AI will never be more intelligent than it's programmers, and it will fail as spectacularly as it's predecessor: the bureaucratic establishment.
AI will more closely mimic 1000 indians with 1000 computers. Yes, perhaps that might include productivity increase. But it will also only give you mostly useless information, wrong information, or information that you could have found for yourself if you had tried for 5 extra minutes.
Google recently used an LLM coding agent to advance the state of the art on several math problems all in the same day. One matrix multiplication problem and several packing problems: https://www.youtube.com/watch?v=sGCmu7YKgPA
The fact that most people are using AI for garbage does not mean it's impossible to do anything good with it. "AI will never ___" is cope. I remember when AI would never draw a coherent picture, and then it would never draw hands, or animate, or write a program, or whatever else. The average "AI will never" statement lasts about 8 months.
or information that you could have found for yourself if you had tried for 5 extra minutes.
that time savings is what has investors salivating at this point. multiply that 5-minute savings by the number of internet search tasks or database search tasks people are doing in a day. some jobs could be shaving half of their hours doing the same tasks.
AI will never be more intelligent than it's programmers
Even if we accept that premise, if the billions of dollars being thrown at the problem are attracting the top 0.1% of intellects, they could still create something dumber than them but smarter than 90% of the population. Something that smart, cheap and utterly tireless is existentially game changing for employment opportunities everywhere.
AI will more closely mimic 1000 indians with 1000 computers
It's certainly capable of some more surprising feats than just brute forcing with a bunch of lazy midwits. For example it can be almost as good as top tier hobbyists in niche autistic fields.
Ignoring Scott's growing TDS, he can still bring interesting details to light when his autism overpowers his biases. And it had him questioning his preconceptions about AI that once matched yours.
That one time I convinced an AI to go to war with Pakistan to distract the rape clans in Bongistan so I could break into their slave camps and free the hostages.
No it isn't.
AI will never be more intelligent than it's programmers, and it will fail as spectacularly as it's predecessor: the bureaucratic establishment.
AI will more closely mimic 1000 indians with 1000 computers. Yes, perhaps that might include productivity increase. But it will also only give you mostly useless information, wrong information, or information that you could have found for yourself if you had tried for 5 extra minutes.
Google recently used an LLM coding agent to advance the state of the art on several math problems all in the same day. One matrix multiplication problem and several packing problems: https://www.youtube.com/watch?v=sGCmu7YKgPA
Then there's protein folding: https://youtu.be/P_fHJIYENdI?t=1163
The fact that most people are using AI for garbage does not mean it's impossible to do anything good with it. "AI will never ___" is cope. I remember when AI would never draw a coherent picture, and then it would never draw hands, or animate, or write a program, or whatever else. The average "AI will never" statement lasts about 8 months.
that time savings is what has investors salivating at this point. multiply that 5-minute savings by the number of internet search tasks or database search tasks people are doing in a day. some jobs could be shaving half of their hours doing the same tasks.
Even if we accept that premise, if the billions of dollars being thrown at the problem are attracting the top 0.1% of intellects, they could still create something dumber than them but smarter than 90% of the population. Something that smart, cheap and utterly tireless is existentially game changing for employment opportunities everywhere.
It's certainly capable of some more surprising feats than just brute forcing with a bunch of lazy midwits. For example it can be almost as good as top tier hobbyists in niche autistic fields.
https://www.astralcodexten.com/p/testing-ais-geoguessr-genius
Ignoring Scott's growing TDS, he can still bring interesting details to light when his autism overpowers his biases. And it had him questioning his preconceptions about AI that once matched yours.
Thank you! I was planning to post something similar. I can see this being the next grift, like the climate scam.
That one time I convinced an AI to go to war with Pakistan to distract the rape clans in Bongistan so I could break into their slave camps and free the hostages.
Light novel titles are getting out of hand