They're referring to the training data being garbage, not the diffusion process.
1,682 karma · joined March 21, 2016
Great men I admire: - Jimi Hendrix - Claude Shannon - Carl Sagan
Footguns, learning curve? Sure, but I'll still make that trade.
RAG - Retrieval augmented generation - how can the retrieval be done during training? RAG will always remain external to the model. The whole point is that you can augment the model by injecting relevant context into the prompt at inference time, bringing your own proprietary/domain-specific data.
> You don't need half an acre to sleep.
Where do I put my vegetable garden and fruit trees?
How can I relax on my back porch listening to the birds and creek?
How do I get away from all the hustle and bustle of dense city living?
> plenty of stuff available at walking distance. And you can't have that without high population density and yeah "pods".
Exhaust, brake dust, sirens, litter, concrete jungle, noisy neighbors with thin walls, massive crowds/traffic... there are tradeoffs to living in a dense city.
I think the Pareto optimal living conditions in the US today are the suburbs bordering the rural outskirts. Especially now with WFH prevalence.
> That’s why one of the possible threats of AI is if one would “forget” how to reason. Or even worse: brain rewires itself to never being able to reason again.
The "WALL-E/Idiocracy"-ification of humanity. It won't get so bad that we lose our ability to reason, but we're definitely losing something. Perhaps it's just our inevitable evolution.
Never heard of the program but on its face that sounds pretty bad. Grift, scam, or just inefficient govt? Not sure but not a good argument for keeping it around!