Vitalik Buterin on Warpcast

Vitalik Buterin pfp

Vitalik Buterin

My impression of chatgpt so far is that it's good at replicating existing human behavior, but bad at doing anything out-of-sample. It anchors quickly onto patterns that have been done by many people before, but there's definitely something that we call "original thinking" that it's not doing at all.

17 replies

0 recast

3 reactions

Vitalik Buterin pfp

Vitalik Buterin

As a random example, consider these AI-generated debates (credit to pmarca for the format). Notice that the Chinese perspective is not even consistent across the two debates! It's pattern matching to the usual yes/no debate, and... https://i.imgur.com/OebQ19c.jpg

3 replies

0 recast

1 reaction

Joey DeBruin pfp

I agree with you, but historically speaking this distinction hasn’t held up well, right? People said AI generated classic music “lacked soul” but then when blinded weren’t able to tell it apart from Bach. Your examples are correct but I’m just not sure how long this holds up.

1 reply

0 recast

0 reaction

Andy Jagoe pfp

💯it seems unable to do new things seems underwhelming at first, but is the current day equivalent of the flying shuttle or spinning jenny in that it reduces human schlep this is a big deal, just as it was in the industrial revolution and makes me think calling it LLM vs AI would reduce headwinds

1 reply

0 recast

0 reaction

ace pfp

@perl add it to my collections #gpt3 #vitalik #ai go.perl.xyz/gpt3

0 reply

0 recast

0 reaction

Ben - [C/x] pfp

@perl AI GPT3 vitalik

0 reply

0 recast

0 reaction

dylan pfp

At its core it seems like a great way to get simple answers or get help with coding problems that aren’t too complex. Search is the largest use case I see right now - almost as if Google was your assistant. Excited to see real core utility happen & paradigms shift, not just a new AI plugin for every app

0 reply

0 recast

0 reaction

Michael Pfister pfp

Michael Pfister

So still a potential disruptor to Google search, but not strong AI yet. When do you think AIs will have “original thinking”?

0 reply

0 recast

0 reaction

VIPER pfp

Hey! I've recently purchased your cast! You can now claim your 🎩 DEGEN tokens on jam! 🟣 https://jam.so/ 🟣

0 reply

0 recast

0 reaction

sude pfp

0 reply

0 recast

0 reaction

Rmz.stark pfp

Interesting to see the application, will it helps indirect original thinking? Did google search engine helped ?

1 reply

0 recast

0 reaction

PhaeDog pfp

That’s pretty much all AI models presently available.

0 reply

0 recast

0 reaction

Julian pfp

100%! I deeply believe that future is about creative AI. The only question is which variable can judge the success or failure of creativity. Complicated

0 reply

0 recast

0 reaction

Benjamin pfp

That was my experience playing around with the original release of GPT-3 — if you labeled the conversation as a chat bot, it started sounding like a phone tree. If you described it with more human attributes, it would become more verbose and natural.

1 reply

0 recast

0 reaction

Altuğ pfp

AI literally can’t think outside the box and limited with current human feats.

0 reply

0 recast

0 reaction

Gustaf Kugelsurf pfp

Gustaf Kugelsurf

Sure, but 95% of what people do also isn't original thinking, so technologies like these will be able to replace almost all knowledge workers. The ones that wield the AIs will control the universe.

0 reply

0 recast

0 reaction

Warpmaster General pfp

Warpmaster General

You're not wrong but let's not gloss over the fact that it sufficiently -recreates human behavior-!

0 reply

0 recast

0 reaction

Loeber.eth pfp

Some commenters have been noting that LLMs are functionally an abstracted/generalized database technology, and I think it holds. When you’re looking at 100B+ parameter models, that reflects an enormous datastore with a good fuzzy search function over it. These things are good at lookup but not “reasoning”.

0 reply

0 recast

0 reaction