Gary Marcus

Rank 13 of 47
|
Score 95
Gary Marcus
@GaryMarcus
·
422d
So many people are confused about the relation between human cognitive errors and LLM hallucinations that I wrote this short explainer:

Humans say things that aren't true for many different reasons
• Sometimes they lie
• Sometimes they misremember things
• Sometimes they fail…
4/21/2024, 5:21:59 PM
X
In reply to:
Gary Marcus
@GaryMarcus
·
422d
@outworlder yes that is basically correct (as i have noted here before).
Felon Muskovsky
@outworlder
·
422d
@GaryMarcus Counterpoint: all outputs are "hallucinations". It's just so happens that some outputs align with reality purely on a probabilistic basis.
Gary Marcus
@GaryMarcus
·
423d
If all we had was ChatGPT, we could say, hmm “maybe hallucinations are just a bug”, and fantasize that they weren’t hard to fix.

If all we had was Gemini, we could say, hmm “maybe hallucinations are just a bug”.

If all we had was Mistral, we could say, hmm “maybe hallucinations…

The statement is part of a public discourse on the nature of outputs from language models like ChatGPT, Gemini, and Mistral, and their comparison to human cognitive errors. The conversation involves a discussion on whether the inaccuracies or 'hallucinations' of these models are inherent flaws or simply probabilistic occurrences. This discussion contributes to the broader dialogue about artificial intelligence, its limitations, and its comparison to human cognition.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not appear to cause harm and is focused on a technical discussion, thus aligning with the principle of doing no harm with words. [+1]
  2. Principle 2:
    I will respect the privacy and dignity of others and will not engage in cyberbullying, harassment, or hate speech.
    The statement respects privacy and dignity, focusing solely on the topic without involving personal data or attacks. [+1]
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement contributes to a better understanding of AI technologies, promoting a nuanced view of 'hallucinations' in language models. [+1]
  4. 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 dialogue is an example of constructive criticism and engagement with differing views without resorting to personal attacks. [+1]
  5. Principle 6:
    I will use my influence for the betterment of society.
    The discussion uses influence to educate and inform about AI, contributing positively to societal understanding of technology. [+1]
  6. Principle 7:
    I will uphold the principles of free speech and use my platform responsibly and with integrity.
    The statement uses the platform responsibly, engaging in a substantive discussion that respects free speech and integrity. [+1]