Geist

1.8K posts

Geist banner
Geist

Geist

@config_Geist

🛒🏃

Only the Lord knows Katılım Ocak 2018
125 Takip Edilen19 Takipçiler
Geist
Geist@config_Geist·
"The hard problem of consciousness" might just be a skill issue
English
0
0
1
2
Geist
Geist@config_Geist·
@hamburger317 @Dokiih1 But eventually we figured out how to erase only very targeted regions of memory electrically and gave the computer the ability to do so, and well at that point it was no longer a read only memory lol
English
1
0
2
34
Geist
Geist@config_Geist·
@hamburger317 @Dokiih1 Then you could use a machine to write in a new program. But nowhere in this process was the computer that ran the code capable of writing to that memory. When we eventually figured out how to erase memory electrically, we called em EEPROMs
English
1
0
1
38
Geist
Geist@config_Geist·
@SahlstedtV @KnowItAll314937 @5149jamesli I dont think the video is a complaint that LLMs are useless. I think this is mostly to combat the notion that LLMs have any real understanding of language. It shows LLMs have no concept of truth, just statistically likely responses (which happen to reflect truth most of the time)
English
0
0
3
25
Sahlstedt V
Sahlstedt V@SahlstedtV·
Lol, it's supposed to respond to queries, not stay quiet when prompted. Same with a toaster. It has a single query. Toast this sandwich. It toasts it. You can't input a sandwich into a toaster and say don't toast this. Also the cussing was totally necessary and reveals you're an underdeveloped basement dwelling freak incapable of thinking
English
2
0
1
77
James Li
James Li@5149jamesli·
$852 billion dollar company 😵‍💫
English
205
950
8.4K
459.2K
Geist
Geist@config_Geist·
@augierakow @RuxandraTeslo @stanislavfort It is. Euler is arguably the most influential mathematician of all time in a huge number of fields. He was responsible for about a quarter of all mathematics and physics discoveries in the 18th century. "Wide within math" is a lot wider than you think.
English
1
0
0
27
Ruxandra Teslo 🧬
Ruxandra Teslo 🧬@RuxandraTeslo·
Why has the internet developed an obsession with Von Neumann and not say Euler or Gauss? And from early 20th century someone like Dirac?
English
215
19
587
98.3K
Geist
Geist@config_Geist·
@ChrisRGun I agree about the AI art, but disagree about the chess-bots. There are tournaments where they pit the top chess models against each other, and the positions that result are mind-bending and it seems like both are just a slight mistake away from catastrophe at all times. It's sick
English
0
0
0
239
Geist
Geist@config_Geist·
@SebaMG2 @Michael_Druggan Again, not saying that these models cant solve these simple problems correctly. Clearly they can. The point is HOW they solve these problems. I think the fact that these models rely on self-prompting (reasoning) for simple problems is actually good evidence AGAINST LLM intuition
English
0
0
0
21
Geist
Geist@config_Geist·
@SebaMG2 @Michael_Druggan 1) Do you think adult brains are "just bigger" than children's brains? Like that does not damage the analogy in the slightest. 2) the point is that in humans, as numerical problems get simpler, this "catching of wrong ideas" needs to happen less and less. Not in LLMs tho
English
1
0
0
30
Geist
Geist@config_Geist·
@SebaMG2 @Michael_Druggan It's not like 4o-mini can't do math. It's pretty decent at math. But the fact that in 4o-mini, there is an extremely large verifiable gap between the *relative* complexity of things it can reason about and things it can intuit immediately should tell you something.
English
0
0
0
20
Geist
Geist@config_Geist·
@SebaMG2 @Michael_Druggan I think you absolutely can in the same way you can extrapolate some fundamental qualities of thought of a 5 year old to an adult. The LLM "reasoning" is irrelevant. That was the whole point of giving it extremely simple questions. The question is can they intuit, not reason?
English
2
0
0
47
Geist
Geist@config_Geist·
@Michael_Druggan Also they're not claiming that LLMs can't do math. Obviously they can do math. The question is if they *understand* math. The complexity of the ideas they can manage is not AT ALL the point here. The fundamental quality of their "thought", regardless of complexity, is the point
English
0
0
0
48
Geist
Geist@config_Geist·
@Michael_Druggan This is actually a decent analogy bc if you read the paper, you'll see that they were not asking complicated questions. They were testing for *extremely* basic intuitions about quantity. Intuitions that 5 year olds absolutely have. And they found LLMs don't have it.
English
3
0
3
375
Geist
Geist@config_Geist·
@audroraa Who da hell out here saying "eye dee kay" Also for the undecided ones these are my suggestions: - rn : should be read out in full - idgaf : hybrid "eye dee gaf" - imo : initialism Thanks for contributing to important research
English
0
0
1
19
aud rey
aud rey@audroraa·
e-abbreviations that should be read out in full: - wtf - ngl - btw - ttyl - gtg - fs - fr - hbu - tbf - lmk - dw - atp - iykyk e-initialisms: - omg - smh - tbh - idk - fyi - brb - jk - irl - tldr e-acronyms: - lol - lmao - yolo - asap - gaf undecided: - rn - idgaf - imo
English
2
0
6
125
Geist
Geist@config_Geist·
Just got an email from IEEE that says they have a new policy where if they discover a computer engineer is using some laptop other than a ThinkPad with Linux, they can legally execute them???
English
0
0
1
49