> You'd open your phone and nothing would work, not even the connection to the cellular network since your phone company has management components running on AWS.
I'm pretty confident, sitting here on a different continent, that I wouldn't have that problem at all.
I think parent is trying to say that the number of services that rely themselves on AWS or some of their internal services is enormous and potentially impacts many.
> Intel GPUs are pretty worthless right now for inference, let alone training.
Why? Is it hardware limitations or deficient software stack?
In any case, as nvidia is ignoring the gaming market, AMD and Intel both have a vacuum to fill, but I don't see them doing it.
Intel, especially, should have rushed out gaming cards as soon as nvidia shortages were predictable (2023, at least).
The problem I have is not that an old GPU can't be used for gaming (it can), it's that I am forced to upgrade from a perfectly working 660ti that once played far cry 3, 4 and 5 (and every other game I threw at it) because my kernel doesn't have support anymore (older drivers don't work on a newer kernel).
A new GPU with the power of a decade-old GPU but all the latest software support would be great. Right now, if I want to see certain web-pages (WebGL), I have to buy a new GPU, but there isn't any low-priced GPUs anymore.
> A new GPU with the power of a decade-old GPU [...]
Asus tried, with GDDR5 versions of the GT710 circa 2021 in a very sensible fanless, "bus-powered", single slot design; alas being an older design did not help much with stability with Nouveau drivers, though on the Windows side it did 10 as well as XP which wasn't a bad deal for spending a week digitizing a bunch of tapes with a good old BT868 capture card!
Bonus point for proper VGA and DVI connectors too, just one weird omission - a half-height card only supplied with a full height bracket?
Memory bandwidth is really low on most of the cards - even if you hyper-optimized the software stack, can't get around that. Compute isn't great either.
I think the cards are probably fine for gaming though. I haven't used these cards, just the numbers I've seen floating around are pretty underwhelming.
Define low price, because looking at sold listings on ebay, you can get a used RTX 3070 for around $250, which is as capable as any 10-year old 8 GB card, and certainly more capable than a 660 Ti.
Still not great given that I paid about the same for a new-in-unopened-box RTX 3070 a couple years ago, but price increases no lower-end cards without enough VRAM for AI at least haven't skyrocketed like the top end of the market.
> There is probably a power advantage of being able to convert n filetypes to m different formats where n and m are very large.
If they've been in business this long, they must have, by now, internal filetypes for specific classes of formats. For example, all rgb files first convert to an RGB format, and then to the destination format. Or, all transcription first convert to an internal format, and then to the destination format.
Not sure how you'd handle .ps or .pdf files in this way, though.
> I would expect code written by the latest frontier models with a "find all security issues make no mistakes" pass to be more secure than average human code written in the pre-AI era (not a high bar).
"If you write code as cleverly as you can, you are, by definition, too stupid to debug it."
Do you disagree with this? Because if you don't, then the rate of bugs/RCEs in a completely AI-written-software world will be exactly the same as the rate of bugs in a completely human-written-software world.
A combination of human-authorship + AI-review, with absolutely no AI-written code is going to have the lowest rate of bugs, because there is always one entity who understands the code (because they wrote it) and one entity who is smarter than the author who can find bugs.
The AI-generated code I see from SOTA models is absolute spaghetti, and they start running into problems spotting bugs after even a small number of refinements/changes to the software (say, 50).
The problem with all this spaghetti is that they are quick to write something that even they can't easily read. Even mediocre but experienced humans are not so eager to write something that they can't easily debug.
IOW, for any non-trivial problem, with AI-generated code you are guaranteed to get spaghetti that it cannot understand. With humans, it's only a probability that that will happen.
Well the AI machines don't have to rely on cleverness to find vulnerabilities, so maybe they're not limited by Kernighan's Law. Does throwing virtually unlimited sufficiently clever monkeys at spaghetti code find these vulnerabilities better than humans? Can you prove they can't?
At the point you starts asking to prove a negative, like "Can you prove that there's no God?", you should have the self-awareness to realise that no argument will convince you, because you are proceeding on the power of belief and belief alone.
If you don't have the self-awareness to realise that you are asking the same question as "Can you prove there's no God?", then I'm afraid no proof will ever be sufficient for you.
We're not talking about God, we're talking about a machine, here on Earth, we can poke and prod and explore and test. It's perfectly within the realm of logic to prove it can't do things. Can you prove the AI machine cannot solve the halting problem for arbitrary programs? Yes, that's a negative you can prove, so your objection about proving negatives needs qualification.
Asking you to justify a limitation you asserted doesn't mean I cannot be convinced. You're asserting Kernighan's law holds for AI, I'm saying there are other ways to debug than applications of cleverness. Does throwing more attempts at that spaghetti find vulnerabilities even when the thing that wrote it can't easily understand it? I'm not saying AI must be able to do it because you can't prove it can't. I'm asking what supports your claim that it can't.
> No no no, don’t burst that AI bubble just yet. I haven’t finished all my side-projects.
You'll be fine; the GPUs will be sold off for pennies on the dollar, and if you can't run one at home there are plenty of people who will buy them to host for cost of hosting +5% markup.
> The stock exchange is the Adam Smith way of directing the assignation of resources to the most efficient company, and it's supposed to be liquid enough to allow fine-grained decisions.
So? OAI isn't on any stock exchange right now. Their value is not determined purely by supply and demand.
> Why do I sense a schadenfreude here? It's like people are rooting for their failure.
You sound surprised. Why?
After all, they have been selling their product on the basis of driving the value of human labour as close to zero as possible (your labour included). Why wouldn't people want such an endeavour to fail?
I'm pretty confident, sitting here on a different continent, that I wouldn't have that problem at all.
reply