It doesn’t seem like AIs are accelerating at all if you ask me. We seem to be plateauing. Smaller and open weight models are catching up to the closed weight frontier models on benchmarks. If AI labs were able to use the smartest models to accelerate their development, the likes of OpenAI and Anthropic would be accelerating away from their competitors. But no such thing is happening. The focus has shifted from intelligence to cost and speed.
The breakthroughs in mathematics are impressive. But it doesn’t feel that different from what machine learning has done with Chess, Go and protein folding. They’re finding patterns in our systems and in nature. That’s what they’ve always been good at.
Mostly I'm seeing breakthroughs in the ways we use LLMs. Agentic harnesses, MCPs, etc - its the wild west still but we've come a long way from a basic chatbot. Gains are now coming from tools that make better use of the LLMs existing capabilities, and put guardrails on their worst tendencies.
I personally feel like we've plateaued in raw model intelligence (even regressed, I find sonnet 4.6 to perform better than Opus 5) but we've gifted them new skills that allow them to run for longer and explore the search space more thoroughly, making them more effective at the same level of "intelligence".
Take two smart people and a problem to solve. Give one person the tools, the other person nothing. The one with the best tools wins. At some point its more about abilities than raw intelligence. Watching a "frontier" model fumble with basic syntax is still common, but not if you give it treesitter.
We are also using these things to like 1% of their ability. The amount of manual process in the world that could be documented once and then automated is staggering. Decoding and transitioning such process is time consuming, and change is slow until fast.
> So let’s go with the 5%. Or, in plain English: we pretty much don’t have a clue about how the world really works.
World is not same as the universe. We may not know about 95% of the stuff that makes up universe, but that doesn't mean we don't know much about our world. World is what matters to us as earthlings. We knew a lot about what's available to us - resources, information etc.
- please rename your substack from signal and chaos to signal and noise but unfortunately noise is all i see
- i went through your list of posts and there is nothing but one opinion after the other with no hard facts to back any of the claims
- i see what you guys are doing there and even here on HN "riding the AI wave"
- honestly i would appreciate if someone could give far more balanced takes with actual facts to back them up
- your point "ai is accelerating....." explain to me how the same architecture transformers is accelerating? infact after the leading frontier labs pull all their gimmicks in this year and maybe the next, they are going to run out of gas unless they invent a new architecture
No one is calling out the bit where the guy goes full Mulder:
> A few days ago the US Department of War released the fifth tranche of records related to objects that appeared to have a faster mode of transportation. Anyone taking UFOs seriously used to be ridiculed. Not long ago the disclosures started. Will the implementation of “alien technology” be next?
>> In other words, he and most of his contemporaries were convinced that science was pretty much done. This was of course before Einstein’s theories of relativity, quantum mechanics and today’s fruitless search for dark matter.
>> Science wasn’t done then and it’s not done today.
Ah, the good ol' "someone nearly 150 years ago made an incorrect assertion, which means that anybody that ever makes a similar assertion ever again is automatically wrong" argument.
14 comments
[ 0.32 ms ] story [ 31.1 ms ] threadThe breakthroughs in mathematics are impressive. But it doesn’t feel that different from what machine learning has done with Chess, Go and protein folding. They’re finding patterns in our systems and in nature. That’s what they’ve always been good at.
I personally feel like we've plateaued in raw model intelligence (even regressed, I find sonnet 4.6 to perform better than Opus 5) but we've gifted them new skills that allow them to run for longer and explore the search space more thoroughly, making them more effective at the same level of "intelligence".
Take two smart people and a problem to solve. Give one person the tools, the other person nothing. The one with the best tools wins. At some point its more about abilities than raw intelligence. Watching a "frontier" model fumble with basic syntax is still common, but not if you give it treesitter.
Finding patterns in data is a pretty succinct description of what intelligence does, no?
This is relevant, I think
World is not same as the universe. We may not know about 95% of the stuff that makes up universe, but that doesn't mean we don't know much about our world. World is what matters to us as earthlings. We knew a lot about what's available to us - resources, information etc.
- i went through your list of posts and there is nothing but one opinion after the other with no hard facts to back any of the claims
- i see what you guys are doing there and even here on HN "riding the AI wave"
- honestly i would appreciate if someone could give far more balanced takes with actual facts to back them up
- your point "ai is accelerating....." explain to me how the same architecture transformers is accelerating? infact after the leading frontier labs pull all their gimmicks in this year and maybe the next, they are going to run out of gas unless they invent a new architecture
> A few days ago the US Department of War released the fifth tranche of records related to objects that appeared to have a faster mode of transportation. Anyone taking UFOs seriously used to be ridiculed. Not long ago the disclosures started. Will the implementation of “alien technology” be next?
>> Science wasn’t done then and it’s not done today.
Ah, the good ol' "someone nearly 150 years ago made an incorrect assertion, which means that anybody that ever makes a similar assertion ever again is automatically wrong" argument.