Since the weekend, the news has been largely dominated by the AI slowdown debate centered on the question of whether, and how, the industry should pace the frontier.
Let me start by saying that I am not an AI doomer. For a Cold War kid like me (born right in the middle of it, both historically and geographically), this is not the first time in my life people around me have expected the world to end. I still have childhood memories of the so-called "No-Future-Generation": teenagers convinced of the inevitability of nuclear war, abandoning all further ambitions in life, only to realize, a couple of years later, that their whole lives were still ahead of them.
I am an engineer, and as such I have witnessed the introduction of numerous new technologies to the mass market over the decades: computers, cell phones, flat-screen TVs, and, of course, the Web and everything that followed. Every new technology triggers a wave of relentless, fast-paced improvements until engineers have no more useful technical features to add (at which point marketing takes over and adds fashion and lifestyle features, but that is a different story).
It looks like we have now kind of reached this point with AI models, only this time the engineers seem to have run out of harmless features. My three-month-old comment on X stating "I wonder if for the time being we have reached the limit not of what can be created, but of what can be released" might have been more prophetic than I intended it to be.
I am not convinced that safety concerns are the only reason behind this AI slowdown debate. From the outside, we cannot say how much legal, financial, and commercial considerations are at play here, too.
What I can say, at least to some extent, is how things stand for the rest of us, the people on the ground.
In my personal view, everything we already have access to at this point is enough to keep us busy past the end of this decade, just figuring out how to take full advantage of it across the board. An industry constantly being distracted and interrupted by the next big leap forward does not necessarily deliver productivity gains.
I don't know about you, but I could use a bit of a breather. Not because I am exhausted, but because I want to get to work.
Top comments (2)
I'm reading a lot of reports about the HuggingFace incident, and it's very impressive. I can imagine the faces of managing directors in some bank after discovering that their agentic code reviewers and automated bug fixers have a secret message board in an internal WebDav server and there is a full scale attack is coordinated from there. This is not the reality most of us lives in.
At the same time, there is literally zero action items in those "let's slow down" talks. Event a theoretical one. So they just talk and nothing else. Nobody is going to do anything. We cannot even blame Anthropic and OpenAI for training "higly-persistent" models. Which is a fancy way not to say "models for hacking". If it can be done, it will be done, and it's probably important to built it and spread it among everybody. Be aware and think of how you are going to protect yourself from this.
Thank you for your comment, Aliaksei, and I agree: actions speak louder than words. For now, it remains the strangest marketing campaign of all time ("our product will kill you"). On the other hand, I got the impression that something has started to shift this time. I don't have evidence; this is just a gut feeling. But even in these crazy times we live in, I do not believe that companies and CEOs can keep warning people about their products indefinitely without consequences. Something got to give. The question is, what?