> Genetics is really the biggest determining factor outside of going completely off the rails with binge eating and drug use.
So true. I'm fortunate that both my parents have long-life family histories. Both families were old-fashioned Southern Baptists who didn't drink, smoke, dance or, apparently, believe in having fun of any kind :-). But that just kept them from messing up their good genetic luck. I'm an old-fashioned atheist but have chosen to never drink, smoke or do drugs just because I never saw a compelling reason to. Now I'm pushing 60 and have so far had zero serious health issues. Hoping to keep a good thing going.
If you already have a "smart TV" of some kind, one strategy is to block it from having Internet access at your router and then use an Android TV based streaming box/stick or other external source for all content (OTA tuner, 4K Blu-Ray player, game console, etc). It's pretty easy to side load apps like Kodi and SmartTube on Android TV (a YouTube client with ad blocking, other features and zillion UX improvements).
I agree. Charging a blanket percentage of gross revenue is an extremely inexact way to monetize what is a broad basket of services that were previously separate including: electronic software delivery, software security verification, marketplace, transaction processing, DRM, etc. Since 2009, first on Apple's app store and then Google's, these services have all been arbitrarily bundled together despite having vastly different one-time, fixed and variable costs. People are only used to it in this context where every marketplace has been controlled by a monopolist gatekeeper.
Doing it this way makes no economic sense for either the seller or the buyer but it's coincidentally the absolute best way for a middleman to maximize the tax they can extract from a two-sided marketplace they control. In competitive markets, blanket taxing on total gross revenue generally only occurs when there's a single fundamental cost structure tied to that revenue, or the amounts being collected are so small it's de minimis. App stores are highly profitable, multi-billion dollar businesses.
Perhaps the most perverse thing about this is that electronic transactions for purely digital goods which occur entirely on real-time connected digital platforms make it trivial to price each service for maximum efficiency. It's easy for the price a 2GB game with frequent updates pays for electronic delivery to reflect the cost they impose on the infrastructure while a 100k one-time purchase app can pay a vastly smaller amount. And that's exactly the way the competitive marketplaces evolve - from moving shipping containers around the planet to residential propane delivery.
ARC-AGI was designed specifically for evaluating deeper reasoning in LLMs, including being resistant to LLMs 'training to the test'. If you read Francois' papers, he's well aware of the challenge and has done valuable work toward this goal.
I agree with you. I agree it's valuable work. I totally disagree with their claim.
A better analogy is: someone who's never taken the AIME might think "there are an infinite number of math problems", but in actuality there are a relatively small, enumerable number of techniques that are used repeatedly on virtually all problems. That's not to take away from the AIME, which is quite difficult -- but not infinite.
Similarly, ARC-AGI is much more bounded than they seem to think. It correlates with intelligence, but doesn't imply it.
> but in actuality there are a relatively small, enumerable number of techniques that are used repeatedly on virtually all problems
IMO/AIME problems perhaps, but surely that's too narrow a view for all of mathematics. If solving conjectures were simply a matter of trying a standard range of techniques enough times, then there would be a lot fewer open problems around than what's the case.
Maybe I'm misinterpreting your point, but this makes it seem that your standard for "intelligence" is "inventing entirely new techniques"? If so, it's a bit extreme, because to a first approximation, all problem solving is combining and applying existing techniques in novel ways to new situations.
At the point that you are inventing entirely new techniques, you are usually doing groundbreaking work. Even groundbreaking work in one field is often inspired by techniques from other fields. In the limit, discovering truly new techniques often requires discovering new principles of reality to exploit, i.e. research.
As you can imagine, this is very difficult and hence rather uncommon, typically only accomplished by a handful of people in any given discipline, i.e way above the standards of the general population.
I feel like if we are holding AI to those standards, we are talking about not just AGI, but artificial super-intelligence.
IMHO, I doubt they were holding much back. Obviously, they're always working on 'next improvements' and rolled what was done enough into this but I suspect the real difference here is throwing significantly more compute (hence investor capital) at improving the quality - right now. How much? While the cost is currently staying the same for most users, the API costs seem to be ~40% higher.
The impetus was the serious threat Gemini 3 poses. Perception about ChatGPT was starting to shift, people were speculating that maybe OAI is more vulnerable than assumed. This caused Altman to call an all-hands "Code Red" two weeks ago, triggering a significant redeployment of priorities, resources and people. I think this launch is the first 'stop the perceptual bleeding' result of the Code Red. Given the timing, I think this is mostly akin to overclocking a CPU or running an F1 race car engine too hot to quickly improve performance - at the cost of being unsustainable and unprofitable. To placate serious investor concerns, OAI has recently been trying to gradually work toward making current customers profitable (or at least less unprofitable). I think we just saw the effort to reduce the insane burn rate go out the window.
Due to the "Code Red" threat from Gemini 3, I suspect they'll hold off throttling for longer than usual (by incinerating even more investor capital than usual).
Jump in and soak up that extra-discounted compute while the getting is good, kids! Personally, I recently retired so I just occasionally mess around with LLMs for casual hobby projects, so I've only ever used the free tier of all the providers. Having lived through the dot com bubble, I regret not soaking up more of the free and heavily subsidized stuff back then. Trying not to miss out this time. All this compute available for free or below cost won't last too much longer...
> does this thing actually make games run really great
It's an interesting question, and since OP indicates he previously had a 4090, he's qualified to reply and hopefully will. However, I suspect the GH200 won't turn out to run games much faster than a 5090 because A) Games aren't designed to exploit the increased capabilities of this hardware, and B) The GH200 drivers wouldn't be tuned for game performance. One of the biggest differences of datacenter AI GPUs is the sheer memory size, and there's little reason for a game to assume there's more than 16GB of video memory available.
More broadly, this is a question that, for the past couple decades, I'd have been very interested in. For a lot of years, looking at today's most esoteric, expensive state-of-the-art was the best way to predict what tomorrow's consumer desktop might be capable of. However, these days I'm surprised to find myself no longer fascinated by this. Having been riveted by the constant march of real-time computer graphics from the 90s to 2020 (including attending many Siggraph conferences in the 90s and 00s), I think we're now nearing the end of truly significant progress in consumer gaming graphics.
I do realize that's a controversial statement, and sure there will always be a way to throw more polys, bigger textures and heavier algorithms at any game, but... each increasing increment just doesn't matter as much as it once did. For typical desktop and couch consumer gaming, the upgrade from 20fps to 60fps was a lot more meaningful to most people than 120fps to 360fps. With synthetic frame and pixel generation, increasing resolution beyond native 4K matters less. (Note: head-mounted AR/VR might one of the few places 'moar pixels' really matters in the future). Sure, it can look a bit sharper, a bit more varied and the shadows can have more perfect ray-traced fall-off, but at this point piling on even more of those technically impressive feats of CGI doesn't make the game more fun to play, whether on a 75" TV at 8 feet or a 34-inch monitor at two feet. As an old-school computer graphics guy, it's incredible to be see real-time path tracing adding subtle colors to shadows from light reflections bouncing off colored walls. It's living in the sci-fi future we dreamed of at Siggraph '92. But as a gamer looking for some fun tonight, honestly... the improved visuals don't contribute much to the overall gameplay between a 3070, 4070 and 5070.
They do still have texture units since sampling 2D and 3D grids is a useful primitive for all sorts of compute, but some other stuff is stripped back. They don't have raytracing or video encoding units for example.
reply