me trying to use normal fucking email without divulging confidential information to copilot
07.10.2025 11:19
👍 4
🔁 0
💬 0
📌 0
me trying to use normal fucking email without divulging confidential information to copilot
Sam says a lack of compute means he must make an impossible choice:
Whether to use his LLM to i) cure cancer or ii) give free education to everyone.
Similarly, I can't achieve BOTH an Olympic gold AND a platinum album.
Don't ask me whether I can achieve either.
www.linkedin.com/posts/axelco...
Even OpenAI's best models still suck at quoting their training data or external sources.
GPT 4o can be useful to help interpret legal provisions pasted into the chat
Ask it to draw upon anything outside of the chat and it breaks down.
The first quote is wrong and the latter two are just made up.
Please nnnno stop
You are giving them far too much credit.
Any bets on whose companies are up next for a major three-letter-agency audit?
The fact that this seems not just plausible but inevitable is both terrifying and kind of funny.
Isn't that the point?
I think that's just the handle of the person/company that owns the rights.
One of several "read later" apps into which I piled hundreds of links that I never "read later".
Google plans to offer Gemini to US children under 13 without verifiable parental consent (parents can opt out if they use Family Link).
As @epicprivacy.bsky.social notes, this seems like a blatant COPPA violation. I don't think Google would have had the gall to do this when Khan headed the FTC.
Globe-straddling yet cuddly. A truly awful look.
Confirmed. My (MANUAL) dash-measurement process reveals that this is an en-dash.
Signal’s new Windows update prevents the system from capturing screenshots of chats
I guess basically all US regulation relates to China now. Or "woke" stuff.
Don't get me wrong—doing business with Chinese companies can present real risks to human rights and security
But for a few years we had tech enforcement (from the DoJ and FTC at least) that wasn't quite so nakedly political
Practically? Just modify the terms and go right ahead. Ethically? I don't know.
"Isn't that chapter about the Library of Congress a little on-the-nose?"
bsky.app/profile/acyn...
This is beyond terrible. A speculative fiction novel with this plot would be so cartoonish and trite as to be unreadable. And I've enjoyed some pretty hackneyed novels.
I have never regretted judging a book by its cover. I found a lot of amazing novels that way. And disregarded a bunch of terrible ones (I assume).
The EU AI Act has a bunch of stuff that is basically about writing and reading the manual. There could be some interesting cases arising out of those bits.
California is particularly bad for particularising its laws into oblivion.
A business can spend more time and money trying to prostrate itself into one of a law's hundreds of exemptions than it would have spent actually complying with the thing.
(In seriousness I'd probably suggest documented, mandatory training for anyone who goes near the thing)
I think we need a definition of "know". And "how". And "to interpret". Another few pages/months won't hurt.
bsky.app/profile/robe...
There was an error in the original, here's a corrected version. I left a stray "facilitate" in (e)(1) which has been updated to "replace".
Damn I think that's my mistake. I might have left some of the old language in there.
This is what happens when lobbyists ask for “clarity” in a regulation.
I hope I'm not being naive but right now I genuinely feel like I'm in a flourishing industry.
Oh you like clarity? Here's 120 pages of it!
"Automated decisionmaking technology does not include... calculators... provided that they do not replace human decisionmaking."
There go a few hilarious calculator-based lawsuits. Just kidding—some firm will try. Back to abacuses just in case, I guess.