Hey thanks for replying! It’s a great question, and one I don’t have a good answer for.
My immediate gut instinct is that we SHOULD still be teaching kids coding, and everything else that was on their curriculum previously. Just as calculators helped us work faster, we still want to understand what they’re doing. Even if AI can augment or replace some of our knowledge in future, it’s still important to think for yourself and be able to vet what you’re generating with AI. But this is an answer built around what’s happening right now, not tomorrow.
When I think about an answer in relation to my experience over the last year, it feels a lot more hopeless, at least to me. I don’t think we’re going to know what adults should be focusing on in the short term, let alone what kids are learning for their future careers.
The situation is changing so rapidly in my workplace (at least for software engineers) that AI workflows we were using 3 months ago feel ancient. By not being on the latest frontier model, we’re losing productivity. We’re all rolling our own custom software to speed up our personal workflows. It’s faster and cheaper to do this in many cases than to buy an off-the-shelf product and learn/master it.
So already, I feel like the days of companies selling software subscriptions feels very numbered. Adobe and Atlassian are going to have their business model completely destroyed IMHO. But that’s not even the scary part.
Have a look at the number of Linux CVEs found this year compared to previous years - it’s humbling. Governments are paying very close attention, and export controls on AI models are in full swing. They’re very capable cyberweapons.
So the future I’m seeing is an incredibly uncertain one for everyone due to the exponential nature of it all, both from a geopolitics standpoint as well as a career standpoint for many industries. Last year I thought we were somewhere through the hype cycle and it was starting to slow down, but now I’ve had more hands on experience with the frontier models, these days I’m thinking we’re still at the beginning of an exponential curve, an explosion of change that’s going to happen so fast many of us will struggle to cope, and our slow legal and economic systems will struggle to keep up.
I think a lot of people are just assuming it’ll all blow over once the bubble pops, but like all the other bubbles, it’s formed because there is a very real technological leap happening, and it’s not going away.
So right now I’ve never been more sure of one thing: that it’s harder than ever to see what’s coming. So to try and answer your question just feels impossible, it’s all still moving too fast.
Perhaps on reflection one thing we should be teaching our kids is the importance of continued learning and bettering ones self. I fear the world is in for a rough decade from all this (not to mention climate change and late-stage capitalism), and anything we can do to prepare our kids to roll with the punches in that future would be wise.
Thanks for such a detailed response. I suppose asking you to essentially predict the future of this particular industry was a bit ambitious.
My immediate gut instinct is that we SHOULD still be teaching kids coding
As is mine, which for me I think, even if I take it as read that LLMs are the future of programming, comes from the same place that makes me think in spite of Python, programming courses should start with C. But it’s become pretty obvious to me that a lot of people don’t have that desire to “go under the hood”, and still seem to manage, so part of me thinks maybe the same will become true of LLM users.
these days I’m thinking we’re still at the beginning of an exponential curve
a lot of people are just assuming it’ll all blow over once the bubble pops, but like all the other bubbles, it’s formed because there is a very real technological leap happening, and it’s not going away
I’m one of the people you called out who “hasn’t used a paid frontier AI model to write code” (though I didn’t downvote), so I have to take your word for it re the rate of change, but it does seem a little contradictory to me to believe we’re on an exponential slope but also in a bubble. I guess the former is from a technological perspective and the latter financial?
Have a look at the number of Linux CVEs found this year compared to previous years - it’s humbling.
That’s one of the things that has pushed me away from the intensely skeptical end of the spectrum. But I don’t really see how this can fail to flatten out eventually. While secure software is possible at all (by which I suppose I mean cryptography, hardware, and airgaps still work as advertised), I think if anything AI will simply accelerate us towards an equilibrium, where the supply of vulnerabilities that can be found by LLM-guided brute force search is patched and exhausted.
I fear the world is in for a rough decade from all this … and anything we can do to prepare our kids to roll with the punches in that future would be wise.
Regarding the frontier models, there’s a idea called Karpathy’s Funnel which notes the massive difference in experience with AI models. It funnels the population of the world like this: (borrowing a lot from a Karpathy tweet with some paraphrasing)
6B people (~75% of the population) have barely come in contact with LLMs.
1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and AI slop, to this part of the world the whole thing feels crazy still.
20M people (0.2%) see frontier-grade LLMs, complex projects that used to take them weeks/months can now be completed by agents with a prompt. You typically need a budget to get this experience. There are multiple levels in this tier as well from those that use ‘open-weight’ models to expensive monsters like Claude Fable. Models are Powerful tools at this level.
5k People (~0.00006%) have access to frontier-grade systems internally inside AI labs, they have the budget to put the frontier on automatic, spin up hundreds of agents, and burn their huge salary’s worth in tokens.
At these lower tiers of the funnel, they believe in the future the AI companies promise because they’ve literally bought in enough to actually get the experience.
Hey thanks for replying! It’s a great question, and one I don’t have a good answer for.
My immediate gut instinct is that we SHOULD still be teaching kids coding, and everything else that was on their curriculum previously. Just as calculators helped us work faster, we still want to understand what they’re doing. Even if AI can augment or replace some of our knowledge in future, it’s still important to think for yourself and be able to vet what you’re generating with AI. But this is an answer built around what’s happening right now, not tomorrow.
When I think about an answer in relation to my experience over the last year, it feels a lot more hopeless, at least to me. I don’t think we’re going to know what adults should be focusing on in the short term, let alone what kids are learning for their future careers.
The situation is changing so rapidly in my workplace (at least for software engineers) that AI workflows we were using 3 months ago feel ancient. By not being on the latest frontier model, we’re losing productivity. We’re all rolling our own custom software to speed up our personal workflows. It’s faster and cheaper to do this in many cases than to buy an off-the-shelf product and learn/master it.
So already, I feel like the days of companies selling software subscriptions feels very numbered. Adobe and Atlassian are going to have their business model completely destroyed IMHO. But that’s not even the scary part.
Have a look at the number of Linux CVEs found this year compared to previous years - it’s humbling. Governments are paying very close attention, and export controls on AI models are in full swing. They’re very capable cyberweapons.
So the future I’m seeing is an incredibly uncertain one for everyone due to the exponential nature of it all, both from a geopolitics standpoint as well as a career standpoint for many industries. Last year I thought we were somewhere through the hype cycle and it was starting to slow down, but now I’ve had more hands on experience with the frontier models, these days I’m thinking we’re still at the beginning of an exponential curve, an explosion of change that’s going to happen so fast many of us will struggle to cope, and our slow legal and economic systems will struggle to keep up.
I think a lot of people are just assuming it’ll all blow over once the bubble pops, but like all the other bubbles, it’s formed because there is a very real technological leap happening, and it’s not going away.
So right now I’ve never been more sure of one thing: that it’s harder than ever to see what’s coming. So to try and answer your question just feels impossible, it’s all still moving too fast.
Perhaps on reflection one thing we should be teaching our kids is the importance of continued learning and bettering ones self. I fear the world is in for a rough decade from all this (not to mention climate change and late-stage capitalism), and anything we can do to prepare our kids to roll with the punches in that future would be wise.
Thanks for such a detailed response. I suppose asking you to essentially predict the future of this particular industry was a bit ambitious.
As is mine, which for me I think, even if I take it as read that LLMs are the future of programming, comes from the same place that makes me think in spite of Python, programming courses should start with C. But it’s become pretty obvious to me that a lot of people don’t have that desire to “go under the hood”, and still seem to manage, so part of me thinks maybe the same will become true of LLM users.
I’m one of the people you called out who “hasn’t used a paid frontier AI model to write code” (though I didn’t downvote), so I have to take your word for it re the rate of change, but it does seem a little contradictory to me to believe we’re on an exponential slope but also in a bubble. I guess the former is from a technological perspective and the latter financial?
That’s one of the things that has pushed me away from the intensely skeptical end of the spectrum. But I don’t really see how this can fail to flatten out eventually. While secure software is possible at all (by which I suppose I mean cryptography, hardware, and airgaps still work as advertised), I think if anything AI will simply accelerate us towards an equilibrium, where the supply of vulnerabilities that can be found by LLM-guided brute force search is patched and exhausted.
Pretty bleak. But I definitely agree.
Regarding the frontier models, there’s a idea called Karpathy’s Funnel which notes the massive difference in experience with AI models. It funnels the population of the world like this: (borrowing a lot from a Karpathy tweet with some paraphrasing)
At these lower tiers of the funnel, they believe in the future the AI companies promise because they’ve literally bought in enough to actually get the experience.