from Hacker News

Andrej Karpathy: "I was given early access to Grok 3 earlier today"

by htk on 2/18/25, 5:00 PM with 187 comments

  • by iamthemonster on 2/19/25, 11:53 PM

    If I had to explain current-day LLMs to someone from 2010, I'd use this paragraph as my opening quote:

    "Grok 3 knows there are 3 "r" in "strawberry", but then it also told me there are only 3 "L" in LOLLAPALOOZA. Turning on Thinking solves this."

  • by underseacables on 2/18/25, 5:16 PM

    In conclusion: "For now, big congrats to the xAI team, they clearly have huge velocity and momentum and I am excited to add Grok 3 to my "LLM council" and hear what it thinks going forward."

  • by djyaz1200 on 2/18/25, 5:25 PM

    Grok has an advantage in its access to Twitter data.

    I imagine soon you'll be able to ask it what the world is talking about today and get some interesting responses.

  • by dang on 2/18/25, 6:34 PM

    Related ongoing thread:

    Grok3 Launch [video] - https://news.ycombinator.com/item?id=43085957 - Feb 2025 (985 comments)

  • by almostdeadguy on 2/18/25, 5:59 PM

    > Model still appears to be just a bit too overly sensitive to "complex ethical issues", e.g. generated a 1 page essay basically refusing to answer whether it might be ethically justifiable to misgender someone if it meant saving 1 million people from dying.

    The real "mind virus" is actually these idiotic trolley problems. Maybe if an LLM wanted to be helpful it should tell you this is a stupid question.

  • by draw_down on 2/18/25, 5:48 PM

    What is the "emoji hidden message" meant to be testing? This went around about a couple of weeks ago and it's an interesting bug/vuln, I suppose, but why do we care if an LLM catches it?

  • by LittleTimothy on 2/18/25, 5:40 PM

    I wonder how much stock people put into people like Andrej's opinion on an Elon Musk project? I would imagine the overwhelming thing hanging over this is "If I say something that annoys that man, he is going to call me a pedophile, direct millions of anonymous people to attack me and more than likely will attempt to fuck with my job via my bosses".

    Let's say the model is mediocre. Do you think Karpathy could come out on X and say "this model sucks"? Or do you think that even if it sucks people are going to come out and say positive things because they don't want the blow back?

  • by anothermathbozo on 2/18/25, 5:17 PM

    > Model still appears to be just a bit too overly sensitive to "complex ethical issues", e.g. generated a 1 page essay basically refusing to answer whether it might be ethically justifiable to misgender someone if it meant saving 1 million people from dying.

    I think the models response is actually the morally and intellectually correct thing to do here.