Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Apidays Paris 2023 - AIvolution or AIPocalypse,...

apidays
December 16, 2023

Apidays Paris 2023 - AIvolution or AIPocalypse, Cyril Vart, Fabernovel

Apidays Paris 2023 - Software and APIs for Smart, Sustainable and Sovereign Societies
December 6, 7 & 8, 2023

AIvolution or AIPocalypse
Cyril Vart, Fabernovel, an E&Y company

------

Check out our conferences at https://www.apidays.global/

Do you want to sponsor or talk at one of our conferences?
https://apidays.typeform.com/to/ILJeAaV8

Learn more on APIscene, the global media made by the community for the community:
https://www.apiscene.io

Explore the API ecosystem with the API Landscape:
https://apilandscape.apiscene.io/

apidays

December 16, 2023
Tweet

More Decks by apidays

Other Decks in Programming

Transcript

  1. We ask our smartphones :questions: It helps us :choose a

    film to watch: in the evening It helps us make medical decisions: We use it for our :banking: We use it for :directions: We use it on our daily journeys to get from A to B AI was already part of our everyday lives
  2. AI, a revolution underway. Has the time finally come (now)?

    Sources: New York Times, ADN, CBSnews (And why the answer is probably yes)
  3. Of course — deus ex machina? In popular culture and

    science-fiction, AI is often portrayed negatively — with conscious, defiant or evil powers… set on ruling the world, instilling fear and fantasy.
  4. April 2022 November 2022 March 2022 July 2022 March 2023

    Evolution of Images Generation, in 1 year
  5. In 1956, the Dartmouth summer research project was organized on

    Artificial Intelligence. A brief history of Artificial Intelligence EXPERT SYSTEMS Each decision-making rule is explicitly integrated into the machine through code. MACHINE LEARNING; A range of techniques that enable the machine to learn by example DEEP LEARNING; Advanced techniques using deep neural networks GENERATIVE AI Automatic content generation and autonomous learning 1950’s 1960’s 1970’s 1980’s 1990’s 2000’s 2010’s 2020’s In 1997, IBM Deep Blue defeated Kasparov, the world champion. In 2016, AlphaGo defeated the professional Go player Lee Sedol.
  6. Speed or acceleration ? Hand writing recognition Vocal recognition Image

    recognition Text understanding Language recognition AI vs. Human AI above human performances AI below standards Human performance baseline Normalized average AI systems performance 2017 - Transformer models 2013 - Word-Vectors
  7. Generation Data collection 9 GENERATIVE AI Automatic learning and content

    generation Generative AI refers to Artificial Intelligence and Machine Learning algorithms that use existing content to generate new content as part of their learning process. Training Text Images Sound Textual content generation Information extraction Image generation Image description Objects recognition Vocal recognition Voice generation → → →
  8. 16 years 7 years 10 years 2 months AI: from

    a niche market to a mass market tool ChatGPT, 100 million users in 2 months ChatGPT Initially designed for specialised applications and niche markets, AI has transformed into a mass consumer product. What was previously the domain of experts and researchers is now available to the general public with several hundred million users. WhatsApp Internet Netflix (streaming) Mobile phones Sources: World of Statistics, CNBC, EY Fabernovel 3.5 years
  9. You said Generative AI ? It’s not all about ChatGPT

    MIDJOURNEY 3D & Gaming Code Text Image Video Sound Bing Chat
  10. The best website to find AI solutions For every need,

    there's an AI to meet it (well, almost every need)
  11. From… The integration of lines of code by a human

    so that the machine produces a result Natural language Automatic content generation by machine self-learning Communicate with AI using simple sentences i.e. “prompts” A tool which can be accessed by all and partly free for the general public The need to master coding to communicate with AI Costly software needs installing on specific machines From query to deliverable, a first-ever for AI Barrier to entry To… Content generation
  12. Complex and dangerous soldering tasks Automatic transcription of meeting minutes

    or medical appointments Document research Radiology analysis AI: from technology to a job AI Natural language processing Machine Learning Spoken language Robotics Vision Deep learning Text classification Text to speech Industrial robotics Troubleshooting tools for technicians Speech to text Image recognition Autonomous vehicles Predictive analysis Automatic translation Information extraction
  13. But requiring human supervision to work Generative AI • Ingests

    data • Processes the information quickly • Recognises shapes • Suggests a probabilistic approach (and sometimes aberration, false information) • Optimizes Human • Understands the context • Abstraction / concepts • Critical thinking • Empathy • Emotion • Innovation
  14. Will AI cause mass job destruction? (And why the answer

    is probably no) The 5 observations of EY Fabernovel which lead us to believe that generative AI will not be a major driving force in destroying jobs: 1. Artificial intelligence is already at play in the workplace 2. Now white-collar workers are creating the traction 3. AI is changing the list of tasks rather than taking over a job entirely 4. AI improves the employee experience 5. AI is driving new opportunities for companies and employers
  15. How generative AI is transforming our daily work environment Generative

    AI can expand our capacities to optimise work: by doing repetitive and time-consuming tasks and exploring new tasks which were difficult to carry out without changing the time and resources allocated. Note taking Summarising information Formalising documents and deliverables Document research and access to information
  16. And is already part of many work applications Microsoft uses

    a new conversational agent in Word, Excel, PowerPoint, Outlook, Teams and OneNote: Microsoft Copilot, which is capable of processing files and creating content. Generative AI is now included in lots of professional software. Some businesses already use these tools on a daily basis. For example marketing agencies use them to produce different customised advertising banners. However, the current uses of existing or future tools have not yet become widespread. But adopting these new uses is likely to accelerate thanks to the firepower of some tech giants, for example the deployment of generative AI in Microsoft Office — which makes up for 46 % of desktop software used by corporates. Zoom offers real-time transcription of meetings, generating text from the audio and automated minutes. Since November 2022 Notion includes a conversational feature which can draft new content, summarise long documents and extract key information from disjointed notes. Source: BDM/Statista
  17. Second round: after blue-collar workers, now it’s white-collar workers’ turn

    Unlike previous industrial revolutions which affected blue- collar jobs or lower qualified jobs, this time round generative AI is affecting white-collar office tasks. But this will probably also mean upskilling of these office jobs thanks to the automation of certain less fulfilling tasks repositioning employees towards advisory tasks involving greater expertise.
  18. Middle management: women are particularly affected A report published in

    August 2023 by the UN’s International Labour Organization (ILO) suggests that automation due to generative AI will have a greater impact on higher income countries, and more specifically office jobs. 5.5% of jobs in these wealthy countries are at risk, vs. only 0.4% in lower income countries. The impact on jobs will also affect women more, as they make up a larger proportion of office jobs. Indeed, the automation risk is twice as high for women than for men, especially in higher and middle-income countries. Automation potential vs augmentation potential: shares of total employment ILO
  19. Evolving the tasks rather than transforming jobs Generative AI is

    more likely to transform certain tasks rather than the jobs per se. Generative AI will enable us to modify the way we carry out certain tasks. But will also allow upskilling via new tasks requiring human analysis, creativity and critical thinking. “I see all these tools as assistants for some of the tasks, the tasks are replaced but not the jobs. A job requires skills in human relations, politics, and reporting which will never be automated.” Pierre-Yves Calloc'h, Chief Digital Officer at Pernod Ricard, during an interview with EY Fabernovel for this study “AIpocalypse or AIvolution?” (2023)
  20. How will my job be affected? 5 questions to consider

    to see whether AI will steal my job 1. Can my job be easily split up into simple and well-defined steps? 2. Is my job made up of several repetitive tasks? 3. Does my job require creative, critical or strategic thinking? Yes No Yes No Yes No 4. Does one of my roles require analysing results produced from data? 5. Does my job require the ability to perceive and respond to emotions? Yes No Yes No Job certified at risk
  21. Generative AI will have a moderate impact for a large

    proportion of jobs Low or no impact Example: piano tuner, wine merchant, carpenter… High impact for jobs depending on the nature of their daily tasks. Example: notary clerk, data scientist, marketing consultant… Moderate impact on some tasks for the majority of professions whose core taks will not change. Example: teacher, salesperson, telephone operator, financial analyst, lawyer… Source: Orders of magnitude estimated by EY Fabernovel Generative AI changes impacting jobs Number of jobs impacted
  22. A job is merely a list of tasks that the

    employee masters Main tasks which define her job: 1. Review financial data and prepare annual and monthly reports. 2. Determine long term financial forecasts. 3. Analyse market trends to identify investment opportunities. 4. Work with management teams to develop financial strategies. 5. Monitor financial performance and point out discrepancies vs. forecasts. 6. Study economic and sector regulations to assess their impact on company business. 7. Submit financial recommendations to management. Her job Let’s use Victoria, a financial analyst as an example→
  23. Tasks where generative AI can be used Changes in the

    tasks of Victoria’s job Generative AI can simplify a certain number of tasks and save time for other higher value-added assignments. Her job will migrate towards tasks which require intrinsically human skills such as critical thinking and communication. Review financial data and prepare annual and monthly reports. Determine long-term financial forecasts. Analyse market trends to identify investment opportunities. Study economic and sector regulations to assess their impact on company business. Work with management teams to develop financial strategies. Study economic and sector regulations to assess their impact on company business. Submit financial recommendations to management.
  24. How will Generative AI impact the future of enterprises? Generative

    AI will bring significant transformations at almost every level Supply HR IT Finance Marketing CRM Sales Today Medium term Long term Design products Write product descriptions Create purchase orders Scan CVs Preselect candidates Write job descriptions Respond to candidates' questions Write employee evaluations Generate, fix, comment on code Summarize files Document the code Detect spam Suggest system performance improvements Check consistency Provide insights Benchmark costs Identify potential savings Identify attractive markets for expansion Brainstorm marketing ideas Create advertisements Personalize advertisements, newsletters, etc. Create a fully individualized customer journey Analyze customer options Respond to customer complaints Proactively approach inactive customers Call cold prospects Customize sales pitches Offer promotions Support customers Create product videos Create training material
  25. New professional opportunities as new jobs emerge Head of AI

    and Automation: Will determine the strategy, implementation and development of AI and automation initiatives within the company, identifying opportunities where these technologies can improve and optimise existing tasks and processes working hand-in-hand with company teams. Responsible AI use Manager: This manager is responsible for developing, implementing and adopting responsible use of AI systems within the company, whilst taking into account issues such as the company’s Code of Ethics, data transparency, reliability and traceability of AI generated data and content. They can also measure the psychological impact of the use of AI on individuals. Specialist in Art valuation and AI digital content: This professional values, selects and recognises the value of works of art generated by AI, and is also capable of detecting and analysing deepfake content. As a digital art conservationist, they organise art exhibitions for works created by AI tools, whilst ensuring their authenticity. Given their in-depth knowledge of image and video generation mechanisms, they are also responsible for checking that the digital content is authentic, often aided by AI tools, to protect the public from misleading information. Knowledge base expert: To take advantage of full potential available from generative AI within a company, Knowledge base experts are needed. Generative AI is based on processing existing data, therefore internal governance rules and processes are essential for listing, centralising and updating the vast quantities of available information.
  26. 30 To harness the potential of generative AI, you must

    know how to :speak to the machine!:
  27. Companies are already significantly assisted by generative AI In the

    short-term, the impact of generative AI will occur mainly via “internal” innovations which improve what we already do. But we will probably also see companies appearing whose business model is entirely based on generative AI technologies. Zelda in an infinite Open World For example, in the video game sector, generative AI could be used to create new game levels in real-time, offering an unlimited “open world” experience. Advertising agency using generative creation An advertising company could use generative AI to create customised visual content (models, photos, paintings…), thereby reducing costs and design time. Investment funds based on AI In the areas of decision-making, AI could minimise discrepancies and increase efficiency, opening the way for more lucrative business models. Specialised companies and ESN 2.0 with expertise in generative AI are emerging, offering services varying from training to maintenance. Supporting job transformation will also be crucial, involving the development of training modules to acquire the new skills needed to work with this technology.
  28. The near future will see job verticalization and enhanced IT

    sectors • The requirements of generative AI in IT infrastructure, namely as regards hardware, the cloud and infrastructure, strengthen companies in this industry such as Nvidia. • General AI solutions such as those offered by the digital giants can be both costly and risky. We see companies such as Mistral and IMOK emerge, who specialise in job verticalization and generative applications for specific sectors. • The ecosystem of design, installation, and maintenance is currently creating new jobs. Roles such as “Prompt Engineer”, “AI Integrator” and “Generative Experience Designer” are emerging to solve ergonomic or user experience issues and rationalise queries in the field of generative AI within organisations. • A boost in productivity for freelance jobs such as lawyers, consultants and graphic designers. AI also opens new opportunities in training, promoting the acculturation of emerging technologies. In the medium-term, generative AI will increase the demand for advanced infrastructures, boost service jobs thanks to specialised solutions, and create a need for experts within companies. AI is not only a technological breakthrough, but it will also stimulate a new way of defining the needs and roles in different sectors.
  29. The long-term future will see generative AI and quantum computing

    used together opening up opportunities for advanced business applications Building synthetic brains Autonomous space probes AI trained to help handicapped children Manikins to assist firemen Artificial intelligence motivator The future combination of generative AI and quantum computing promises major advance with higher computing powers. These two technologies used together will enable the creation of highly efficient AI assistants, more advanced business applications and machines capable of operating totally autonomously. They will adapt their work to fit the context, making business systems more agile and responsive. For example, in the future could we develop new vaccines in the space of a couple of days when faced with a new virus? Furthermore, progress in building synthetic brains using quantum computing could challenge our conceptions of creativity and intelligence, whilst spurring important ethical issues regarding the control of such advanced technologies. The end of Moore’s law
  30. Will companies opt for the rebound effect or the 4-day

    week? Generative AI will improve employee productivity, saving time durably provided there is an internal conversation about the correct ways of using this time, implementing tools and setting up training programmes. The time gained will drive companies to make important structural and strategic decisions: will they adopt the Jevons paradox (rebound effect) or decide to embrace the 4-day week? Regardless of the rebound effect of increased productivity, there is huge potential for greater added value which companies must decide how to share out fairly between employees, clients, investors, and the planet… Now is the time to ask the right questions. Indeed, the Jevons paradox implies that the use of this technological breakthrough may lead to increased use to stretch employee efficiency even further encouraging them to complete even more tasks in the same amount of time. On the other hand, companies may also choose to reallocate this time to promote a better work-life balance to improve staff retention and promote employee well-being.
  31. The big reward is better skills, the danger is downskilling

    Thanks to artificial intelligence the time spent on repetitive tasks has now been freed up, this time can now be used to maximise reskilling and upskilling. Companies can now choose to use the time saved to accelerate the development of their employees’ skills, company wide. This new model placing the employees as learners at the centre of the process has significant resource and cost implications, requiring initial investments in training programmes and advanced teaching tools. The long-term impacts for businesses are positive: a more skilled and more versatile workforce. Employees will be better equipped to face the challenges posed by the market, to innovate and, all in all to contribute to the company’s success.
  32. Social impact The digital divide may widen Are we moving

    towards employment polarisation? As more and more tasks are automated, will middle-skilled jobs gradually disappear? This trend could lead to a society where the winners will be those who use these technologies with ease while the others will be pushed aside, as their skills are made obsolete. The danger lies in the fact that not everyone will benefit fairly from this revolution. If we do not succeed in training and upskilling everyone faced by these new technologies, the risk is that we create a profoundly unequal society, where only a select few will reap the benefits of AI . “Employment polarisation is a real risk. AI will destroy middle-class jobs and managers with added-value roles will reap the benefits at the expense of lower paid, less skilled employees. AI will decimate middle managers if they are not retrained.” Arno Pons Delegate General Digital New Deal Think Tank “The greatest risk is that not everyone benefits from these transformations and that we do not have the means to upskill everyone.” Pierre Deheunynck, Chairman of the Board of France compétences,
  33. Whilst the technological feats of generative AI are indeed impressive,

    their environmental impact is by no means meagre due to the huge amount of energy, water and electronic components such as graphic cards, that they consume. There is little or no transparency from AI companies as regards their energy consumption so the calculation of their model's carbon footprint cannot be easily benchmarked. The environmental impact of these models is far from neutral Environmental impact Sources, research papers: On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?, ESTIMATING THE CARBON FOOTPRINT OF BLOOM, Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models, Carbon Emissions and Large Neural Network Training However, independent studies have done so and have published consistent outcomes: • 700,000 litres of fresh water to cool down servers and 552 tonnes of carbon emissions just for training ChatGPT3, which is the equivalent of 205 return flights between Paris and New-York • plus the daily impact of users’ daily queries must be added: 500ml of water for every 20–50 questions on ChatGPT3, whilst the carbon impact has yet to be calculated • plus regular retraining with new data. One solution could be that each company creates its own specific uses, more selective databases and language models to replace these greedy general AI applications.
  34. Generative AI may intensify cyber-risks Generative AI technologies will increase

    the risks of cyber crime in many areas: creating content for phishing e-mails, manipulating models by poisoning data which alter the algorithms and produce false data, violation of personal data, creating deepfakes, theft of intellectual property or malicious use of content generated for phishing, amongst others. Business impact To prepare for these new risks and threats , several steps need to be taken but will probably not provide sufficient protection: • Review or adapt cybersecurity processes and protocols • Strengthen authentication to avoid new cases of identity theft. • Review backdoor and front door protection measures to deal more efficiently with password hacking by generative AI technologies • Constantly retrain staff on the new risks faced by the company’s daily use of AI.
  35. 1 Generative AI… 2 Will change employment tasks but will

    not massively wipe out jobs. Will evolve organisations in the long-term. Costs, complexity, and the inertia of large corporates to efficiently deploy strongly impacting new technologies being the main reasons, much like structural digital projects which take three to five years to complete. 3 Stretches beyond the mere technical deployment, impacting companies’ strategy, their economic model and their organisation. Now is the time to ask the right questions and reflect on how AI should be integrated into governance. 4 Requires a systemic approach involving several business lines working together rather than embarking on isolated initiatives. An approach which does not require radical change, simply because the most relevant and impacting uses are still to be identified. 5 Makes way for upskilling. The benefits of generative AI make investments in employee training possible so as to develop their expertise, increase their engagement and offer new opportunities both for them and their employer. Talents who can boast proficient use of AI tools and technologies and AI skills will have a strong competitive advantage for recruitment in tomorrow’s world.
  36. Knowledge Democratize access to knowledge with few questions, enabling an

    understanding of the realm of possibilities and fostering a mindset of continuous learning. Time saving Save a considerable amount of time and focus on delivering value and quality service to the customer. Self-improvement Allows us to practice and self-assess by simulating questions on the issues encountered. Hyper-personalisation Enables the creation and display of content and solutions that precisely address the needs expressed by the user. Keep in mind The benefits of Generative Artificial Intelligence
  37. EY | Building a better working world EY exists to

    build a better working world, helping create long-term value for clients, people and society and build trust in the capital markets. Enabled by data and technology, diverse EY teams in over 150 countries provide trust through assurance and help clients grow, transform and operate. Working across assurance, consulting, law, strategy, tax and transactions, EY teams ask better questions to find new answers for the complex issues facing our world today. EY refers to the global organization, and may refer to one or more of the member firms of Ernst & Young Global Limited, each of which is a separate legal entity. Ernst & Young Global Limited, a UK company limited by guarantee, does not provide services to clients. Information about how EY collects and uses personal data and a description of the rights individuals have under data protection legislation are available via ey.com/privacy. EY member firms do not practice law where prohibited by local laws. For more information about our organization, please visit ey.com. © 2023 EY Fabernovel – All rights reserved. SCORE France N° 2023-084 - ED None Credits: Midjourney/@hGabha, Netflix, Adam Nadel/AP Images, Universal This publication has been prepared for general informational purposes only and is not intended to be relied upon as accounting, tax, legal or other professional advice. For specific questions, please contact your advisors. ey.com/fr About EY Fabernovel EY Fabernovel is a leading international provider of consulting services for strategic transformation and the creation of innovative services. Fabernovel was founded in 2003 by Stephane Distinguin, and became EY Fabernovel on 5 July 2022, following a merger with EY Consulting to become the leader in the convergence of digital and ecological transitions in Europe. The multidisciplinary teams include developers, designers, creatives, analysts, data engineers, marketing specialists bring their convictions and solutions across the entire digital value chain, from the consultancy stage through to the completion of products for daily use, and on to marketing campaigns and promotion of transformation strategies. The team: Stephane Distinguin, Robin Bruniaux, Benjamin Fallot, Cyril Vart, Diego Ferri, Joachim Martin, Yi Cao & Marine Chatras. External contributors: Pierre Deheunynck, Arno Pons, Pierre-Yves Calloc'h.