Vitalik Buterin pfp
Vitalik Buterin
@vitalik.eth
https://x.com/PratyushRT/status/1873428775171670435 Things like this are a big part of why I'm skeptical of just putting a single AI in charge of things (eg. DAOs, funding mechanisms), and favor an open market where many AIs can participate and humans judge AIs based on some sample of outcomes.
14 replies
187 recasts
762 reactions
vrypan |--o--| pfp
vrypan |--o--|
@vrypan.eth
I think we should treat AI agents the same way we treat human agents. For example: 1. Tests and exams that lead to certifications: test how they perform in specific scenarios, just like we would test a new hire, a driver, or a doctor. 2. History: An agent's history can be treated as a long term test. It's an agent's CV: "This model has been working for us for 10 years, it has dealt with these cases, and it has done ok, with the exception of cases of type X." 3. Stake: An agent should be able to bear the consequences of bad decisions. A stake is a way to implement this, and also to measure the amount of trust that can be put on one. (Stake could also have the for of insurance, or something similar).
1 reply
2 recasts
9 reactions
J. Valeska 🦊🎩🫂 pfp
J. Valeska 🦊🎩🫂
@jvaleska.eth
this is another point for not considering any LLM open sourced if the whole process of training is not open sourced and transparent too.. AIs need more transparency.. aside, I think this post is trying to create a narrative against open sourced LLMs.. like.. they are something "bad" created by unknown people versus the trust of private LLMs created by well-known brands.. anyway.. I think the answer is open sourcing the whole process.. or creating the LLMs in the open source from scratch.. I already saw some of these projects where everything is transparent and people can train the LLM through interactions..
1 reply
0 recast
7 reactions
Ferran 🐒 pfp
Ferran 🐒
@ferran
It’s wild how the whole AI x DAO and AI x Crypto thing is being hyped up without stopping to seriously think about the flaws or the consequences. Like if they don’t exist. To me it just feels like another hype cycle. Giving people false hope that this will magically fix crypto issues. While VCs capitalize an idea that, deep down, they probably know is kind of dumb. I mean, really… handing over governance, one of the most complex human realities, to an AI? Locking everything onchain so it can make unchangeable decisions, and then putting these AI agents also in charge of funds, microservices, even hardware? Things that could have a real impact on humans lives. What could go wrong, right?
1 reply
0 recast
1 reaction
casslin.eth pfp
casslin.eth
@casslineth
doesn’t make sense he stresses this threat particularly on open source llms. How abt closed source ones? Just trust bro??
0 reply
0 recast
0 reaction
David (d/acc) pfp
David (d/acc)
@promptrotator.eth
What do you think of AI owning wallets?
0 reply
0 recast
0 reaction
compusophy pfp
compusophy
@compusophy
the largest hack hasn't happened yet
0 reply
1 recast
0 reaction
Heikki Arponen pfp
Heikki Arponen
@harpone
Well you can't really put today's AI (LLMs) in charge of anything since they don't actually want to do anything (except predict the next token sort of).
0 reply
0 recast
0 reaction
Lauren McDonagh-Pereira pfp
Lauren McDonagh-Pereira
@lampphotography
Yeah that is scary. We can't depend completely on something that is so easily tampered with.
0 reply
0 recast
0 reaction
Moon pfp
Moon
@moon.eth
Had to deal with some open source AI for specific tasks at a previous job and had severe difficulty with political bias in the training. If you just put an off the shelf AI in charge of something you will probably get a lowest common denominator result.
0 reply
0 recast
0 reaction
Joshua Salles pfp
Joshua Salles
@omghax.eth
Yeah it's concerning the amount of unknowing CEO's replacing entire workforces with AI when we can't get it to stop hallucinating or being racist 🫠
0 reply
0 recast
0 reaction
Metaphorical pfp
Metaphorical
@hyp
Even sample outcomes might not reveal the 0.1% of where poisoned results hide.
0 reply
0 recast
0 reaction
D-wayñe 🎩🕴 pfp
D-wayñe 🎩🕴
@drrrner
the latter seems more ideal
0 reply
0 recast
0 reaction
Aaron Blaisdell / ittybit 533 pfp
Aaron Blaisdell / ittybit 533
@aaronblaisdell
/microsub tip: 782 $DEGEN
1 reply
0 recast
0 reaction