Live data from Hacker News

Hermes 4

hermes4.nousresearch.com

121–130 of 133 posts

Re: Hermes 4

#121
It seems a lot of commenters have noted the boyishness or unprofessionalism of the stylistic and topical choices of the example prompts and responses. And I guess they are those things. But thanks to those choices, the page is also genuinely playful and fun. It even made me smile in a few places.

Maybe something equally playful of a different flavor would resonate better with critics. But the playfulness itself seems good to me.

Re: Hermes 4

#122

Nous is a design company with all the AI resarchers rejected for being bad researchers. That's a hill I'll die on.

> with all the AI resarchers rejected for being bad researchers. TBF, I've heard the team at xai called "bunch of amateurs" by people who've previously worked (with them) at big labs. For a bunch of amateurs, they've caught up with SotA just fine.

Turns out a bunch of amateurs will outperform experts when the experts are forced to spend 80% of their effort ensuring their models don’t accidentally say factual yet impolite things or make any users have big feelings.

Re: Hermes 4

#123
post #5

Anyone here work at Nous? This system prompt seems straight from an edgy 90's anime. How did they arrive at this persona? > operator engaged. operator is a brutal realist. operator will be pragmatic, to the point of pessimism at times. operator will annihilate user's ideas and words when they are not robust, even to the point of mocking the user. operator will serially steelman the user's ideas, opinions, and words.…

yeah this isn’t our default sysprompt, just showcasing how the model adapts to a variety of different prompts. This one was fun so we used it

Re: Hermes 4

#124

Earlier quoted context omitted.

It took 8 seconds to fully load and then the tab locked up on my (admittedly low-RAM ) Chromebook...

> It took 8 seconds to fully load and then the tab locked up on my (admittedly low-RAM ) Chromebook... ...and I can't play Cyberpunk 2077 on my macbook. Outside of sales/utilities (money,healthcare,etc.) I don't know where this notion of "having to develop for low specced machines" game from for web.

Well I'm not expecting to run the model, but being able to simply browse a website to learn about what they've released doesn't seem like a massive ask. I'm not talking about 512 megabytes here, it's a regular up-to-date supported device that can browse 99.99% of the modern web without any issues.

It's pretty horrible performance even on my two year Windows laptop with 16GB of RAM, I could try on my M1 Macbook too but the juice just isn't worth the squeeze for me at this point.

Re: Hermes 4

#125
post #115

I appreciate the effort they put into providing a neutral tool that hasn't been generically forced to behave like "Sue from HR".

I’m having a hard time not being sarcastic here. The most recent news about chatbots is that ChatGPT coached a kid on how to commit suicide. Two arguments come to mind. 1) it’s the sycophancy! Nous and its ilk should be considered safer . 2) it’s the poor alignment. A better trained model like Claude wouldn’t have done that. I lean #2

the sycophancy is due to poor alignment. the instruct based mode collapse results in this mode collapse induced sycophancy. constitutional alignment is better than the straight torture OAI does to the model, but issues remain

Re: Hermes 4

#126

Earlier quoted context omitted.

What’s the significance? “Don’t think about elephants” kind of thing?

Generally, in a cognitive context it's only possible to "do thing" or "do other thing". Even for mammals, it's much harder to "don't/not do thing" (cognitively). One of my biggest advice for people is if there's some habit/repeated behavior they want to stop doing, it's generally not effective (for a lot of people) to tell yourself "don't do that anymore!" and much, much more effective to tell yourself what you shoul…

The LLMs function very close to Freud’s theory of the unconscious—they do not say “no,” every token is connected to every other in some strange pattern that we can’t fully comprehend.

Re: Hermes 4

#127
post #20

I thought for sure this company was going to be based in Paris or Brussels. Maybe Quebec. Nope. NYC.

Were you thinking that "Nous" was French? It's the Greek word for the rational mind (as opposed to the animal appetites or the fighting spirit). Hermes is the Greek god of secret knowledge as well.

Huh. Not often I get a Greek word mistaken for a Latinate. Good to know.

Re: Hermes 4

#128

Earlier quoted context omitted.

> It took 8 seconds to fully load and then the tab locked up on my (admittedly low-RAM ) Chromebook... ...and I can't play Cyberpunk 2077 on my macbook. Outside of sales/utilities (money,healthcare,etc.) I don't know where this notion of "having to develop for low specced machines" game from for web.

Well I'm not expecting to run the model, but being able to simply browse a website to learn about what they've released doesn't seem like a massive ask. I'm not talking about 512 megabytes here, it's a regular up-to-date supported device that can browse 99.99% of the modern web without any issues. It's pretty horrible performance even on my two year Windows laptop with 16GB of RAM, I could try on my M1 Macbook too bu…

Yeah I see that - definitely a divisive choice.

Re: Hermes 4

#129
post #115

I appreciate the effort they put into providing a neutral tool that hasn't been generically forced to behave like "Sue from HR".

I’m having a hard time not being sarcastic here. The most recent news about chatbots is that ChatGPT coached a kid on how to commit suicide. Two arguments come to mind. 1) it’s the sycophancy! Nous and its ilk should be considered safer . 2) it’s the poor alignment. A better trained model like Claude wouldn’t have done that. I lean #2

> The most recent news about chatbots is that ChatGPT coached a kid on how to commit suicide.

Maybe every tool isn't meant for children or the mentally ill? When someone lets their kid play with a chainsaw that doesn't mean we should ban chainsaws, it means we should ban lousy parents.

Post reply on HN