33
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 23 Sep 2024
33 points (100.0% liked)
TechTakes
1437 readers
174 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
Was salivating all weekend waiting for this to drop, from Subbarao Kambhampati's group:
Ladies and gentlemen, we have achieved block stacking abilities. It is a straight shot from here to cold fusion! ... unfortunately, there is a minor caveat:
Looks like performance drops like a rock as number of steps required increases...
correct me if I’m reading this wrong — the results are that LLMs are much, much worse than classical AI at planning block placement for SHRDLU? that seems pretty damning
Yes, the classical algo achieves perfect accuracy and is way faster. There is also a table that shows the cost of running o1 is enormous. Like comically bad. Boil a small ocean bad. We'll just 10x the size and it will achieve 15 steps inshallah.
Imo, this is like the same behavior we see on math problems. More steps it takes, the higher the chance it just decoheres completely. I can't see any reason why this type of thing would just "click" for the models if they are also unable to do multiplication.
I mean this just reeks of pure hopium from OAI and co that things will magykly work out. (But the newer model is clearly better^{tm}! I still don't see any indication that one day that chart is just going to be 100s across the board.)
I feel this shouldn't at all be surprising, and continues to point to Diverse Intelligence as more fundamental than any sort General Intelligence conceptually. There's a huge difference between what something is in theory or in principal capable of, and the economics story of what that thing attends to naturally as per its energy story.
Broadly, even simple things are powerful precisely because of what they don't bother trying to do until perturbed.
Ultimately, I hypothesize the reason why VCs like the idea of LLMs doing simple things far more expensively than otherwise is already possible, is because, They literally can't imagine what else to spend their money on. They are vacuous consumers by design.
So what you're saying is that Tim Apple could save us from these people by selling Marc Andreessen a billion dollar iphone?
It's Veblen goods again, isn't it?
Honestly, Yes. The hardest thing for a rich person to do is spend their money. Eventually this catches up with them: to spend no money is to lose it comparatively, to spend money is to risk not getting it back. So a great deal of the money world revolves primarily around persuasion, and the very odd things that happen along the way.
It also helps to recognize how much many of these people see all of this as a competition, and trying to out-unique/out-possess their peers