8 ms·

More likely Musk liked xAI's CTO more than X's CTO, and parted ways with one of them. Which is his prerogative, as the owner of both companies. I'm sure Haofei

by ein0p 2y ago

More likely Musk liked xAI's CTO more than X's CTO, and parted ways with one of them. Which is his prerogative, as the owner of both companies. I'm sure Haofei would love to be the CTO of one of the strongest AI models in existence. I would, if I were in his shoes.

CursedSilicon 2y ago
"strongest AI models in existence" [citation needed]
- llm_nerd 2y ago
  To be fair they said one of the strongest models. We're at the point now where any heavily monetized player can build one of the strongest models. I mean, especially with all of the incredibly open Chinese companies releasing cutting edge research now that most American firms have gone quiet or only releasing their tailings.
  - foobarian 2y ago
    Technically my toy Markov chain I trained on Gutenberg books is one of the strongest models.
    - ikiris 2y ago
      If you used real chain it definitely has the most strength.
- ein0p 2y ago
  See the leaderboard tab here: https://lmarena.ai https://lmarena.ai. Or just try it. It's excellent. Usage is free of charge so far. Subjectively, on my prompts, it seems substantially stronger than GPT 4o and 4.5. I also like its more "informal" tone, and its willingness to discuss the more controversial topics.
  - uxcolumbo 2y ago
    I think it’s important to not support products by people who want to destroy democracy and who openly sieg heiling or claiming Hitler wasn’t responsible for the Holocaust.
    - gwern 2y ago
      On the bright side, Google's new Gemini-2.5-pro comes with no Holocaust-denying issues, and is definitely a lot better than Grok, and is also free and fast. (I've been using it the past 2 days and am pretty impressed - the first high-end Google model I'd willingly use instead of GPT-4.5 or Claude-3.7.)
      - ein0p 2y ago
        Meh. Google can keep its censorship and algorithmic narrative steering for itself, IMO. Quis custodiet ipsos custodes, or something like that.
        gwern 2y ago
        The Gemini models have become notorious for being rather uncensored and unrestricted, with an oddly base-model-like feel, since at least last November.
        ein0p 2y ago
        Good to know, I will give them another try, thanks. Cooler heads might have prevailed over there.
    - arandomusername 2y ago
      [flagged]
      - fashion-at-cost 2y ago
        [flagged]

by ein0p 2y ago

CursedSilicon 2y ago

"strongest AI models in existence" [citation needed]

llm_nerd 2y ago
To be fair they said one of the strongest models. We're at the point now where any heavily monetized player can build one of the strongest models. I mean, especially with all of the incredibly open Chinese companies releasing cutting edge research now that most American firms have gone quiet or only releasing their tailings.
- foobarian 2y ago
  Technically my toy Markov chain I trained on Gutenberg books is one of the strongest models.
  - ikiris 2y ago
    If you used real chain it definitely has the most strength.
ein0p 2y ago
See the leaderboard tab here: https://lmarena.ai https://lmarena.ai. Or just try it. It's excellent. Usage is free of charge so far. Subjectively, on my prompts, it seems substantially stronger than GPT 4o and 4.5. I also like its more "informal" tone, and its willingness to discuss the more controversial topics.
- uxcolumbo 2y ago
  I think it’s important to not support products by people who want to destroy democracy and who openly sieg heiling or claiming Hitler wasn’t responsible for the Holocaust.
  - gwern 2y ago
    On the bright side, Google's new Gemini-2.5-pro comes with no Holocaust-denying issues, and is definitely a lot better than Grok, and is also free and fast. (I've been using it the past 2 days and am pretty impressed - the first high-end Google model I'd willingly use instead of GPT-4.5 or Claude-3.7.)
    - ein0p 2y ago
      Meh. Google can keep its censorship and algorithmic narrative steering for itself, IMO. Quis custodiet ipsos custodes, or something like that.
      - gwern 2y ago
        The Gemini models have become notorious for being rather uncensored and unrestricted, with an oddly base-model-like feel, since at least last November.
        ein0p 2y ago
        Good to know, I will give them another try, thanks. Cooler heads might have prevailed over there.
  - arandomusername 2y ago
    [flagged]
    - fashion-at-cost 2y ago
      [flagged]