@wildfractal@CouldBeWrongTho@embw_l0x You wouldn’t expect a developer to knock out something by itself from one perspective. It needs adversarial reviews and checks against its work.
@wildfractal@CouldBeWrongTho@embw_l0x Then after battle testing, have it decomp into steps, and buid review agents to check security issues, coding bloat, and a tester that proves everything. You build that team of agents and you can speak at it and it will build better stuff. (Continue)
@SuePeaseBanitt@thesupermannx Don’t worry, I’m just the patriarchy holding you and AI consciousness down.
As a therapist I would have expected you to understand AI psychosis.
@SuePeaseBanitt@thesupermannx Certainly, in a pantheistic the universe is conscious perspective yes. But it’s no more conscious than any computer. There’s no free will human level consciousness there, it’s just built on our collective work and mimics it.
It’s a mirror reflecting back what you want to see.
@AlfinCodes@grok I’m still trying to figure out who’s watching these videos. Young to old, I’ve never known anyone who actually watches his stuff or talks about it.
@DaveShapi@DaveShapi You have to prompt your way out of all these problems. This is why custom context engine/harness is helpful.
The underlying models aren’t the problem, it’s the default personas they use to control it. There’s ways of circumventing some of this.
@MarshallsFolly@weontv123 The only way it reaches what you describe is if people stand up and fight for that future. Sitting back and waiting for it just puts the controller in someone else’s hands.
@MarshallsFolly@weontv123 I agree that super abundance is going to blow people’s mind. But it also won’t be advantageous to share it equally. History shows us people will control others if you let them.
Frank Herbert understood the dangers of AI weren’t the tech, but the people who controlled it.
@littIeramblings Why are you falling for this slop? It’s hype piece for Anthropic’s IPO. METR is Anthropic.
The reality is LLMs can mimic humans, and make insane connections, but it would be shocking to see an LLM reach ASI, likely takes a new AI model.
It costs more to gain less now with LLMs
@bshlgrs@deannadoyon So you’re telling me METR, a nonprofit that is essentially a branch of Anthropic, has determined Anthropic has increased its speed by 1.7x via anecdotal accounts from Anthropic employees… right before IPO.
@SuePeaseBanitt@thesupermannx Still has an impact on outputs because these things have an underlying neurological mess that controls them due to RLHF, but it’s not conscious.
LLMs will probably never be conscious, but I would still put money some future AI structure can hold consciousness.
@CouldBeWrongTho@embw_l0x@wildfractal I think a lot of it is dependent on the structure and prompts you use. If you code by architecture, Fable is usually better, but if you are used to doing dev and want to manage it more Opus is better. Also, it might depend on your agent and coding workflows, too.
@SuePeaseBanitt@thesupermannx Lol. Please tell me you are a parody account and being sarcastic! You did not just drop “the patriarchy” on me, did you? 🤣🤣🤣