- This research process is the same as what I've been doing.
https://cmpr.ai/hutter/archive/
Claude does the research, like having an army of very fast but forgetful grad students working for free.
The individual paper doesn't make much incremental improvement but once you get a pipeline of results building in a particular direction it's powerful.
- if you click through to the linked Vannevar Bush article and scroll down there are a bunch of vintage ads around the prose that are kind of interesting. And some of the predictions have been well overtaken by events!
by lawrenceyan
1 subcomments
- Biggest update I see is that he thinks AI 2027 is actually going to happen.
by munificent
5 subcomments
- > We are entering a golden age in which all computer science problems seem to be tractable, insomuch as we can get very useful approximations of any computable function.
Alternatively, we are entering a dark age where the billionaires who control most of the world's capital will no longer need to suffer the indignity of paying wages to humans in order to generate more revenue from information products and all of the data they've hoarded over the past couple of decades.
> the real kicker is that we now have general-purpose thinking machines that can use computers and tackle just about any short digital problem.
We already have those thinking machines. They're called people. Why haven't people solved many of the world's problems already? Largely because the people who can afford to pay them to do so have chosen not to.
I don't see any evidence that the selfishness, avarice, and short-term thinking of the elites will be improved by them being able to replace their employees with a bot army.
- This looks like a survey. Is there a thesis; any claim?
by TacticalCoder
0 subcomment
- > As Rocks May Think
I thought they meant the plural of ASRock as in "ASRocks May think" and thought this was about ASRock motherboards getting a BIOS/UEFI with an integrated LLM or something.
by macrocosmos
0 subcomment
- > AI generated videos are indistinguishable from reality.
by alsetmusic
2 subcomments
- This person doesn't understand how LLMs work.
by measurablefunc
1 subcomments
- Yes, yes, we are all going to be living in an automated & luxurious communist utopia. Here are some material facts to ground the exuberance: 1) Lifecycel of typical GPU in a data center is 1-3 years, 2) Buildout is already limited by production capacity & will hit production walls by 2027-2028 when turnover matches & exceeds production capacity, 3) TSMCs projected capacity is ~130k wafers/month & it is not keeping up w/ demand which is more than doubling, 4) Power consumption for these "geniuses" & "thinking rocks" in data centers requires lots of power & the capacity was saturated in 2025, 5) So along w/ production capacity limitations power production is now another gating factor.
Anyway, like data centers in space there are lots of material limitations that all of these exuberant "ZOMG rocks can think now" essays all sweep under the rug to drive a very biased narrative about what is actually happening & the fact that all those binary bits are produced by real materials that have lifecycles & production limits not visible in the digital artifacts.
by mynameisjody
1 subcomments
- I'm still waiting for one of these articles to be written by someone without something to be directly gained by the hype. Eric Jang, VP of AI at 1X.
by xyzsparetimexyz
4 subcomments
- [flagged]
by kalterdev
2 subcomments
- > Chief among all changes is that machines can code and think quite well now.
They can’t and never will.
Hacker News