Gary Marcus

Rank 14 of 47
|
Score 83
8/8/2024, 3:33:48 AM
X
In reply to:
Simon Willison
@simonw
·
528d
@GaryMarcus A lot of people misunderstand the difference between training models and feeding prompts into them - I think it’s important for us to help people understand that, it’s fundamental for understanding what these things can and cannot actually do https://simonwillison.net/2024/M…
Simon Willison
@simonw
·
528d
@GaryMarcus The difference her is critically important: training a shared model on private documents is a five alarm fire, ethically - an enormous betrayal of trust

Building an LLM-based personal assistant that selectively pulls private data into the context (RAG) is responsible engineering
Simon Willison
@simonw
·
528d
@GaryMarcus I think you may have misinterpreted that - it’s from this podcast interview, but the segment you quote here doesn’t look to me like it’s about training - it sounds more like a description of RAG or tool use

https://assets.gatesnotes.com/8a…
Gary Marcus
@GaryMarcus
·
528d
@simonw Come on, people, I always have receipts. This is mentioned in my upcoming book, source is an interview w Bill Gates
Simon Willison
@simonw
·
528d
@GaryMarcus “Sam Altman has acknowledged wanting to train on everyone's personal documents”

I missed that, when and where did he say that?
Gary Marcus
@GaryMarcus
·
528d
⚠️ Something terrifying just snapped in place:
• Sam Altman has acknowledged wanting to train on everyone's personal documents (Word files, email etc)
• ChatGPT has gathered unprecedented amounts of personal data
• Sam founded WorldCoin, known for their eye-scanning orb
•…

The statement 'I do not misunderstand that difference, sir.' is a direct response to a previous comment in a conversation about the differences between training models and feeding prompts into them. The tone is assertive and aims to clarify the respondent's understanding of the topic. The conversation involves a substantive engagement with public issues related to the ethical implications of training language models on personal data, which qualifies it as public discourse.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and is a clarification of the respondent's understanding, aligning with the principle of striving to do no harm. [+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 the privacy and dignity of others, as it does not engage in cyberbullying, harassment, or hate speech. [+1]
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement does not explicitly promote understanding, empathy, or compassion, but it does contribute to a constructive dialogue by clarifying a point of misunderstanding.
  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 statement engages in constructive criticism and dialogue without resorting to personal attacks or ad hominem arguments. [+1]