RE: https://mstdn.ca/@teledyn/116652708401285794
"every single warning that paper made about large language models has now happened at scale"
1. The hallucination problem before anyone had a word for it.
2. Bias amplification
3. Environmental cost
4. Documentation — the training datasets being assembled were too large for anyone to actually audit
Groggy morning thought: A business that hosts a ton of open standard network APIs for a large number of time/self/thought management data (to-do, event, calendar, note, mindmap, wiki, etc) but they will never ship an app. They are solely a hosting service.
Users pay a monthly fee, cancel at any time, all of their data can be downloaded at any time, no surveillance ads, ... all of the good behavior we should demand from service providers.
Maybe the business even open the service source code and maintain container images for self-hosters and to reduce impact if they go out of business. They might also maintain a number of freely licensed client libraries in a number of coding languages.
The bet is that there are enough people who will pay a subscription to have someone host their info but only if it is 100% standards-based and portable and only if they can use (or make) whatever app they want.
@drahardja Yes! Too late to vote by mail now, so drop your ballot in an actual ballot box.
(I wish we had Democracy Sausage at voting locations. It's an Australian thing.)
This is how the AI bubble bursts: https://www.theverge.com/ai-artificial-intelligence/917380/ai-monetization-anthropic-openai-token-economics-revenue
There is no conceivable way to break even for the AI industry—let alone to repay an investment that requires $2Tn a year from now to the end of the decade. That's about 3% of the entire planetary GNP. Just to break even.