About this talk
This talk explores the impact of artificial intelligence on human relationships and collective productivity. The speaker argues that while AI enhances individual performance and speed, it simultaneously undermines collaboration and trust within teams. They highlight Goodhart's law, which states that optimization of measurable metrics can degrade their true value. Through case studies of AI-based educational projects, the speaker illustrates how focusing on shared goals rather than individual metrics can foster meaningful connections among participants. Ultimately, the discussion raises critical questions about the nature of technology's role in society and emphasizes the need for AI to serve collective human interests.
Full transcript
We forgot to ask who AI was supposed to be smart for. 10 years of making it faster, more powerful. Is it working? Yes. Okay, we're back to it. And the question of who it was actually serving barely made into the brief. Since ChatGPT launched in 2022, we've heard endlessly about augmented human beings. How many times have you heard about an augmented society? For the past few years,
teams have become more productive, deliverables shipped faster, individuals are more autonomous. And yet, if you ask those same teams what has changed in the way they work together, many will tell you that something had been lost. Not in the actual output, but in the fabric of human relationships. We were sold a revolution. And yes, it is one. Just not maybe the one we think. This is not
a technological revolution, this is a human revolution. One that is redefining how we think, how we behave, how we work, how we learn, how we interact with one another. And at the core of it, what values guide us to ensure that we remain human. And that revolution hasn't really begun. Because we actually asked the wrong question. We actually asked, how can AI make each individual more productive?
What we never asked was, what is it doing to the collective in the meantime? That is not a criticism. It's actually what we've done is actually optimizing what we're supposed to optimize. The problem is what the brief never really included. There is a law in behavioral economics that says when a measure becomes a target, it ceases to be it ceases to be a good measure. This is
named the Goodhart's law. It's actually everywhere. If you look at it, if you look at productivity for instance, if we measure productivity by hours worked, everyone stays late everyone stays late in the office. If we measure engagement by the number of Slack messages, you might have actually your Slack on fire. The numbers keep on going up. The actual value does not necessarily follow. AI amplifies this mechanism
at a groundbreaking scale. It optimizes what is measurable. Execution speed, individual output, autonomy. What it does not measure is actually disappearing from the equation. And Yelena actually talked about it. We talked about trust. Trust is the first thing, shared meaning, intrinsic motivation. Not by intention, but because it was actually never in the brief. Studies are beginning to document the real cost of this. If you take for
instance a first study back to back in January 2026 on knowledge worker, the study showed that collaboration with AI improves engagement over over short-term period. However, the same study identifies that there is a collaboration paradox. So, if you take this study and you take it into the the longer horizon, intrinsic motivation erodes, well-being deteriorates, and fatigue sets in. There is another HBR analysis that goes further. Overall
team performance declines after AI integration. This is Goodhart's law at scale, at human scale. We optimized where we're supposed to optimize, which is the measure, and we degraded what it was meant to represent. Human relationships are difficult. They involve friction. Think about misunderstandings. Think about conflicting needs. Think about the actual work of repairing bonds. That friction is not a flaw in the fabric of human relationships. It
is what weaves it. AI is designed to eliminate that Interactions are smooth, non-judgmental, and free of any expectations of reciprocity. That is its true value proposition. There is a professor of MIT named Sherry Turkle that said precisely that by seeking to design perfect AI companions, we actually embark on a journey towards forgetting what it means to be human. So now let's look at the results of that
at the team level. We have now start a star-shaped networks where AI occupies the center, but where peer-to-peer connection is well, actually the connections where you can find trust, innovation, resilience actually dissolves. Various analyses of enterprise platform are documenting right now massive decline in colleague-to-colleague interactions in favor of centralized request to AI. At the society level, what it means a system designed to constantly validate the users
individuals believe does not build shared reality. It actually fragments it. As you were saying. Uh there is a former French minister in the 1800s named Alexi Togville um and what he said, he said that um and he identified that the capacity to associate and to build shared meaning is the beating heart of living democracies. And we were talking about democracies right before. So if we all believe
in democracy and in our humanity, which I hope we all do, we do need to deliberately design AI for the good of the collective and not only design it, but adopt it for the good of the collective. Today I wanted to share with you two of the projects that I actually uh launched recently. Uh and in these two projects, I tested the new approaches in how I
teach AI and how I actually use AI in my platforms. So the first project that you actually see on the left uh happened in January in January 2016 in Casablanca, Morocco. I was asked to train 100 students and NEETs. So NEETs are um young people not in education, employment, or training. And I was asked to train them on AI for 2 weeks. I could have decided to
done a simple program where uh the focus was on providing dozens of tools and exercises and methodologies and then the the idea was to would have been to aim for individual adoptions that would have answered their individual needs. Instead, even though this approach would have been fast, measurable, it would have been collectively useless. So what I decided to do is to take like another approach and to
use AI as a lever for a shared project and not as a personal tool. So, what we've done together for the with the students is that we actually decided to have a training focused on building an entrepreneurial project in groups. The constraint of that project was that they had to find solutions to solve um a real problem in their local environment. Most of the students, except for
ChatGPT, never used any of the tools that they were supposed to use during the training. So, it was kind of funny. However, what we saw is that during these 2 weeks, what happened was not only a technological adoption. What happened is that they articulated what they already knew collectively together. So, they used AI as a tool to make the solution come to life, but most of all
what they've done, they actually built built real relationships with their teammates, and most importantly, they regained confidence in themselves. In the second project that you see on the right with InterSkills mentoring platform, the the guiding question was how does AI strengthen the mentor-mentee relationships without claiming to replace it? Think about, for instance, a development plan that you need to build for a mentee. Should it be AI,
or should we leave it to AI to make the plan? So, you come, you set the goals, and AI magically generate the plan, or is it the role of the mentor-mentee relationship or the mentor and the mentee to build it together first, and then use AI to maybe improve it or make suggestions to make it more efficient? So, what we decided to do is that AI AI
scope for AI would um only going to deliver the content. The platform will organize the path. And the mentor helps the mentee reflect, apply, and grow. So, technology was not at the center of these two projects. The relationship was. I wanted to end slowly but surely this keynote with three questions for you guys today. The first one is what does the system do to the fabric of
human relationship? And was that in the brief? The second question is was this technology designed for the real user or the imagined user? And the distance between the two is actually the measure of the blind spot between what we and what we feel in terms of the real field experience that we actually have. The third one is does the system train the muscles of the collective or
atrophies them? This is not an ethical question. This is a resonance question. A team that no longer knows how to solve problems together will certainly be faster in familiar situation. However, it will be more fragile in new ones. technology, AI, but humanity and human beings. The crisis we are facing is not about the machines. It's about us. What we will choose to measure. What we will choose
to optimize. What we will choose to stop leaving out of the brief. What is at stake blah blah. What is at stake is a human revolution. What it becomes is a question of what we choose to value. Thank you very much. That was a great presentation. Um would you agree with the statement that we have a window of opportunity to make sure that AI supports us instead
of us supporting AI? Not sure. Okay. No, I don't I don't think that specifically when I work with a lot of companies, um SMEs, institutions, we look at it in the way we frame our prompts. We are starting to speak in their language while AI was supposed at first to speak our language. So, we are shifting in the way we are approaching AI. So, instead of actually
supporting our singularity, we tend to all of us to go towards a more standardized approach. So, I think it goes in the opposite way right now. Thank you so much. Will you be around later on? I think you said couple of very interesting hypotheses. I could imagine that the one or the other would have an interest to talk to you. So, feel free to reach out to
Leila. I think that's an interesting discussion to And it's way of way from the usual technology discussion we have in this group of people. So, that was really interesting. Thank you >> Thank you.
