Thank you very much. My concern is rather in the direction of inserting ads or “promotional information” into the training material, much like SEO plagues search today. If the info is from the web it can still be malicious, even if you run your own LLM.
which previously failed since ads and SoC were the driver of the Web, not information.
Can you elaborate on why you think the ads wouldn’t sneak in again? The semantic web is a fantastic concept, but I don’t immediately see the AI connection. AI doesn’t magically pay for authored content and there is still an incentive to somehow get ads into LLM answers.
So does the stellarator. What’s the argument here?
Nah, not impossible people build stellarator type Fusion reactors with large freeform metal parts in that tolerance region that are exposed to liquid helium.
I suspect that most people don’t subscribe to Spotify to listen to white noise but other music. So they might not lose a lot of revenue because white noise is not their core value proposition.
Only of those people subscribed to Spotify to listen to white noise. I suspect it’s a side effect…
Yes, but a binary gate reacts to a change in inputs exactly once by adjusting its own state. If the inputs change faster the frequency will change of course, but that’s not the point. Neurons will fire pulse trains with different rates for two different inputs that a binary system would both interpret as “on”. It’s a much more analog and continuous system in that regard.
That’s just plainly wrong. If neurons are “activated” (the binary analogy) it starts firing, but at varying rates depending on how far above it’s threshold the activation happened. A bit like an activation level to frequency converter, but non-linear.
They fire at different rates are though.
Can I donate to EFF instead?
Good job on #10. I love it.
It’s a last-mile thing. Artificially boosts the download numbers which most customers look at.