Gonna delete the thread (the point about data contamination is real but I don’t want to let Adobe off the hook) Thanks for your reporting!
REGULATION
-
Adobe’s Ethical Firefly AI Trained on Midjourney Images
By
–
🚨Scoop from me and @BrodyFord_: Adobe has touted Firefly as a safe and ethical alternative to competitors like Midjourney… but it also quietly trained Firefly on Midjourney images. Adobe’s ‘Ethical’ Firefly AI Was Trained on Midjourney Images bloomberg.com/news/articles/…
→ View original post on X — @strongreporter, 2024-04-12 14:08 UTC
-
NYT lawsuit fair use argument against AI training
By
–
The first four pages of the NYT lawsuit are worth reading – the legal argument rests on "fair use", and one of the factors in whether or not something is fair use is how much it competes with the content it has duplicated https://
nytco-assets.nytimes.com/2023/12/NYT_Co
mplaint_Dec2023.pdf
… -
Email Policy Changes: User Communication Challenges
By
–
Similarly I don't think "emailing our users to inform them about this policy change" is reasonable – I have accounts with well over 1,000 different services (according to my 1Password), the chance of me reading a policy email from any of them is pretty small!
-
Account Access Loss and Digital Identity Management Challenges
By
–
That assumes you can still login to your account – at this point I have plenty of accounts on services that I can no longer access – because they're associated with an old email address, for example
-
AI Interpretability Circuit Discovery Disclosure Concerns Analyzed
By
–
the reason why sholto and trenton avoid answering dwarkesh's question about the highest level circuit discovered by the interpretability team but casually mention circuits for deception and love must be the industry's reluctance to disclose the consciousness circuit
-
Crafting Crisis Apologies for Controversial AI Policies
By
–
An interesting exercise in imaginary crisis management is trying to compose an apology that can help regain the trust lost by the fact that this policy made it to the point of being announced in the first place
-
Platform Trade-offs: Removing User Anonymity for Review Accountability
By
–
Presumably the rational here is that they have millions of existing reviews which are a huge asset to their platform, and they don't want to remove them, but they also don't want them to stay anonymous And maybe they've decided that angering a subset of their users is worth it?
-
OpenTable privacy policy change displays user data without consent
By
–
I'm finding this privacy breaching decision by @OpenTable baffling: "To build on the credibility of our review program, starting May 22, 2024, OpenTable will begin displaying diner first names and profile photos on all diner reviews. This update will also apply to past reviews"
-
AI Safety: Regulating Model Naming Practices
By
–
AI safety demands the creation of a law that AI researchers should not be allowed to name their own models anymore. The only thing worse than a hostile AGI would be a hostile AGI called Phat Goose.