Tanishq Mathew Abraham agrees but opposes slowing AI research.
As my entire feed is criticizing Anthropic, I think that the team there genuinely believes what they’re saying. It’s not a marketing/anticompetitive tactic. They genuinely believe these models are dangerous and that AI research should be slowed down.
Positive users defended Anthropic's principles and encouraged working there, while negative users dismissed their safety beliefs about AI models as insincere posturing or a cultish power grab.
@banburismus_ I liked your pitch.
if you want to use interpretability to make models better, rather than to secretly make them worse, come to @GoodfireAI
@finbarrtimbers well, other people's ai research should be slowed down
As my entire feed is criticizing Anthropic, I think that the team there genuinely believes what they’re saying. It’s not a marketing/anticompetitive tactic. They genuinely believe these models are dangerous and that AI research should be slowed down.
@finbarrtimbers Everybody should go work for them (if they're good 😅) - you should apply!
As my entire feed is criticizing Anthropic, I think that the team there genuinely believes what they’re saying. It’s not a marketing/anticompetitive tactic. They genuinely believe these models are dangerous and that AI research should be slowed down.
@finbarrtimbers So why are they not slowing down internally if they genuinely believe this?
@casper_hansen_ I think they feel the dynamics of the race are such that they have to be in the lead to control it.
@finbarrtimbers There are no beliefs. Only posturing. Their real intentions are to gain as much power as possible through whatever it takes. I think it’s really that simple.
@finbarrtimbers if they did, every other ai lab would see it as a chance to take over anthropic's sota model position on the leaderboards almost directly.
meanwhile there doubling down on models with bigger weights than opus.
thats like the opposite of safety and slowing down.
@finbarrtimbers agreed, but it's an unfortunate mindset to have imo
As my entire feed is criticizing Anthropic, I think that the team there genuinely believes what they’re saying. It’s not a marketing/anticompetitive tactic. They genuinely believe these models are dangerous and that AI research should be slowed down.
@finbarrtimbers remember when OAI did it wayback, history repeats
@finbarrtimbers Right but you can’t simultaneously believe AI research should slow down while accelerating your own. Thats saying one thing and then doing another
@immortaldip I don’t, actually! What are you referring to?
@irohsharpeniroh @finbarrtimbers Right but then the belief is that capable AI is dangerous in other people’s hands and not that research should slow down. In other words, they believe they should seek maximum power because it’s safer to be the most powerful.
@finbarrtimbers They are true believers.
In a technology they don't really understand.
@casper_hansen_ @finbarrtimbers no this is a coherent position. if they have confidence in their internal controls and visibility then it might be reasonable to say "I can safely advance research" while not trusting "arbitrary third party" to advance research safely
@MartinShkreli @finbarrtimbers I think I just spent my tokens for the week on trying to get it to read a PDF.
@finbarrtimbers Dario and a huge portion of the team defected from Sam/OAI for a reason. It was a high risk to start a new venture when ChatGPT was basically the only game in town. People can think it's misguided, but the people calling it all "marketing" are clearly misguided.
If they truly believe in that then they won't be releasing the model publicly.
Public release allows them to collect a vast amount of data and that essentially starts the recursive loop to better the model, all the while preventing others from bettering their models and in turn reducing competition.
This might a classic anti trust case brewing, given the power the company already has in the market and about to go public for a Trillion dollars and doing shit like this.
@MartinShkreli @finbarrtimbers a ragebait machine
@AdvicebyAimar Every other lab is trying to take over Anthropic’s SOTA model position! It’s an aggressive competition.
@finbarrtimbers true, they have perfected over years on lesswrong how to rationalize their god complex
Tanishq Mathew Abraham agrees but opposes slowing AI research.
As my entire feed is criticizing Anthropic, I think that the team there genuinely believes what they’re saying. It’s not a marketing/anticompetitive tactic. They genuinely believe these models are dangerous and that AI research should be slowed down.