The Dual Threat of AI: Job Displacement and the Need for Human Oversight

Welcome back to Market Movers.

The opening bell.

We are standing at the edge of a structural phase transition in the global economy.

4 years, the artificial intelligence narrative has been about automation and efficiency, but could the market be mispricing the real threat, AKA the measurability gap.

Now as AI drives the cost of execution.

The new bottleneck isn't necessarily intelligence.

It is trust and human verification.

Now the danger is a mountain of unverified debt quietly building up inside our financial systems as agents scale faster and that our capacity to actually audit them and as the dominant revenue model shifts from selling software to underwriting liability, how and where to investors.

Find the real alpha while joining me to weigh in this morning is Christian Catalini, founder of MIT crypto Economics lab.

Good morning, Christian.

Thank you so much for joining me.

So let's start out with this idea of counterfeit utility.

Tell us what exactly it is and what investors should actually be paying attention to when it comes to.

Thank you and good morning.

So I think we're all experiencing the rapid acceleration of capabilities with AI.

Uh, Especially over the last few months, uh, really starting in December, model capabilities have jumped ahead when it comes to coding and engineering tasks, so we're probably entering a period of fast takeoff, uh, where, uh, you know, we will be able to do a lot more with the technology.

But as a result, there's also a side effect of it, which is as the models become more capable, our capability to really monitor them, verify and check their work doesn't scale at the same speed.

We're still bottlenecked by our, by our biology, by our tools, and so that gap between what the models can do and and what we can safely.

Really, you know, certify is increasing that can create accumulated risk and debt that may come due at a later date.

I think we're seeing early traces of that as companies ship software and then they realize that, you know, some of that software has major vulnerabilities.

A few days ago we've seen a Chinese company release a private key that was really critical to their system in the wild.

Um, I think we'll see a lot more of this, and so companies need to pay a lot more attention to how they scale these agentic systems going forward. and Christian, many of us have heard the term SAS and you point out that we're moving from software as a service to liability as a service, meaning that companies will be valued on how well they handle risk and not just output.

So practically speaking, what does a business leader need to change about their business model today in order to survive that.

Yeah, so the terrifying part of the of the change is, of course, that software is very easy to generate and so many have been worried that traditional software as a service, a big part of it is really, you know, resting on the ability to ship better code.

In the wild that serves customers' needs is not a moat anymore.

Now that's not fully true, you know, things like distribution, uh, network effects, the scale, and the understanding of your customer needs still really matter, but I would say in a world where agentic capabilities really scale, the critical dimension that every CEO needs to keep in mind is, are we scaling our verification capabilities at the same pace as our as our ability to really ship new products and services.

Um, that's what we call the liability as a service.

You're not just selling software, you're often selling labor, right?

So you can essentially use gentech workflows to recreate what usually would have taken maybe a number of employees to do, uh, but you cannot do that if you also don't understand the potential risk of what you're shipping.

Uh, and so the ability to underwrite that agentic code, that agentic output, is becoming, I, I think, the core new mode, and investors should pay close attention to the companies that are investing in verification infrastructure versus the ones that are not. and you bring up an important point because we are all paying attention to the labor market, especially since it is that day today, and we are awaiting that press conference with thatcher Powell, but you outline a dual threat on the talent side, so artificial intelligence is wiping. entry level jobs.

The junior analyst roles where people actually learn the ropes.

So senior experts may be basically relying on coding their own replacements.

But if you're running a major institution, how do you stop this collapse in human oversight?

So it's, it, it's a tension, right?

And we're seeing major restructuring, uh, recently, for example, Block, right, laid off 40% of, of their staff.

I think we're going to see a lot more of that, especially in tech.

Tech is the sector that is the closest to the transformation happening on the engineering side and the capabilities of these models.

So I would expect many tech companies, and we're hearing rumors about meta and many others, uh, being among the first ones, uh, to move on this, uh, but it's going to be a broader phenomenon.

I think people will realize. that they can do more with less people and probably the most challenging part of this is that if you're, you know, fresh out of college and you're just starting your career, you're really at the level of capabilities often that the models are at.

Now, of course the models are getting better, so that level of capabilities keep rising and rising, and even very senior experienced, uh, you know, staff, staff people will have to really increase their capabilities and use the tools, use the models to lift themselves up the value chain.

Uh, so for a CEO dealing with this transformation, I think it's important to realize that you need a constant pipeline of talent that will be able to, uh, you know, scale those models that scale them safely and with the right verification capabilities.

You will likely need the very top talent in your industry as the models get better because that talent will have a lot more leverage than it did even in the past, and you'll have to fundamentally reshape also how you train people.

Now the good news is that I think This a lot of the same tooling that is creating disruption in the labor market that is displacing things we used to do is also creating the tooling that can get us back to the frontier.

So learning will be much faster, mastery will be much faster, and so we're going from a period where maybe your apprenticeship as a fresh out of college was that period where you would learn kind of the extreme educators of that profession, whether in law, medicine, finance, to essentially learning this at an accelerated rate through AI tools.

OK, Christian, well, we will have to leave it there for today.

Thank you so much for joining us as always and thank you so much for sharing all of your insights and your perspective.

Thank you so much, Remy.