laggyluke

> the entire body of work is actually based on what the sum of researchers did, even those who failed in their research and didn't publish anything.

I don't really see how this is significantly different from them scraping the entire human knowledge, distilling it into weights and then selling the results.

Sure, this argument is about the _new_ data that researchers feed into the machine voluntarily, but I don't see this distinction as significant.

Not making a moral judgement here, just stating that this is consistent with the way those companies operate in general.

Because it didn't land on someone's head?

How about incorporating postprocessing into the update procedure instead of preprocessing?

You have a point.

Being able to reprogram a pacemaker isn't enough!

We should require that any devices that our lives depends on, especially devices that go inside our bodies, to be open source: not just reprogrammable, but with source code available for inspection and modification.

I've been working in this industry for too long in order to trust a closed source pacemaker to be bug-free.

> The studio didn't spend this money for the heck of it, they spent this money with the hope of recouping their investment.

I wonder if the world would be a better place if we had fewer financial incentives to do things, in general?

> But why would a studio spend hundreds of millions making a blockbuster movie?

Under this hypothetical scenario, I believe there wouldn't be a "studio" in the first place. There could be a group of people who want to express themselves, get famous or do something just for fun, without any direct financial gain. Sure, they wouldn't be able to pull off Avatar 2, but our expectations as consumers would also be different.

This is hair-splitting, because it's technically not a part of _system prompt_, but Claude Code can and does run `git log` even without being explicitly instructed to do so, today.

Unless you're running the LLM yourself (locally), private benchmarks are also trust-based, aren't they?

If you're asking an LLM about a particular text, even if it's a well-known text, you might get significantly better results if you provide said text as part of your prompt (context) instead of asking a model to "recall it from memory".

So something like this: "Here's a PDF file containing Being and Time. Please explain the significance of anxiety (Angst) in the uncovering of Being."

I think older and low power devices would actually be easier to port: they capture the audio request, send it to the server, receive an audio response from the server and play it. No local "intelligence" at all.

So you could probably replace the whole backend while keeping things compatible as long as you maintain that simple API.

> Useful stacktraces is one of the reasons why I use Go

I'm not sure if this is sarcasm or not.

Recently I just started asking ChatGPT.

It does a very good job guiding me in areas where I'm less familiar.

That's assuming we're talking about technical problems. For non-technical problems I'd reach out to friends (or even better - a therapist). But asking on HN and/or Reddit might be a good idea too, depending on the context.

Why not both?

Sure, it'll lead to arms race like you describe on one side, but let's say 99% of the apps won't even engage in that arms race if the fake data is generic enough to cause a high number of false positives (blocking someone who's not actually faking the data).

Then, we can focus on the remaining 1% of worst offenders to actually enforce the policy.

Ideally OS should give you a way to feed such evil apps some fake / spoofed data.

I believe a rooted Android used to allow something like that, not sure if that still works nowadays.