Glad to hear you’re not giving up on Tumblr after all! 🙏 (Also I'm sure you didn't mean it literally but still I feel obliged to protest that my post was NOT intended as a callout of Garbage Day, which is a wonderful publication and I love it like a son.)
The big deception around which everything else flows is that these models are autonomous in any way that actually matters, as opposed to simply automated colaters of billions of examples of lossily compressed, corrupted-to varying-degrees existing material. The blitzkrieg about 'is it REALLY sentient?', which was demolished almost instantly but still managed to sow the seeds it needed to, was about reframing it to their advantage - if you asked questions like 'should a word processor come with billions of unlicensed books/forum discussions/shitposts preloaded for its users to incorporate at semi-random?' or should 'Photoshop do the same with billions of existing images?' the answers would seem hilariously obvious.
It's not really obvious to me what the non-mischievous/destructive use case is for image/video generating AIs to accept prompts with the names of real people. Like, what is that for, other than creating fake images? I feel like a lot of potential issues would be solved if they just...didn't do that.
Important thing to keep in mind about Meta: they *did* make React, which is responsible for the “component web,” arguably the most important thing to happen to the open internet since AJAX. So I guess there’s a chance that it’s wasn’t a fluke and there are people trying to do more stuff like that.
1. Definitely a good take, I have heard email called a “federated” system over 10 years ago.
2. Follow-up: Google has also done this with calendar-ing. If you don’t have Google Calendar (or Outlook I guess?), you have to jump through so many hoops to do things like add an event
3. I also remember when GChat/GoogleTalk federated through the XMPP until one day they just shut it off. Only the nerds cared, and nobody seemed to mind when they couldn’t hear from them anymore.
The line about reporters covering the previous election reminds me of Marshall McLuhan's concept of the rear-view mirror - the way I understand it, he thought we could only look at brand new media or technology in terms of previous, similar concepts that are already established. Hence "horseless carriage" and that kind of thing. When TV first became a thing, you could see people treating it like radio that also sends pictures, and early shows were attempts at figuring out what could be done with the medium. You could see the same thing happen with the internet, podcasts, youtubers, microblogging, etc., and I think AI is pretty similar in this way - all the old problems are still there, and now they're taking fancy new forms that have their own complications, and plus, it's new enough that everyone is still screwing around with it trying to figure out what it's good for.
Internet 1.0 was the first big grift. But no one calls it that. Then crypto was the second big grift. Now AI is the third. And it's always funded by the same set of VCs; libertarian technocrats preaching the power of the individual over society. I attended My Effin' Life at The Beacon Theatre in NYC earlier this week, where Geddy Lee responded to the question about Sapolsky's book about the lack of free will by saying that he would walk back the philosophy of Rush's earlier song lyrics, inspired by Ayn Rand, and that we need to practice social responsibility. None of these VCs give a shit about social responsibility. They all adhere to the very binary thinking and find cover under Kahneman's writings. The latter's nobel winning philosophy has failed us as there are clearly more than 2 systems that drive the individual's actions, namely emotion and morality. We are all actors in networks and the key thing is to find balance between the two. It's never easy. But well structured networks can be both generative (coming from the individual) and sustainable (coming from the institutions).
Imagine writing a magazine about social media and complaining every issue about how Twitter isn't Twitter anymore, while in the same breath having a hard time comprehending the need for the fediverse to exist. Like, holy shit, Ryan, we're watching you become less self aware while you watch machines attempt to appear more self aware in real time at this point.
Everyone stop being ridiculous for like five minutes
Glad to hear you’re not giving up on Tumblr after all! 🙏 (Also I'm sure you didn't mean it literally but still I feel obliged to protest that my post was NOT intended as a callout of Garbage Day, which is a wonderful publication and I love it like a son.)
The big deception around which everything else flows is that these models are autonomous in any way that actually matters, as opposed to simply automated colaters of billions of examples of lossily compressed, corrupted-to varying-degrees existing material. The blitzkrieg about 'is it REALLY sentient?', which was demolished almost instantly but still managed to sow the seeds it needed to, was about reframing it to their advantage - if you asked questions like 'should a word processor come with billions of unlicensed books/forum discussions/shitposts preloaded for its users to incorporate at semi-random?' or should 'Photoshop do the same with billions of existing images?' the answers would seem hilariously obvious.
It's not really obvious to me what the non-mischievous/destructive use case is for image/video generating AIs to accept prompts with the names of real people. Like, what is that for, other than creating fake images? I feel like a lot of potential issues would be solved if they just...didn't do that.
Important thing to keep in mind about Meta: they *did* make React, which is responsible for the “component web,” arguably the most important thing to happen to the open internet since AJAX. So I guess there’s a chance that it’s wasn’t a fluke and there are people trying to do more stuff like that.
Re: Meta becoming the Google of ActivityPub:
1. Definitely a good take, I have heard email called a “federated” system over 10 years ago.
2. Follow-up: Google has also done this with calendar-ing. If you don’t have Google Calendar (or Outlook I guess?), you have to jump through so many hoops to do things like add an event
3. I also remember when GChat/GoogleTalk federated through the XMPP until one day they just shut it off. Only the nerds cared, and nobody seemed to mind when they couldn’t hear from them anymore.
The line about reporters covering the previous election reminds me of Marshall McLuhan's concept of the rear-view mirror - the way I understand it, he thought we could only look at brand new media or technology in terms of previous, similar concepts that are already established. Hence "horseless carriage" and that kind of thing. When TV first became a thing, you could see people treating it like radio that also sends pictures, and early shows were attempts at figuring out what could be done with the medium. You could see the same thing happen with the internet, podcasts, youtubers, microblogging, etc., and I think AI is pretty similar in this way - all the old problems are still there, and now they're taking fancy new forms that have their own complications, and plus, it's new enough that everyone is still screwing around with it trying to figure out what it's good for.
See, I know what all those words mean because I teach middle school...it is indeed a Big Trend. The boys won’t stop singing it. Constantly.
Internet 1.0 was the first big grift. But no one calls it that. Then crypto was the second big grift. Now AI is the third. And it's always funded by the same set of VCs; libertarian technocrats preaching the power of the individual over society. I attended My Effin' Life at The Beacon Theatre in NYC earlier this week, where Geddy Lee responded to the question about Sapolsky's book about the lack of free will by saying that he would walk back the philosophy of Rush's earlier song lyrics, inspired by Ayn Rand, and that we need to practice social responsibility. None of these VCs give a shit about social responsibility. They all adhere to the very binary thinking and find cover under Kahneman's writings. The latter's nobel winning philosophy has failed us as there are clearly more than 2 systems that drive the individual's actions, namely emotion and morality. We are all actors in networks and the key thing is to find balance between the two. It's never easy. But well structured networks can be both generative (coming from the individual) and sustainable (coming from the institutions).
> I mean, my god, OpenAI is literally using the same Africa-based third-party moderation contractors that Meta and Google use.
At least the Stasi got the same housing and healthcare benefits as everyone else.
Imagine writing a magazine about social media and complaining every issue about how Twitter isn't Twitter anymore, while in the same breath having a hard time comprehending the need for the fediverse to exist. Like, holy shit, Ryan, we're watching you become less self aware while you watch machines attempt to appear more self aware in real time at this point.
Thanks for offering Garbage Day Free of Charge! 👍