L
@etchedpixels @cstross A lot of the recent legislative moves in the UK & EU both are rather highly incompatible with the use of "unlocked" Open Source systems.
But a bootloader-locked Android could easily enforce this rule.
I'm wondering at which time running a FLOSS Operating System on your devices will require a license/permit.
@em_and_future_cats @mathew @jzb It is not. No LLM can ignore the data in the training set. And NotebookLM is definitely not a local instance.
@etchedpixels @jzb I'm inclined to slightly disagree and think this is denialism (understandably, given how bad their ethics are; it'd be much easier if indeed they had no useful function).
The problem is that, despite all the scenarios where they're inappropriate and wrong, they do.
And we're unwilling (as a society) to fully consider their risks and costs, because "there's no glory in prevention".
That's the challenge we need to overcome.
@jzb On the plus side, Ansible (because it's so freaking widespread and well documented, and it is mostly fairly easy to tell if the answer would do the thing one asked for) is a fairly successful area to apply GenAI to.
Combine with ansible-lint, shellcheck etc in the precommit hook, and the results are actually rather impressive.
@jzb Is is an inherent limitation of how LLMs currently exist and are implemented.
They do strive to minimize it through scale, but it's also a reason why they do get "creative" in their answers.
Like with any stochastic algorithm, they perform best if you can (cheaply) validate the result. e.g., does a program pass the tests still?
This is much harder for complex questions about the real world.
@em_and_future_cats @jzb LLMs are notoriously bad at both using only the information they have, all the information they have, and telling whether they did.
So yes, can be useful, but every answer needs to be validated against facts.
@jzb It's important to note though that you also paint a different possible reality: one in which we don't have to fear automation, but would get to welcome it.
Automation shouldn't be bad.
Wouldn't change the inane energy requirements as implemented currently. But it's not the tech that's necessarily evil: it is the people driving it.
@Purple I've seen a lot of such results of LLM use (I've also seen genuinely constructive use, when properly handled, so I don't want to be seen as a complete naysayer).
So yes, GenAI is definitely a factor.
I'd also raise the point that most have gone through repeated infections with a neuropathologic disease that literally alters the brain.
And we all live in a world of massive external stress that melts our minds.
Sooo LLMs are arriving at the best of all times.