Conversation

Notices

1. Embed this notice
   Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:12:58 JST Cassandra Granade 🏳️‍⚧️

   I have yet, in two years of AI hype, to hear of a concrete use case that AI proponents believe that LLMs will solve, together with any justification for why one might expect that LLMs will solve that use case.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:14:18 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     That's pretty fucking basic, really. If you're advocating for a technology *as a commercial product* and not as a research direction, you absolutely should be able to say that it solves some problem, and that you are justified in that claim.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:16:30 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     For all that I've criticized cryptocurrency and NFTs, they both easily pass that low bar, such that the critique moves into what the costs of that technology are, how it fails in practice, and what indirect consequences might result from its use.

     But AI? We can't even get to that stage! If you're ever lucky enough to see a concrete use case being proposed, there's basically zero argumentation or evidence available to support the claim that AI has anything to offer for that use case.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:18:26 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg Yes and no... even crypto nominally meets some concrete goal. Albeit badly, at extreme environmental cost, in ways that create lots of room for financial scams and that dodge regulatory oversight. But still, you can say something like "I would like to anonymously transfer currency from A to B" and show that it nominally accomplishes that task.

     Don't get me wrong, I hate cryptocurrency... only trying to argue that AI is even worse than that.

   • Embed this notice
     Jer Warren (nyquildotorg@fedia.social)'s status on Thursday, 12-Dec-2024 05:18:27 JST Jer Warren

     in reply to

     @xgranade sounds like crypto. Weird.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:22:02 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     At least for me, personally, that's what makes AI marketing feel so much like gaslighting. There's no there there, but I'm constantly being told that there's something I'm missing. I read descriptions of AI products, and I feel like I'm losing my mind — I cannot parse said descriptions into any concrete understanding of what those products are or what they're expected to do. It's tech hype distilled to its purest fact-free form, and it feels impossible to meaningfully argue with it.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:24:04 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg I broadly agree, but even there, you *could* use blockchains for many of those applications. A plain old database would work much, much better and with far fewer drawbacks, but the blockchain approach would, in some technical way, address the use case.

     That made it much easier, I think, to argue that existing databases work better, because there's at least a concrete task that you can argue about and compare performance on.

   • Embed this notice
     Jer Warren (nyquildotorg@fedia.social)'s status on Thursday, 12-Dec-2024 05:24:05 JST Jer Warren

     in reply to

     @xgranade maybe I should have said "blockchains," because the "currency" aspect of crypto that it facilitates extremely badly is really the only use case. But everyone in the world was pitching using it for all sorts of things where a regular old database would be better.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:26:32 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg Again, none of that is a defense of blockchains, but rather me trying to take lessons from why it was possible to argue against blockchain hype while still feeling sane. Speaking personally, I felt like I could argue specifics and show that the factual claims made by crypto were either incorrect or incomplete.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:29:14 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg Absolutely. Outside of a very few limited cases which had their own problems, I don't think I've ever seen a proposed application of blockchains that wasn't badly replicating existing technologies.

   • Embed this notice
     Jer Warren (nyquildotorg@fedia.social)'s status on Thursday, 12-Dec-2024 05:29:16 JST Jer Warren

     in reply to

     @xgranade yeah, I didn't mean it couldn't be done, just that you'd have to be going out of your way to make it worse in every single way —except you'd get to be propping up the idea of blockchains, which is essentially just crypto evangelism in disguise.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:38:12 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     As an addendum, this is why stuff like "open-source" AI falls so flat for me. I could argue all day long about whether it meets some OSI definition, what it actually means to be open-source, and whatnot. At its core, though, even if you meaningfully made an AI model "open-source," it wouldn't get you any closer to having an actual problem that your AI model solves.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:41:45 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg Absolutely, and a few artists did even make it pretty rich by buying in early to those pyramid schemes... a fact that NFT grifters always loved to point to. Not so much the artists who bought into that grift later and were left holding the bag.

   • Embed this notice
     Jer Warren (nyquildotorg@fedia.social)'s status on Thursday, 12-Dec-2024 05:41:46 JST Jer Warren

     in reply to

     @xgranade one of the bits of the NFT hype cycle I had to admire a little bit, despite how fucked up it was, was the early movement to paint any critique of NFT tech as being against artists getting paid. It was a master stroke of grifting. You'd try to explain to some poor sap how the thing they're involved in isn't doing any of the things they think it is, and you were the villain. Genius.
     
     I live in fear of them figuring out an AI equivalent.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:44:08 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg Wash trading and self-dealing definitely were both huge parts of the scam, yeah.

   • Embed this notice
     Jer Warren (nyquildotorg@fedia.social)'s status on Thursday, 12-Dec-2024 05:44:10 JST Jer Warren

     in reply to

     @xgranade a non-zero number of those that "got rich" were just moving eth from one of their wallets into another one of them. No actual money ever changed hands, yet "sales" were registered.

   • Embed this notice
     Cassandra Granade 🏳️‍⚧️ (xgranade@wandering.shop)'s status on Thursday, 12-Dec-2024 05:45:51 JST Cassandra Granade 🏳️‍⚧️

     in reply to

     • Jer Warren

     @nyquildotorg I don't have any numbers either, so I'll readily admit that I'm not being entirely rational here, but I'd absolutely bet that you're right.

   • Embed this notice
     Jer Warren (nyquildotorg@fedia.social)'s status on Thursday, 12-Dec-2024 05:45:52 JST Jer Warren

     in reply to

     @xgranade I don't have good numbers, but my gut says a very large percentage of the high profile ones were exactly this.