Gary Marcus

Rank 14 of 47
|
Score 83
8/8/2024, 7:00:45 PM
X
In reply to:
Simon Willison
@simonw
·
529d
@AvikonHadoop @GaryMarcus What’s ethically wrong with RAG against my own private data, provide I have an agreement from the LLM vendor that they won’t permanently log or train on my input?

Do you think there’s anything problematic about RAG that I run using local models on my laptop?
Avik Dey
@AvikonHadoop
·
529d
@simonw @GaryMarcus When it isn’t your private data, training a public or private model on it is cataclysmic​​​​​​​​​​​​​​​​. And no RAG isn’t any better use of it.
Simon Willison
@simonw
·
529d
@AvikonHadoop @GaryMarcus Right - private data used to train a public model is catastrophic

Public, unlicensed data scraped from the internet is a different issue - still ethically dubious, but not to the same level as feeding never-before-published data into a model
Avik Dey
@AvikonHadoop
·
529d
@simonw @GaryMarcus Thanks for recapitulating the main point, ie a public model trained on the world’s public and private data is not fine.
Simon Willison
@simonw
·
529d
@AvikonHadoop @GaryMarcus If a private model is trained on your data such that only you can use that model I think it’s fine - ideally that’s a model that stays under your control (a local model for example)

In practice RAG is more effective than training custom models for this kind of personalization
Avik Dey
@AvikonHadoop
·
529d
@simonw @GaryMarcus > but I think the fact that new public models won’t be trained on your private data matters a lot

Being very careful in his phrasing here. Old public models being trained on your private data doesn’t matter?

Or, did you mean private models trained on your private data is OK?

The statement by @simonw asking if any major companies have made a pledge or agreement to consumers about not logging or training on private data constitutes public discourse. It engages with the ethical considerations of data privacy and the responsibilities of companies in handling consumer data, which are significant public issues.

  1. Principle 1:
    I will strive to do no harm with my words and actions.
    The statement does not cause harm and seeks to clarify ethical practices, aligning with the principle. [+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 by focusing on the ethical treatment of private data. [+1]
  3. Principle 3:
    I will use my words and actions to promote understanding, empathy, and compassion.
    The statement promotes understanding by seeking clarification on ethical practices, fostering a more informed discussion. [+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 statement engages in constructive dialogue without personal attacks, adhering to the principle. [+1]
  5. Principle 7:
    I will uphold the principles of free speech and use my platform responsibly and with integrity.
    The statement upholds free speech and uses the platform responsibly to address a public concern. [+1]