Gary Marcus

Rank 35 of 47
|
Score -42
3/24/2025, 3:24:10 PM
X
In reply to:
Andrew Allen
@andrewallenxo
·
11d
@GaryMarcus so are you saying there's no point in trying to address these as 'bugs to be fixed' in LLMs?
Gary Marcus
@GaryMarcus
·
11d
@andrewallenxo there will always be errors, but better architectures might not confabulate whole cloth in the way that LLMs do.
Andrew Allen
@andrewallenxo
·
11d
@GaryMarcus I don't think it'll ever be fully solved. But reduced hallucinations are probably possible?
Gary Marcus
@GaryMarcus
·
11d
If only I had a nickel for every time someone told me wrongly that hallucinations had been solved…
James Pogue
@jhensonpogue
·
11d
It seems like it ought to be a bigger deal that ChatGPT routinely invents fake sources and even fabricates ISBN numbers to make them look real

The statement suggests a shift from relying on uncertain solutions to investing in research for alternative approaches, addressing the issue of errors in language models. It promotes a proactive and research-oriented mindset.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement aligns with the principle of doing no harm by advocating for research to address issues rather than relying on uncertain solutions. [+1]
  2. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    It promotes understanding and constructive action by suggesting investment in research, which can lead to better solutions. [+1]
  3. Principle 4:
    I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
    The statement engages in constructive dialogue by proposing a practical approach to a problem rather than dismissing it. [+1]