Back
Our experienced team is dedicated to providing strong support at every stage of your project
To promote greater balance in the workplace, we combine excellence and independence. We merge the high standards of traditional consulting firms with the flexibility, accessibility, and technology offered by our platform
Thanks to a community of 4,000 meticulously selected independent professionals, OMS offers various solutions tailored to your needs.
We meticulously select consultants and executives, focusing on academic excellence and professional achievements.
Are you looking for a new balance in your career? Are you seeking to join a dynamic and stimulating network?
LinkedIn
Privacy policy
Legal notices
© 2023 All rights reserved.
Accueil - AI, these professional experts help us shift our focus toward GenAI
Click on one of the contributors to read their interview.
Dimitri Carbonnelle
Expert in environmental issues and technologies
Claire Lelièvre
Data & Sustainability specialist
Grégoire Mialon
Post-doctorate researcher in AI
Marceau Longeanie
Chief Product Manager, OMS
If we are to stay the course, the Artificial Intelligence (AI) revolution is just a question of time. 65% of business leaders in France have made AI a priority investment with 88% of them expecting a return on investment within a period of 5 years*.
While GenAI shows promising prospects for future development, it also credibly makes us apprehensive with respect to the limitations of its use, with the risk of mistakes, bias, ethical issues, data security and confidentiality problems, not to mention the related costs of energy and finances.
With this in mind, mastering this tool is one of the cornerstone challenges for today’s companies. To achieve this, companies need to ask the right questions. For instance, which solution should be deployed, what boundaries need to be set (in terms of cybersecurity, ethics and investment), and what is the right strategy? That’s where our freelancers come into play!
At OMS, we help our customers tackle data issues by leveraging the expertise of our network. The latter comprises professionals who are qualified to deal with complex issues and are on hand to assist you in your shift toward GenAI, combining agility with performance while placing people at the center of every process.
Below, 3 OMS freelancers specializing in AI have their say:
Last but not least, Marceau Longeanie, OMS’ Chief Product Manager, offers 4 top tips to support your teams in their transitional use of AI, intended to boost the performance of your business.
Still keen? Contact OMS today and let us know your questions and needs!
Happy reading!
Grégoire Mialon is a post-doctorate researcher in AI who currently works in Meta’s AI research lab.
I first came across AI in 2016 during my studies as an engineer. To me, it was magical that a computer could learn a task, merely from examples. What’s more, the fact that this model could program concepts and programs familiar to us as humans that naturally develop in its neurons was pretty neat. On top of that, the industry-scale apps demonstrated a range of specs and much promise. In 2018, I wrote a thesis at l’INRIA¹ on how to build models that would operate on very limited training or learning data. In parallel, Open IA and other players were toying with a different approach, training a fairly simple but large-scale model, leveraging vast swathes of generic data which did not necessarily correlate with the issue at hand. We would then make tweaks to resolve the specific problem². Upon validation of my thesis, I joined Meta’s AI research lab to concentrate on these methods, which included Large Language Models (LLMs)³. Meta is developing a LLM which they have branded as “Llama.” There are two stages involved in training LLM to conversate. An initial phase to learn a vast set of data. This is followed by training to follow instructions and ensure the ability to converse with the intelligence.
Right now, at Meta, I am training LLMs to deploy tools that answer, what for them, are unfathomable questions based solely on the data on which they have been trained. As an example, I teach LLMs to activate a search engine whenever their training is outdated in answering a specific question, e.g., identifying the winner of the 2023 Rugby World Cup, i.e., information that is too “current“ to be included in training data. We can also train this intelligence to use other tools, including calculators, to steer clear of miscalculations. Until now, this was made impossible since models lacked the capabilities to know when and how to deploy tools, providing a stumbling block to myriad use cases. This will radically change the landscape of LLMs.
When it comes to productivity, AI is a substantial growth driver for companies, with its ability to unearth strategic data from colossal textual corpora. That said, with AI, there’s a price to pay both regarding energy and financials. At present, the challenge for companies is to grasp the capabilities of LLMs, pinpointing exactly what alternative existing and less expensive methods are incapable of. Conversely, companies also need to identify what is beyond the scope of LLMs such as devising a comprehensive marketing strategy since LLMs are incapable of thorough reasoning and planning and are restricted in terms of creativity. You can deploy LLMs to automate readily defined tasks. More complex tasks require an interactive assistant in order to accelerate the output of your work albeit with a human continuously in the loop.
Multiple instances of collaboration exist between France and the United States since many a French person holds a position in America’s large AI labs. Yann Le Cun, who heads up Meta’s AI department, is the high-profile case in point, but there are plenty more examples. What’s more, there is a fascinating ecosystem taking shape in France. Meta, Google DeepMind and Apple have all established major AI hubs in Paris. French startups such as Mistral are being created by former employees of American AI labs. HuggingFace was founded by French people, partly via American funds. As such, Paris is a strategic location for global-leading researchers, not to mention funding. By way of example, Mistral already completed a €100 million capital-raising project; Poolside, a company kick-started by an ex-Github employee, has set up shop in Paris, and Xavier Niel is investing hundreds of millions of euros in a number of AI ventures including the all-new lab – Kyutai. Significant discussions are also ongoing with the French government, with Bruno Le Maire and Jean-Noël Barrot fully aware of the latest AI developments. The United States has taken a step further than Europe on the regulatory front, with its recently published Executive Order which provides for enhanced controls and monitoring of AI models (e.g., maximum number of calculations for training models, surveillance systems to detect potential misconduct and misuse). All this means that France and the United States are working closely. Europe is smaller in size, but not without its assets. For instance, Paris is a strategic hub for AI, as is London which is home to the headquarters of Google DeepMind, with its hundreds of internationally acclaimed researchers. On top of that, talks held with European regulators appear to be constructive.
For me, personally, AI presents more opportunities than dangers. This intelligence could help us discover new medicine and materials. It could also help us gain clearer perspective on the universe we inhabit and generate considerable gains in productivity terms. As with all technologies, misuse is unavoidable and tricky to anticipate in the current market. Long-term, fears abound that LLMs will outsmart humans and take over our planet, but this scenario is extremely hypothetical: the devil is in the scientific details, and for now, no evidence exists to suggest LLMs will develop such capabilities. Granted, the aforementioned issues are engaging, but there are others less talked-about that are very real. One example is the potential impact of US regulations, with the emergence of an AI monopoly fronted by companies like OpenAI. On the flip side, Meta has taken the stance that the field needs to promote open-source platforms with universally accessible tools. On a different note, how will AI affect automation for businesses? And will it impact misinformation?
When performing complex tasks, you need to consider AI as an interactive assistant that can help you get your work done more quickly albeit with a human continuously in the loop.
We cannot predict the future. My view is that models such as GPT will keep improving in the years ahead, which will increase their efficiency and reliability, and this includes images and videos. These models will therefore be even easier to use in companies. Thereafter, we may hit a stumbling block. For some researchers, we will enter a cycle where LLMs successfully produce their own, augmented, training data and will use such data to continuously improve. If I were to hazard a guess about the future direction of research, I would suggest that these models still lack autonomy. Indeed, they are not autonomous learners, which means we simply cannot throw them in the deep end and expect them to learn on their own. They struggle to adapt to new contexts and scenarios. In this respect, my cat is smarter than them!
Claire has worked a freelance consultant for over a decade, collaborating with startups, not-for-profits and large corporations, mainly in the services sector. She specializes both in customer experience (CX) and sustainability.
I have worked in consulting for more than a decade, collaborating with large corporations that mainly operate in the services sector (banking, mass catering, energy, telcos) as well as startups and not-for-profits. My work has focused on digital technologies with an emphasis on customer experience (CX) in addition to matters relating to sustainability. I started work in the AI field seven years ago as part of a project to set up a neobank with AI featuring prominently. I was responsible for supporting the development of banking services and led a three-person team to design all the CX models. I also took charge of building an AI-powered customer relationships platform. In just one year, we engineered a mobile app with support from a banking technology platform that included 40 technology partners and a team of a hundred or so developers working in agile mode.
With a keen interest in digital technologies and the robust capabilities of agile teamwork, I have since continued focusing my energies on large-scale projects that combine technology, agility and a customer-first approach. I spent a year working for a major US company where I was tasked with setting up an innovation lab that would use AI to leverage customer knowledge. Given my intention to make ecology the central focus of my work, I quickly developed an interest in how AI can be used to address environmental challenges such as the fight against global warming, biodiversity protection and water conservation. I worked at the Techstars Sustainability Accelerator in the State of Denver, hoping to fully immerse myself in these issues. For just over a year, I have worked with a global-leading player in mass catering and facility management, tackling digital, consumer and sustainability issues as part of this multinational’s Global Digital&Data team.
Any large corporation seeking to scale up investment in Sustainability depends on access to high-quality data. The development of a strategic action plan for the environment involves a stage for self-assessment so as to understand potential ways of taking action. Assuming your objective is to reduce carbon footprint, you would need to determine the exact sources of emissions. This type of exercise requires you to cross-reference various sets of data in order to plan for potential areas of improvement based on your company’s situation. Reverting to the mass catering, facility management company mentioned previously, you could track the following Key Performance Indicators (KPIs) to measure impact: how much you reduce carbon footprint, water consumption and food waste.
In more practical terms:
1. More advanced AI-powered reporting tools can be deployed to track KPIs and gain better insights into the impact of the company’s decisions. Real-time reporting can also be utilized to flag up alerts caused by potential operating anomalies which adversely affect the environment, e.g., unusual energy consumption.
2. AI can also make proactive decisions and take related action. In the above example, AI would be capable of identifying peak consumption times and would directly respond by lowering the temperature on the premises. Longer-term, assuming AI learns enough knowledge, it can take a precise and proactive approach to optimize how companies manage their energy consumption.
3. As regards the connection with customers through digital solutions, AI enables more accurate insights into customer needs and expectations. The data supplied to customers to guide their decisions, as consumers, is absolutely vital. One notable example is online food shopping: AI makes it easier for customers to access the right data, at the right time, nudging them to consume food more responsibly. Whether proactively or reactively, AI can provide detailed information on the impact of food products. A case in point is AI’s analytical capabilities when it comes to the composition of dishes, issuing consumer data such as the carbon footprint of food, the product origins of dishes, how much water was consumed to produce the food and how the food will affect your health.
AI can proactively suggest alternative solutions, such as price reductions on specific items. Consumers can even ask for personalized recommendations that cater to their needs: how to eat vegetarian, how to source food more locally, etc. Customers are free to choose whether they want such assistance. When deployed in a way that is respectful of humans and nature, AI takes on the role of fully-fledged companion, aiding companies and customers in the generation of models to consumer ever more responsibly.
Protecting our planet and all its species is the cornerstone challenge of this century, both for people and businesses. Human activity – the reason for global warming and the loss of biodiversity –will undergo drastic changes over the coming years, either through informed decision-making or out of necessity. Companies failing to diminish their impact and deliver people- and eco-friendly solutions will find themselves in a serious predicament that will jeopardize their long-term future. This predicament involves a growing lack of interest from informed consumers, difficulties in attracting talent and criticism for practicing greenwashing. Companies that opt for transparent communications and attain carbon neutral status, or even generate a negative carbon footprint, are the true trailblazers in today’s world.
Disruptive innovation calls for new business models with technology at the helm to fast-track wide-scale change.
Admittedly, AI represents disruptive innovation that is highly robust and capable of revolutionizing every single way of working. That said, this intelligence will not enable companies to thrive nor humanity to last unless an industry-leading CSR approach is taken to support its use. Impactful disruptive innovation calls for new business models with technology at the helm to fast-track wide-scale change. Awareness is skyrocketing and AI’s core strength lies in the fact that it can analyze unlimited streams of data to help humans make much faster decisions. Human society now realizes that not many years are left to act, with companies cornered into finding urgent solutions, in order to properly prepare for this battle.
AI, Customer Experience (CX) and Sustainability. Over a decade of consulting experience with large corporations, mainly in the services sector (banking, luxury, fast-moving consumer goods (FMCG), energy, telcos), as well as startups, and is a Shifter on the Shift Project.
I worked for multinationals such as Hewlett Packard, General Electric and SFR with a focus on strategic projects before establishing my own consulting firm, Livosphere, which specializes in new technologies. I am inspired by the idea of anticipating radical technological changes, particularly those brought about by AI: what are its exact capabilities? And how will it reshape our society?
These past three years, my work has become increasingly focused on climate and environmental issues, including the notion of “digital sobriety.” I have written a book entitled “2050 : Crash ou Renaissance” (“2050: Crash or Renaissance”) which explains how we can adapt our society in response to climate challenges by factoring in technology. In this book, I also outline the capabilities of AI, as well as its associated risks and potential abuses.
In an article I wrote a few years back entitled “Jamais sans mon iLad” (“Not one day passes without my iLad”), I recounted a short story to illustrate how AI could normalize our behavior. Based on the concept of self-reference, we will increasingly put our trust in AI with its improving ability to predict our future since we will prefer to follow its advice as opposed to getting it wrong ourselves! The risk involved in this is that we stop taking initiative and lose the ability to overcome setbacks, not to mention our creative faculties! Life would become incredibly dull because we could easily predict it from birth.
In 2024, there is a growing fear that AI will supplant humans, eventually taking us over. Personally, I see the risk of us automatically depending on AI, because it is convenient and even because we are lazy. We notice how AI often gets things right and we end up trusting it, no matter what, even though this intelligence is sometimes wrong. This issue is known as hypovigilance, and it’s when humans no longer check AI’s solution due to overconfidence and mistakes are then made which can cause devastating impact. Yes, AI can get things 99% right, but remember, 99% does not equal 100%! This can have a dehumanizing effect with humans acting as links in a chain controlled by AI which instructs them on what to do.
Within an industry, if an operator regularly checks that AI conducts quality assurance on product manufacturing and observes no problems 99% of the time, they could allow a faulty part to slip through the net, as it were. In the case of vehicles, the effects could be fateful. What does this mean? There is a need to test people’s vigilance by triggering controlled errors in the system to monitor this hypovigilance and by making people more critically aware in their activities using AI. Naturally, the error in question would be flagged with the operator immediately afterward!
Another critical factor is deciding where to integrate AI and robots. I worked for a leather goods company, analyzing their handbag production line to identify tasks that could be automated using robots. We then decided not to automate specific repetitive tasks, which leather workers needed to ease their mental workload. However, we did automate the bag wear test with its requisite set of repetitive actions by the thousands! Humans are essential to any company process introducing AI and robots.
As things stand, the learning phase for neural networks calls for substantial computing power and data. Transfer Learning¹ makes it possible to lower AI’s impact on energy and resources. It uses a generic AI model, which is then adapted to each stakeholder’s individual needs, considerably reducing the most energy-intensive learning phase. As an example, in the auto industry, highway concession companies, cities and public authorities could be interested by AI that can recognize various vehicle models using cameras. Instead of separately training different AI models, they could utilize an open-source model before tailoring it to their personal needs at a later stage. This would dramatically cut down energy consumed by servers.
There are several ways to deploy AI with a view to cutting back energy consumption. AI can anticipate particular flood risks, calculate a roof’s solar energy potential, reduce building energy consumption, and optimize travel for maintenance teams. The use of drones with heat-mapping cameras and built-in AI can also identify buildings most in need of thermal renovation as a result of heat loss. AI is capable of analyzing a tremendous number of images and delivering solutions with precision. Not only does this intelligence help you prioritize, but at the same time, it channels and redirects flows of capital and resources wherever they are required.
Moreover, AI presents a competitive advantage for reporting. It formats data collected by companies from a range of sources so that it matches the layout of CSR reports including the CSRD², markedly shrinking the administrative workload. It goes without saying that data must still be checked as it will not be 100% accurate.
Not only does this intelligence help you prioritize, but at the same time, it channels and redirects flows of capital and resources wherever they are required.
Some students are using AI to boost the skills and knowledge they already have. They are incorporating it into their work and are taking the right approach. In other words, they start the work themselves before completing it using CHAT GPT or other AI tools. They also draw on this intelligence to perform tedious tasks such as writing summaries, drafting plans and monitoring intelligence while remaining critically vigilant and checking AI for its sources and errors. Nonetheless, another cohort of students takes the answers given by GenAI like GPT-4 for granted, without verifying the quality and double-checking for “hallucinations.” As mentioned previously, the risk lies in the fact that students get used to working less hard; they stop learning for themselves and let silly AI errors slip through the cracks.
AI method used to transfer knowledge acquired from solving one problem and applying it to another.
Chief Product Manager, OMS, IT specialist in project management, network infrastructure, blockchain & Lean Management
Artificial Intelligence (AI) is reshaping the professional landscape. To fully harness the benefits of this technology, we must first identify the challenges faced:
AI is a tool, not a substitute!
More often than not, Artificial Intelligence (AI) is regarded as a threat to jobs and positions historically held within companies. However, the rationale behind this outlook is much too simple. The reality is that most forms of AI – especially those currently used on a routine basis – are intended to complement rather than substitute human expertise. Typically described as “low-standard,” these forms of AI are specialized and optimized to perform structured tasks. A case in point is their ability to quickly analyze vast datasets, spotting patterns and trends and making predictions based on such analysis.
That being said, these forms of AI are not capable of forming judgments and lack the intuition, creativity and empathy which characterize human beings. Instead of considering AI as a substitute, it would be wiser to see it as a powerful tool. Used correctly, AI can increase efficiency and productivity and even boost levels of satisfaction in the workplace, by relieving humans of mundane tasks which lets them focus on higher value-added activities.
AI and data protection
With respect to AI, the issue of confidentiality and data protection is key to companies’ digital transformation. The increased deployment of AI to analyze and process data and make predictions calls for extra vigilance over how such information is managed. Often, versions of AI, particularly those made by private companies such as ChatGPT or other language models, operate by tapping into huge data streams in order to produce relevant results. This dependency on data raises a number of credible concerns: who exactly has access to this information? How is the data stored and protected? Do you run the risk of a leak or breach of data? These sources of preoccupation become all the more urgent as sensitive or confidential data can be publicly disclosed whether by accident or design when we interact with these AI systems. Against this background, the introduction of stringent data management protocols is an absolute must for companies, with their employees made aware of the importance of discretion. Additionally, companies must coordinate continuous training on best practices in information security. Deploying AI does not necessarily mean forgoing all aspects of confidentiality, but it does call for proactive management so that data protection is guaranteed in every step of the process.
AI and ethics
AI ethics is an issue that goes above and beyond technology; it resonates with the basic values of our society. In the context of Artificial Intelligence (AI), ethics refers to how we use these tools. Questions arise regarding our intentions of this use and the resulting effects. AI’s ability to process gigantic sets of data and make predictions can prove tempting for less virtuously-minded applications, whether to influence public opinion, heighten surveillance, invent fake news or any other uses that serve to violate civil rights and liberties. Furthermore, AI can inadvertently perpetuate existing biases and inequalities if it is trained on data reflecting such biases. With this in mind, companies crucially must implement responsible practices, taking time to reflect on the right questions before implementing an AI-powered solution. This also involves establishing safeguards to prevent any abuse, actively pursuing continuous improvement while factoring in feedback from the community and experts in the field. Ethically-driven AI should not be seen as a luxury; rather, a necessity, to ensure use that is respectful of and beneficial to one and all.
« Just as we would not let an untrained driver get behind the wheel of a car, we should not let unqualified staff interact with, or depend on, AI systems. ».
AI and the need for training
Without a shadow of a doubt, proper AI training forms one of the foundational pillars for organizations to confidently navigate today’s digital landscape. Just as we would not let an untrained driver get behind the wheel of a car, we should not let unqualified staff interact with, or depend on, AI systems. Proper training not only affords employees with the skills needed to effectively utilize AI, but also develops their understanding of the limitations and potential errors found within these systems. Rather than putting our blind faith in these tools, people are encouraged to critically analyze AI and make informed decisions. The scope of AI training extends beyond its mere use, encompassing a good grasp of ethical issues, privacy risks, and the wider ramifications of AI in the company’s area of expertise. By supplying employees with the requisite knowledge, you improve a company’s safety and efficiency. And the icing on the cake – you foster a culture centering on innovation and versatility, both of which are indispensable for success in this new digital age.
Deploying AI calls for proactive management so that data protection is guaranteed in every step of the process.
Artificial Intelligence (AI) promises to revolutionize the world of work, just as it did so in other fields. Nevertheless, like any significant innovation, AI is not without its share of challenges and opportunities. If we embrace AI – as a valuable tool and not some threat – by overseeing data protection and privacy, by placing ethics at the heart of everything we do and by investing in proper training, we will not just avoid the pitfalls, but also scale new heights for our companies and businesses. As we embark on this journey, a balanced and informed strategy is paramount, to guarantee a future where humans and machines work hand in hand, building a brighter future together.
Partager cet article :
Prendre contact
Contact us