their own data, to make them widely usable and valuable. This will allow the models to support specific downstream tasks all the way across the business. The effect will be to increase a company’s efficacy in using AI to unlock new performance frontiers—elevating employee capabilities, delighting customers, introducing new business models and boosting responsiveness to signals of change. A new era of generative AI for everyone | 6
Consume or customize: Generative AI for everyone Companies will use these models to reinvent the Creating. Generative AI will become an essential Automating. Generative AI’s sophisticated way work is done. Every role in every enterprise creative partner for people, revealing new ways understanding of historical context, next has the potential to be reinvented, as humans to reach and appeal to audiences and bringing best actions, summarization capabilities, and working with AI co-pilots becomes the norm, unprecedented speed and innovation in areas like predictive intelligence will catalyze a new era dramatically amplifying what people can achieve. In production design, design research, visual identity, of hyper-efficiency and hyper-personalization any given job, some tasks will be automated, some naming, copy generation and testing, and real- in both the back and front office—taking will be assisted, and some will be unaffected by the time personalization. Companies are turning to business process automation to a transformative technology. There will also be a large number of state-of-the-art artificial intelligence systems like new level. One multinational bank is using new tasks for humans to perform, such as ensuring DALL·E, Midjourney and Stable Diffusion for their generative AI and LLMs to transform how it the accurate and responsible use of new social media visual content generation outreach. manages volumes of post-trade processing AI-powered systems. DALL·E, for example, creates realistic images and emails—automatically drafting messages with Consider the impact in these key functions: art based on text descriptions and can process up recommended actions and routing them to the to 12 billion parameters when transforming words recipient. The result is less manual effort and Advising. AI models will become an ever-present into pictures. Images created can then be shared smoother interactions with customers. 5 co-pilot for every worker, boosting productivity on Instagram and Twitter. Protecting. In time, generative AI will support by putting new kinds of hyper-personalized Coding. Software coders will use generative AI to enterprise governance and information security, intelligence into human hands. Examples include significantly boost productivity — rapidly converting protecting against fraud, improving regulatory customer support, sales enablement, human one programming language to another, mastering compliance, and proactively identifying resources, medical and scientific research, programming tools and methods, automating code risk by drawing cross-domain connections corporate strategy and competitive intelligence. writing, predicting and pre-empting problems, and inferences both within and outside the Large language models could be useful in and managing system documentation. Accenture organization. In strategic cyber defense, LLMs tackling the roughly 70% of customer service is piloting the use of OpenAI LLMs to enhance could offer useful capabilities, such as explaining communication that is not straightforward and 6 can benefit from a conversational, powerful and developer productivity by automatically generating malware and quickly classifying websites. intelligent bot, understanding a customer’s intent, documentation – for example, SAP configuration In the short term, however, organizations can formulate answers on its own and improve the rationale and functional or technical specs. The expect criminals to capitalize on generative AI’s solution enables users to submit requests through capabilities to generate malicious code or write 4 7 accuracy and quality of answers. a Microsoft Teams chat as they work. Correctly the perfect phishing email. packaged documents are then returned at speed — a great example of how specific tasks, rather than entire jobs, will be augmented and automated. A new era of generative AI for everyone | 7
A look ahead at the fast-paced evolution of technology, regulation and business A new era of generative AI for everyone | 8
A look ahead at the fast-paced evolution of technology, regulation and business A look ahead at the fast-paced evolution of technology, regulation and business Moments like this don’t come around often. The coming years will see outsized investment Figure 1: Each layer of the generative AI tech stack will rapidly evolve in generative AI, LLMs and foundation models. What’s unique about this evolution is that the technology, regulation, and business adoption Applications: Generative AI and LLMs will be increasingly are all accelerating exponentially at the same accessible to users in the cloud via APIs and by being embedded time. In previous innovation curves, the directly into other applications. Companies will consume them technology typically outpaced both adoption as they are or will customize and fine-tune them with proprietary and regulation. data. The technology stack Fine-tuning: The importance of model fine-tuning will create demand for a multidisciplinary set of skills spanning software The complex technology underpinning engineering, psychology, linguistics, art history, literature and generative AI is expected to evolve rapidly library science. at each layer. This has broad business Foundation models: The market will rapidly mature and diversify implications. Consider that the amount of as more pre-trained models emerge. New model designs will compute needed to train the largest AI models offer more choices for balancing size, transparency, versatility and has grown exponentially – now doubling performance. between every 3.4 to 10 months, according to 8 Data: Improving the maturity of the enterprise data lifecycle various reports. Cost and carbon emissions are therefore central considerations in will become a prerequisite for success – requiring mastery of adopting energy-intensive generative AI. new data, new data types and immense volumes. Generative AI features within modern data platforms will emerge, enhancing adoption at scale. Infrastructure: Cloud infrastructure will be essential for deploying generative AI while managing costs and carbon emissions. Data centers will need retrofitting. New chipset architectures, hardware innovations, and efficient algorithms will also play a critical role. A new era of generative AI for everyone | 9
A look ahead at the fast-paced evolution of technology, regulation and business The risk and regulatory environment AI systems need to be “raised” with a diverse Figure 2: Key risk and regulatory questions for generative AI and inclusive set of inputs so that they reflect Companies will have thousands of ways to the broader business and societal norms of apply generative AI and foundation models responsibility, fairness and transparency. When Intellectual property: How will the business protect its own to maximize efficiency and drive competitive AI is designed and put into practice within an IP? And how will it prevent the inadvertent breach of third-party advantage. Understandably, they’ll want to get ethical framework, it accelerates the potential copyright in using pre-trained foundation models? started as soon as possible. But an enterprise- for responsible collaborative intelligence, wide strategy needs to account for all the where human ingenuity converges with Data privacy and security: How will upcoming laws like variants of AI and associated technologies they intelligent technology. the EU AI Act be incorporated in the way data is handled, intend to use, not only generative AI and large processed, protected, secured and used? language models. This creates a foundation for trust with consumers, the workforce, and society, and ChatGPT raises important questions about the can boost business performance and unlock Discrimination: Is the company using or creating tools responsible use of AI. The speed of technology new sources of growth. that need to factor in anti-discrimination or anti-bias evolution and adoption requires companies considerations? to pay close attention to any legal, ethical and reputational risks they may be incurring. Product liability: What health and safety mechanisms need It’s critical that generative AI technologies, to be put in place before a generative AI-based product is including ChatGPT, are responsible and taken to market? compliant by design, and that models and applications do not create unacceptable risk Trust: What level of transparency should be provided to for the business. Accenture was a pioneer in consumers and employees? How can the business ensure the the responsible use of technology including accuracy of generative AI outputs and maintain user confidence? the responsible use of AI in its Code of Business Ethics from 2017. Responsible AI is the practice of designing, building and deploying Identity: When establishing proof-of-personhood depends on voice AI in accordance with clear principles to or facial recognition, how will verification methods be enhanced and empower businesses, respect people, and improved? What will be the consequences of its misuse? benefit society — allowing companies to engender trust in AI and to scale AI with confidence. A new era of generative AI for everyone | 10
A look ahead at the fast-paced evolution of technology, regulation and business The scale of adoption in business Figure 3: Generative AI will transform work across industries Companies must reinvent work to find a path to generative AI value. Business Banking 54% 12% 24% 10% Work time distribution by industry leaders must lead the change, starting Insurance 48% 14% 26% 12% and potential AI impact now, in job redesign, task redesign and Based on their employment levels in the US in 2021 reskilling people. Ultimately, every role Software & Platforms 36% 21% 28% 15% in an enterprise has the potential to Lower potential for Capital markets 40% 14% 29% 18% Higher potential for Higher potential for augmentation or Non-language be reinvented, once today’s jobs are automation augmentation automation tasks decomposed into tasks that can be Energy 43% 9% 14% 34% automated or assisted and reimagined for a new future of human + machine work. Communications & Media 33% 13% 21% 33% Generative AI will disrupt work as Retail 34% 7% 12% 46% we know it today, introducing a new Industry Average 31% 9% 22% 38% 40% of working hours across dimension of human and AI collaboration industries can be impacted by in which most workers will have a “co- Health 28% 11% 33% 27% Large Language Models (LLMs) pilot,” radically changing how work is Public Service 30% 9% 35% 26% done and what work is done. Nearly every job will be impacted – some will Aerospace & Defense 26% 13% 20% 41% Why is this the case? Language tasks account for 62% of total worked time be eliminated, most will be transformed, in the US. Of the overall share of language tasks, 65% have high potential and many new jobs will be created. Automotive 30% 6% 13% 50% to be automated or augmented by LLMs. Organizations that take steps now to High Tech 26% 8% 16% 50% decompose jobs into tasks, and invest in training people to work differently, Travel 28% 6% 15% 50% alongside machines, will define new Utilities 27% 6% 15% 52% performance frontiers and have a big leg Source: Accenture Research based on analysis of Occupational up on less imaginative competitors. Life Sciences 25% 8% 17% 50% Information Network (O*NET), US Dept. of Labor; US Bureau of Labor Statistics. Industrial 26% 6% 14% 54% Notes: We manually identified 200 tasks related to language (out Consumer Goods & Services 24% 6% 13% 57% of 332 included in BLS), which were linked to industries using their share in each occupation and the occupations’ employment level Chemicals 24% 5% 14% 56% in each industry. Tasks with higher potential for automation can Natural Resources 20% 5% 11% 64% be transformed by LLMs with reduced involvement from a human worker. Tasks with higher potential for augmentation are those in 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% which LLMs would need more involvement from human workers. A new era of generative AI for everyone | 11
Embrace the generative AI era: Six adoption essentials A new era of generative AI for everyone | 12
Embrace the generative AI era: Six adoption essentials A new era of generative AI for everyone | 13
Embrace the generative AI era: Six adoption essentials Dive in, with a business-driven mindset Even when new innovations have obvious advantages, 1 diffusing them across an organization can be challenging, especially if the innovation is disruptive to current ways of A bank uses enhanced search to equip working. By experimenting with generative AI capabilities, employees with the right information companies will develop the early successes, change agents and opinion leaders needed to boost acceptance and spread the innovation further, kick-starting the transformation and As part of its three-year innovation plan, reskilling agenda. a large European banking group saw an Organizations must take a dual approach to experimentation. opportunity to transform its knowledge One, focused on low-hanging fruit opportunities using consumable models and applications to realize quick returns. base, empower its people with access to The other, focused on reinvention of business, customer the right information, and advance its goal engagement and prodicts and services using models that are customized with the organization’s data. A business-driven of becoming a data-driven bank. Using mindset is key to define, and successfully deliver on, the Microsoft’s Azure platform and a GPT- business case. 3 LLM to search electronic documents, As they experiment and explore reinvention opportunities, users can get quick answers to their they’ll reap tangible value while learning more about which types of AI are most suited to different use cases, since the questions — saving time while improving level of investment and sophistication required will differ accuracy and compliance. The project, based on the use case. They’ll also be able to test and improve their approaches to data privacy, model accuracy, which included employee upskilling, is bias and fairness with care, and learn when “human in the the first of four that will apply generative loop” safeguards are necessary. AI to the areas of contract management, conversational reporting and ticket classification. A new era of generative AI for everyone | 14
This is a modal window.
Embrace the generative AI era: Six adoption essentials Figure 4: Generative AI will transform work across every job category Take a people-first approach Office and Administrative Support 57% 6% 14% 23% Work time distribution by major Success with generative Sales and Related 49% 13% 14% 24% occupation and potential AI impact 2AI requires an equal attention on Based on their employment levels in the US in 2021 people and training as it does on Computer and Mathematical 28% 32% 23% 17% technology. Companies should Business and Financial Operations 45% 14% 35% 6% therefore dramatically ramp up Lower potential for Higher potential for Higher potential for augmentation or Non-language investment in talent to address Arts, Design, Entertainment, Sports, and Media 25% 26% 26% 22% automation augmentation automation tasks two distinct challenges: creating Life, Physical, and Social Science 27% 20% 25% 28% AI and using AI. This means both building talent in technical Architecture and Engineering 21% 24% 25% 30% competencies like AI engineering and enterprise architecture Legal 33% 9% 58% 0% and training people across the Occcupation Average 31% 9% 22% 38% In 5 out of 22 occupation organization to work effectively with AI-infused processes. In our Management 30% 9% 44% 17% groups, Generative AI can analysis across 22 job categories, Personal Care and Service 29% 8% 31% 32% affect more than half of all for example, we found that hours worked LLMs will impact every category, Healthcare Practitioners and Technical 22% 15% 40% 22% ranging from 9% of a workday at Community and Social Service 29% 7% 59% 6% the low end to 63% at the high end. More than half of working Healthcare Support 27% 8% 31% 34% hours in 5 of the 22 occupations Protective Service 29% 6% 23% 43% can be transformed by LLMs. Educational Instruction and Library 23% 8% 50% 19% Food Preparation and Serving Related 25% 5% 9% 61% Source: Accenture Research based on analysis of Occupational Transportation and Material Moving 23% 4% 7% 66% Information Network (O*NET), US Dept. of Labor; US Bureau of Labor Statistics. Construction and Extraction 15% 4% 7% 75% Notes: We manually identified 200 tasks related to language (out Installation, Maintenance, and Repair 16% 1%9% 75% of 332 included in BLS), which were linked to industries using their share in each occupation and the occupations’ employment level Farming, Fishing, and Forestry 8% 8% 17% 66% in each job category. Tasks with higher potential for automation can Production 14% 2% 8% 76% be transformed by LLMs with reduced involvement from a human worker. Tasks with higher potential for augmentation are those in Building and Grounds Cleaning and Maintenance 9% 0% 7% 84% which LLMs would need more involvement from human workers. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% A new era of generative AI for everyone | 15
Embrace the generative AI era: Six adoption essentials In fact, independent economic research indicates that companies are significantly underinvesting in helping workers keep up with advances in AI, which require Figure 5: Reinventing a customer service job, task by task 11 2more cognitively complex and judgment-based tasks. Even domain experts who understand how to apply To assess how specific jobs will be reinvented with AI, an Accenture analysis decomposed data in the real world (a doctor interpreting health data, one customer service job into 13 component tasks. We found: for example) will need enough technical knowledge of how these models work to have confidence in using them as a “workmate.” There will also be entirely new roles to recruit, including tasks would continue to be performed linguistics experts, AI quality controllers, AI editors, primarily by humans, with low potential and prompt engineers. In areas where generative 4for automation or augmentation. AI shows most promise, companies should start by decomposing existing jobs into underlying bundles of tasks. Then assess the extent to which generative AI might affect each task — fully automated, augmented, tasks could be fully automated — or unaffected. such as gathering, classifying, and summarizing information on why a 4customer is contacting the company. tasks could be augmented to help humans work more effectively — such as using an AI summary to provide a 5rapid solution with a human touch. Importantly, new job tasks might also be needed to ensure the safe, accurate and responsible use of AI in customer service settings, such as providing unbiased information on products and pricing. A new era of generative AI for everyone | 16
This is a modal window.
Embrace the generative AI era: Six adoption essentials Get your proprietary data ready Invest in a sustainable tech foundation Customizing foundation models will require Companies need to consider whether they have the 3access to domain-specific organizational data, 4right technical infrastructure, architecture, operating semantics, knowledge, and methodologies. In the model and governance structure to meet the high pre-generative AI era, companies could still get compute demands of LLMs and generative AI, while value from AI without having modernized their keeping a close eye on cost and sustainable energy data architecture and estate by taking a use-case consumption. They’ll need ways to assess the cost centric approach to AI. That’s no longer the case. and benefit of using these technologies versus other Foundation models need vast amounts of curated AI or analytical approaches that might be better data to learn and that makes solving the data suited to particular use cases, while also being challenge an urgent priority for every business. several times less expensive. Companies need a strategic and disciplined As the use of AI increases, so will the carbon approach to acquiring, growing, refining, emissions produced by the underlying infrastructure. safeguarding and deploying data. Specifically, they Companies need a robust green software need a modern enterprise data platform built on development framework that considers energy cloud with a trusted, reusable set of data products. efficiency and material emissions at all stages of the Because these platforms are cross-functional, with software development lifecycle. AI can also play a enterprise-grade analytics and data housed in cloud- broader role in making business more sustainable based warehouses or data lakes, data is able to break and achieving ESG goals. Of the companies we free from organizational silos and democratized for surveyed that successfully reduced emissions in 12 use across an organization. All business data can production and operations, 70% used AI to do it. then be analyzed together in one place or through a distributed computing strategy, such as a data mesh. Read more on the practices data-mature companies are using to maximize enterprise data value: A new dawn for dormant data: Unleash the intrinsic value of enterprise data with a strong digital core on cloud. A new era of generative AI for everyone | 17
Embrace the generative AI era: Six adoption essentials Accelerate ecosystem innovation Level-up your responsible AI Creating a foundation model can be a complex, The rapid adoption of generative AI brings fresh urgency 5compute-intensive and costly exercise. And for 6to the need for every organization to have a robust all but the very largest global companies, doing it responsible AI compliance regime in place. This includes entirely on their own will be beyond their means controls for assessing the potential risk of generative AI and capabilities. The good news is that there is a use cases at the design stage and a means to embed burgeoning ecosystem to call on, with substantial responsible AI approaches throughout the business. investments by cloud hyperscalers, big tech players, Accenture’s research suggests most companies still and start-ups. Global investment in AI startups have a long way to go. Our 2022 survey of 850 senior and scale-ups is estimated to exceed $50 billion in executives globally revealed widespread recognition 13 2023 alone. These partners bring best practices of the importance of responsible AI and AI regulation. honed over many years, and can provide valuable But only 6 percent of organisations felt they had a fully insights into using foundation models efficiently robust responsible AI foundation in place. and effectively in specific use cases. Having the right network of partners—including technology An organization’s responsible AI principles should be companies, professional services firms and academic defined and led from the top and translated into an institutions—will be key to navigating rapid change. effective governance structure for risk management and compliance, both with organizational principles and policies and applicable laws and regulations. Responsible AI must be CEO-led, beginning with a focus on training and awareness and then expanding to focus on execution and compliance. Accenture was one of the first to take this approach to Responsible AI years ago, with a CEO-led agenda, and now a formal compliance program. Our own experience shows that a principles- driven compliance approach provides guardrails while being flexible enough to evolve with the fast pace of changing technology, ensuring companies aren’t constantly playing “catch up.” To be responsible by design, organizations need to move from a reactive compliance strategy to the proactive development of mature Responsible AI capabilities through a framework that includes principles and governance; risk, policy and control; technology and enablers and culture and training. A new era of generative AI for everyone | 18
This is a modal window.
The future of AI is accelerating A new era of generative AI for everyone | 19
The future of AI is accelerating This is a pivotal moment. For several years, Businesses are right to be optimistic about the generative AI and foundation models have been potential of generative AI to radically change how quietly revolutionizing the way we think about work get done and what services and products machine intelligence. Now, thanks to ChatGPT, they can create. They also need to be realistic the whole world has woken up to the possibilities about the challenges that come with profoundly this creates. rethinking how the organization works, with implications for IT, organization, culture, and While artificial general intelligence (AGI) remains responsibility by design. a distant prospect, the speed of development continues to be breathtaking. We’re at the start of Companies need to invest as much in evolving an incredibly exciting era that will fundamentally operations and training people as they do in transform the way information is accessed, technology. Radically rethinking how work gets content is created, customer needs are served, done, and helping people keep up with technology- and businesses are run. driven change, will be two of the most important factors in realizing the full potential of this step- Embedded into the enterprise digital core, change in AI technology. generative AI, LLMs, and foundation models will optimize tasks, augment human capabilities, and Now’s the time for companies to use open up new avenues for growth. In the process, breakthrough advances in AI to set new these technologies will create an entirely new performance frontiers—redefining themselves language for enterprise reinvention. and the industries in which they operate. A new era of generative AI for everyone | 20
Glossary References ChatGPT is a generative AI chatbot interface built on top of OpenAI’s GPT-3.5 1. ChatGPT sets record for fastest-growing user base - analyst note, Reuters, February 2023 large language model (see below). ChatGPT (and ChatGPT plus, which uses https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/ GPT-4) allows users to interact with the underlying AI in a way that seems 2. The Next Big Breakthrough in AI Will Be Around Language, Harvard Business Review, September, 2020 remarkably accurate and feels surprisingly human. You can ask it to explain https://hbr.org/2020/09/the-next-big-breakthrough-in-ai-will-be-around-language a subject, write an essay, run a calculation, generate some Python code, or simply have a conversation. 3. Accenture Tech Vision 2023 Generative AI is the umbrella term for the ground-breaking form of creative 4. ChatGPT Is Coming to a Customer Service Chatbot Near You, Forbes, January 2023 artificial intelligence that can produce original content on demand. Rather https://www.forbes.com/sites/rashishrivastava/2023/01/09/chatgpt-is-coming-to-a-customer-service-chatbot-near- than simply analyzing or classifying existing data, generative AI is able to you/?sh=730eeab97eca create something entirely new, whether text, images, audio, synthetic data, or more. 5. How AI Transforms Social Media, Forbes, March 2023 Foundation models are complex machine learning systems trained on vast https://www.forbes.com/sites/forbestechcouncil/2023/03/16/how-ai-transforms-social-media/?sh=739221ca1f30 quantities of data (text, images, audio, or a mix of data types) on a massive 6. Large AI Models have Real Security Benefits, Dark Reading, August, 2022 scale. The power of these systems lies not only in their size but also in the fact https://www.darkreading.com/dr-tech/large-language-ai-models-have-real-security-benefits they can quickly be adapted or fine-tuned for a wide range of downstream tasks. Examples of foundation models include BERT, DALL-E, and GPT-4. 7. OPWNAI: Cybercriminals starting to use ChatGPT, Checkpoint Research, January, 2023 Large Language Models (LLMs) represent a subset of foundation models https://research.checkpoint.com/2023/opwnai-cybercriminals-starting-to-use-chatgpt/ that are trained specifically on text sources. GPT-3, for instance, was trained 8. Accenture Technology Vision 2023 14 on almost 500 billion words from millions of websites. Its successor, GPT-4, can take image as well as text as inputs. 9. CXO Pulse Survey, conducted by Accenture Research, February 2023 Fine-tuning is the process by which foundation models are adapted for 10. Accenture Technology Vision 2023 specific downstream tasks using a particular dataset. That can include everything from the hyper-specific (training a model to compose emails 11. The Productivity J-Curve: How Intangibles Complement General Purpose Technologies - American Economic Association (aeaweb.org) based on your personal writing style) to the enterprise level (training an LLM on enterprise data to transform a company’s ability to access and analyze its 12. Uniting technology and sustainability, Accenture, May, 2022 core intelligence). Technology Sustainability Key to ESG Goals | Accenture Data is the fundamental bedrock of generative AI. Not only in training 13. Pace Of Artificial Intelligence Investments Slows, But AI Is Still Hotter Than Ever, Forbes, October, 2022 foundation models themselves, but also in fine-tuning those models to https://www.forbes.com//sites/joemckendrick/2022/10/15/pace-of-artificial-intelligence-investments-slows-but-ai-is-still-hotter-than- perform specific tasks. In an enterprise context, examples might include ever/?sh=853d8124c76c everything from legacy code to real-time operational data to customer insights. 14. OpenAI’s GPT-3 Language Model: A Technical Overview, Lambda, June, 2020 https://lambdalabs.com/blog/demystifying-gpt-3 A new era of generative AI for everyone | 21
Authors Paul Daugherty Bhaskar Ghosh Karthik Narain Group Chief Executive & Chief Strategy Officer, Lead – Accenture Cloud First Chief Technology Officer Accenture Lan Guan Jim Wilson The authors would like to acknowledge Lead – Cloud First, Data & AI Global Managing Director – Thought Tomas Castagnino, Elise Cornille, Leadership & Technology Research Ray Eitel-Porter, Linda King, Amy Sagues, Ezequiel Tacsir and Denise Zheng for their contributions.
About Accenture Contact us Accenture is a leading global professional services company For more information, contact the Accenture Generative AI/ that helps the world’s leading businesses, governments and Large Language Model Center of Excellence at: other organizations build their digital core, optimize their [email protected]. operations, accelerate revenue growth and enhance citizen services—creating tangible value at speed and scale. We are a talent and innovation led company with 738,000 people serving clients in more than 120 countries. Technology is at the core of change today, and we are one of the world’s leaders in helping drive that change, with strong ecosystem relationships. We combine our strength in technology with unmatched industry experience, functional expertise and global delivery capability. We are uniquely able to deliver tangible outcomes because of our broad range of services, solutions and assets across Strategy & Consulting, Technology, Operations, Industry X and Accenture Song. These capabilities, together with our culture of shared success and commitment to creating 360° value, enable us to help our clients succeed and build trusted, lasting relationships. We measure our success by the 360° value we create for our clients, each other, our shareholders, partners and communities. Visit us at www.accenture.com. Copyright © 2023 Accenture. All rights reserved. Accenture and its logo are registered trademarks of Accenture Disclaimer: This content is provided for general information purposes and is not intended to be used in place of consultation with our professional advisors. This document refers to marks owned by third parties. All such third-party marks are the property of their respective owners. No sponsorship, endorsement or approval of this content by the owners of such marks is intended, expressed or implied.