We're at the Inflection Point

 Sorry to break it to you guys, but the world is about to change. It's just a couple to five years and likely economies will probably change drastically. And yes, this is not the same as the industrial revolution, or anything else you draw analogies to in the past. This is one of a kind, and yes, AI is going to actually take jobs away. Not just take jobs, it's going to change the notion of economic value. I don't want to be an AI doomer, but spending a large amount of my time thinking about whether to be skeptical or concerned, having talked to many people on this, and discussing extensively with GPT/Claude, I cannot but shake this idea that things are going to change drastically.

How did we get here? So Dario's point a couple years back was on automating all of software by the end of 2026. Well, engineers are still employed, so haha, shame on me! Let's honestly ask ourselves, what part of our entire day is spent in truly doing work that is out of the box in terms of our cognitive capabilities. Let's be truly honest, and not allude to "this was hard", "that was easy" analogies. If you are authentic, chances are that unless you're in probably the top 1% of software skill wise (and this is probably the level people reach after more than ten years of software experience, or even more I guess), a lot of your work is doing "pattern matching". You've done something before, you apply a similar pattern here. Turns out AI is just the perfect tool for this job. It can do repetitive tasks fine, and has the cognitive reasoning capacity to fill in the gaps when things move away a little. So turns out that if you're not really doing cutting edge stuff, pushing the frontier forward, AI is coming for your job. Infact it's coming for the 99 percentile of all of knowledge work, because a lot of this work is not really out of the box thinking, where out of the box thinking sits in the thin tail of the cognitive compute distribution, say at the level of doing proper academic research.

But hey, the job of a tech person is not to just write code! Yes, and that's fair. But again, ask yourself, if your manager just needed updates on project status, and they need someone to trust so that they take ownership of code, how about the entire team contains only one architect who just reads Claude's reports instead of a junior telling them the status of work, and just communicates to management. Why is this just not plausible? But then "I also present my work and talk to clients". Again, that's fair, but if I give your clients an option to talk to an AI agent at a tenth of the price, delivering the same level of quality as the 99th percentile of people who they're used to interaacting with, what do you think they'll choose? 

But wait, AI can write code, it surely can't do sales or talk to people. I've chatted with people who've done full sales calls with AI, where it even offered a discount and was "very smooth", to quote their words. Sure, it could be nit-picking, but let me assure you that technologically we're there. Why do you think GPT solved IMO and not sales before that? It's simply because OpenAI directed data and compute to solving IMO, no other reason whatsoever. Tomorrow, if they decide, they can simply employ thousands of good salesmen to come in, show their tricks, give them data and GPT will become an expert on it in no less than a month. It's all about data, and compute, and of course, reinforcement learning. A fair bit of environments are essentially verifiable, and honestly people are getting real world outcomes in organic chemistry and drug design, human to human interaction is surely hackable.

I do believe all hope's not lost though. There will be jobs where the effective role of "empathy" will still be valuable, or maybe even go up. Human connection will be at the core of a post scarcity society. That being said, my greatest fear is that we might not end up in a post-scaracity society. Human greed will take over, and the rich will keep these LLMs to themselves, monopolize the shit out of the market, and call the shots on who gets what and does what. A sliver lining to this is open source catches up soon, and then everyone will have access to the best intelligence, which might bring up its own problems though, but atleast censorship and dictatorship not being one of them.

My point of this post was to make a prediction. I see FDEs going up, companies monopolizing the knowledge work market, China massively open sourcing all kinds of LLMs it's creating, until a point it feels that having this tech to itself is also massively valuable. At this point, every major power starts building it's own LLM tech in house. You have to, it's the new nuclear race. There will be massive AI policy decisions taking shape in the next few years, and all of it would be to distribute power and figure out the economics of a post AGI world. When 99 percentile of all knowledge work gets automated, what happens is a vacuum of meaning, and people need to find meaning to survive. I guess that's the point this post goes philosophical, a topic I'm not in the mood to discuss now.

I hope for a positive outcome through all of this, and that everyone ends up happy in a post AGI world. This was just a prediction, and maybe I'm in this echo chamber of talking day in and day out across Professors, PhD students and people at startups. So a prediction was imminent I guess. That's all this is about. It's not to spread any information/misinformation. Just trying to see if my words stand the test of time. That's all.