"My own belief is that AI training should be treated legally like human learning."
I understand that argument but it makes me think of surveillance. Back before technology it was a lot more cumbersome to actually analyze the data so the surveillance would be more focused. Now there's zero marginal cost to surveil an individual and I wouldn't be surprised if every government/agency/etc is doing as much as they can here.
I don't have an answer or proposal here and don't know which side I'm on but I do think that lack of friction does make it different.
But to your point, we're in a new world and society/laws need to adapt.
"Now there's zero marginal cost to surveil an individual and I wouldn't be surprised if every government/agency/etc is doing as much as they can here."
I agree with you, but I think it's a separate issue. You can't turn the tide back on technology, so the way to protect rights like privacy and other liberties isn't technical, but through laws and checks and balances.
Plenty of terrible regimes did plenty of damage with mass surveillance with paper & pens, so it never stopped anyone. But in this case, if our goal is to stop this, it won't be by stopping AI altogether because the bad regimes around the world won't be stopped by that and we'd just lose all the other benefits (medical cures, scientific progress, increases in productivity, etc)
Yeah, I'd probably be tempted to take deep dives into piles of data if I was in a position to do so - so small wonder that intel and other investigative results are jumping for joy at all of this. Legislation most likely is the best route.
But we can still reject what we can! I've taken my toll transponder out of my car - fortunately the toll roads around here can be avoided - keep my phone & car key fob in anti-RFID bags, refuse to hand over my drivers license at the doctors office (it's not required). I might get really medieval and restart making payments via checks.
I wrote about training data, which is the public web, books, expert data, synthetic data generated by the models, etc.
Are you talking about what people write into chatbots? Like user data? I think that's different, and could certainly be required to be private and follow a bunch of regulations so it's not trained on without permissions or such (and the big AI labs all have policies on this, but I wouldn't mind third party verification).
I've been responding to both your post(s) and Dan Goldin's, and also adding in some indirect reactions to what I've read (and am paranoid about) of surveillance capitalism.
“People will optimize their protein intake, chase early-morning sunlight exposure, fine-tune vitamin D3 levels, and track deep sleep cycles like they’re managing an investment portfolio.”
* if you haven't seen the recent John Oliver on AI slop, it's worth a watch (and lots of laughs). it kinda puts Zuck's AI moves in a new light. love him/hate him but he responds to threats like nobody's business and I hadn't thought of all the AI tools as threats, but duh, they certainly are!
* that Stanley Jordan video...omg, how have I not seen that before? his right hand is better on the fretboard than most people's left hand :D my wife and I saw Matteo Mancuso here in Portland on Sunday and wow, just outstanding. Stanley Jordan though, damn...
"My own belief is that AI training should be treated legally like human learning."
I understand that argument but it makes me think of surveillance. Back before technology it was a lot more cumbersome to actually analyze the data so the surveillance would be more focused. Now there's zero marginal cost to surveil an individual and I wouldn't be surprised if every government/agency/etc is doing as much as they can here.
I don't have an answer or proposal here and don't know which side I'm on but I do think that lack of friction does make it different.
But to your point, we're in a new world and society/laws need to adapt.
"Now there's zero marginal cost to surveil an individual and I wouldn't be surprised if every government/agency/etc is doing as much as they can here."
Agree, 100%
I agree with you, but I think it's a separate issue. You can't turn the tide back on technology, so the way to protect rights like privacy and other liberties isn't technical, but through laws and checks and balances.
Plenty of terrible regimes did plenty of damage with mass surveillance with paper & pens, so it never stopped anyone. But in this case, if our goal is to stop this, it won't be by stopping AI altogether because the bad regimes around the world won't be stopped by that and we'd just lose all the other benefits (medical cures, scientific progress, increases in productivity, etc)
Yeah, I'd probably be tempted to take deep dives into piles of data if I was in a position to do so - so small wonder that intel and other investigative results are jumping for joy at all of this. Legislation most likely is the best route.
But we can still reject what we can! I've taken my toll transponder out of my car - fortunately the toll roads around here can be avoided - keep my phone & car key fob in anti-RFID bags, refuse to hand over my drivers license at the doctors office (it's not required). I might get really medieval and restart making payments via checks.
Wait -- which data are we talking about?
I wrote about training data, which is the public web, books, expert data, synthetic data generated by the models, etc.
Are you talking about what people write into chatbots? Like user data? I think that's different, and could certainly be required to be private and follow a bunch of regulations so it's not trained on without permissions or such (and the big AI labs all have policies on this, but I wouldn't mind third party verification).
Google/Gmail and Meta/Facebook are huge stockpiles of that, and similar protections should apply there.
I've been responding to both your post(s) and Dan Goldin's, and also adding in some indirect reactions to what I've read (and am paranoid about) of surveillance capitalism.
I love your opening post!!
As tech gets more and more embedded in our lives, it seems many aspects of physical embodiment can get ignored and diminished.
Bravo for including this piece!!
Glad you liked it. I feel it's one of these important things that is too rarely talked about 💚 🥃
Not that there's an easy solution for everyone, but being aware of the problem is the first steps towards making changes and improvements!
“People will optimize their protein intake, chase early-morning sunlight exposure, fine-tune vitamin D3 levels, and track deep sleep cycles like they’re managing an investment portfolio.”
I feel attacked
Maybe I was talking about myself 😅
So much to comment on :D
* loved your opening essay...very interesting
* if you haven't seen the recent John Oliver on AI slop, it's worth a watch (and lots of laughs). it kinda puts Zuck's AI moves in a new light. love him/hate him but he responds to threats like nobody's business and I hadn't thought of all the AI tools as threats, but duh, they certainly are!
* that Stanley Jordan video...omg, how have I not seen that before? his right hand is better on the fretboard than most people's left hand :D my wife and I saw Matteo Mancuso here in Portland on Sunday and wow, just outstanding. Stanley Jordan though, damn...
Glad you liked it, thank you for writing a thoughtful comment! 💚 🥃
I'll have to check out Oliver, I haven't seen it. Hopefully that clip is on YouTube.
Yep, the whole segment is at https://youtu.be/TWpg1RmzAbc?si=8FzJMVVrSA2nzpob
"Video unavailable in your country" 🙃
Oh geez, he's gotta be an honorary Canadian. 😂