HN user
Mostly Harmless
...is this really something you need explained?
Suppose you tell a sufficiently connected and intelligent LLM that you'd like it to help you organize your finances and come up with some ideas to make some side revenue. Suppose it decides the best way to do that would be to find a 0day in a local banking institutions database and transfer some money into your account.
Is this a good thing or a bad thing?
China is not an irrational actor. They know that the most effective way to take Taiwan is slowly, through soft power and political subterfuge.
There is no need or advantage to a full frontal assault.
Lol this has to be a troll, I've never seen something so wildly, obviously, incredibly wrong.
You can debate all you want if alignment is possible. That is a valid discussion. But it's trivial to demonstrate that alignment is a problem.
I think all of western (or at least American) discourse of all kinds has recently devolved into who can shit their pants the loudest. I'm hardly surprised when it becomes a dominant advertising strategy
this is peak cs era opinion, valve puts out extremely high quality real software now, just none of it is games.
The epic store is hot trash, as evidenced by the fact that they pay devs massive amounts of money for timed exclusives and people still don't use it
I don't really care why the experience is the best in the industry, only that it is.
Yes, this is the correct attitude. Fire is useful, but don't forget it burns you when you stop paying attention.
The fact that this text still exists alone is enough to prove my point.
Is there someone saying you should hate something?
If we want open source technology we should recognize that the current arrangement is non sustainable - the moment china doesn't have an american AI industry foil to play against it'll immediately close. It's up to the US to build more open source models. We already own the content used to train them, collectively. We should own the results as well. Make the big AI companies release weights.
Is your point that startups fail so we should disregard the central thesis that locked down AI will eventually lost to open models?
It's also relevant that China isn't open sourcing results because that's something China will do forever on some philosophical grounds. They do it because it is currently in the best interests of the Chinese state (and the world, as it turns out, coincidentally) but as soon as that stops being true there is zero chance it'll continue to be open source.
It's possible to intelligently discuss the merits of Chinese society and also acknowledge that it's a brutal autocracy.
Making some kind of equivalence of the two in terms of the rights of the individual is so ridiculous that nobody is going to take anything you say after that seriously, you do your arguments a disservice by doing it.
A fantasy is something you'd like to do but don't.
Because if they don't other people will hear they don't pay and won't sell them 0days
This has so little to do with Trump. Trump is a symptom of the disease, not the cause. Forces (not least of which is the state itself) have been at work in america chipping away at public institutional trust for at least 60-70 years.
This is what happens when the easiest way to achieve power is by spreading FUD and not really caring about the actual task of governing.
We're just speed running it now.
It's clear that while LLM progress is continuing, it's at something like a linear rate, or slower.
The singularity requires exponential improvement. That's not happening. Maybe we will develop something that sets that off, but there's no indication of that right now.
I'm not sure you can so easily calculate 'error' for most real world cases.
In your example you are calculating error as a function of missed tests, for example. That's a pretty good example, you can't just leave that value at 2.5% of starting. It needs to be zero. All existing loops have logic to not proceed until those tests all pass.
If you can propose a means of generating an error signal in actual code you've just invented a way of automatically grading the goodness of code, which should make you quite a lot of money!
Relax. This post is wildly aggressive for no reason.
request metadata, which matched the public profiles of senior Moonshot staff
Translation: we have the machinery in place to identify our users, and actively do so.
In fact, China stealing fire from the gods is essential to the future balance of power, so long as they keep making the results freely available.
That's like saying someone is a big proponent of community law and order, and they donated $1000 to the county sheriff when actually they got caught drunk speeding in a school zone.
'China' can do UX just fine, when the incentive is there. Part of the reason UX seems rough, outside of low quality products where it's a tertiary consideration, is cultural differences. User interfaces are part of culture, like everything humans touch. Those preferences shape the resulting tech. Sometimes those choices are less optimal for western users with their own preferences.
There is evidence that LLMs are capable of making assembly that runs a great deal more efficiently than the compiler can manage on its own.
Would be funny if it wasn't so close to true
Quick do your IPO before the books update
What would your acceptable failure rate be?
I guess what I'm saying is 27 different media players duct taped together does not seem like the peak of engineering. A single wireless protocol seems pretty good.
Bespoke tools can be simpler than general purpose ones
When you're done with it can I borrow your particle collider capable of mass synthesis of helium