Crypto gets a lot of hate... but this really puts its utility into perspective: No counterparty risk with random banks or foreign companies, near-instant settlement, vastly lower fees, immediate fx conversion.
There will undoubtedly be people that like or love it and there's nothing wrong with that. Design is rather subjective. Fortunately I'm not in the market for a $300,000+ EV made by Ferrari, so I don't have to lose sleep at night over buying this or not :)
Office workers will eat lunch, take a 1-2hr nap in the afternoon, and also eat dinner with their coworkers within the common 9-9-6 rhythm. It still takes a significant chunk of time, but the actual working time butt-in-chair is closer to 54 hours
Application-specific AI models can be much smaller and faster than the general purpose, do-everything LLM models. This allows them to run locally.
They can also be made to be deterministic. Some extra care is required to avoid computation paths that lead to numerical differences on different machines, but this can be accomplished reliably with small models that use integer math and use kernels that follow a specific order of operations. You get a lot more freedom to do these things on the small, application-specific models than you do when you're trying to run a big LLM across different GPU implementations in floating point.
Yeah, in the same way how pseudo-random number generators are "deterministic." They generate the exact same sequence of numbers every time given the seeds are the same!
But that's not the "determinism" people are referring to when they say LLMs aren't deterministic.
Some people care more about compile times than the performance of generated code. Perhaps even the correctness of generated code. Perhaps more so than determinism of the generated code. Different people in different contexts can have different priorities. Trying to make everyone happy can sometimes lead to making no one happy. Thus dichotomies like `-O2` vs `-Os`.
EDIT (since HN is preventing me from responding):
> Some people care more about compiler speed than the correctness?
Yeah, I think plenty of people writing code in languages that have concepts like Undefined Behavior technically don't really care as much about correctness as they may claim otherwise, as it's pretty hard to write large volumes of code without indirectly relying on UB somewhere. What is correct in such case was left up to interpretation of the implementer by ISO WG14.
Some people care more about compiler speed than the correctness? I would love to meet these imaginary people that are fine with a compiler that is straight up broken. Emitting working code is the baseline, not some preference slider.
> I would love to meet these imaginary people that are fine with a compiler that is straight up broken.
That's not what I said; you're attacking a strawman.
My point was more so that some people prefer the madness that is -funsafe-math-optimizations, or happen to rely on UB (intentionally or otherwise). What even is "correct" in the presence of UB? What is correct in such case was left up to interpretation of the implementer by ISO WG14.
Let's pretend, for just a second, that the people who do, having been able to learn how to program, are not absolute fucking morons. Straight up broken is obviously not useful, so maybe the conclusions you've jumped to could use some reexamination.
a compiler introducing bugs into code it compiles is a nightmare thankfully few have faced. The only thing worse would be a CPU bug like the legendary Pentium bug. Imagine you compile something like Postgres only to have it crash in some unpredictable way. How long do you stare at Postgres source before suspecting the compiler? What if this compiler was used to compile code in software running all over cloud stacks? Bugs in compilers are very bad news, they have to be correct.
> a compiler introducing bugs into code it compiles is a nightmare thankfully few have faced
Is this true? It’s not an everyday thing, but when using less common flags, or code structures, or targets… every few years I run into a codegen issue. It’s hard to imagine going through a career without a handful…
They found a bimodal distribution in failures over the lifetime of chips. Infant mortality was well understood. Silicon aging over time was much less well understood, and I still find surprising.
It's not that uncommon if you work in massive lowish level systems. Clang/LLVM being relatively bug free is the result of many corporate big tech low level compiler swes working with the application swes to debug why XYZ isn't working properly and then writing the appropriate fix. But compiler bugs still come up every so often, I've seen it on multiple occasions.
We're already starting to see people experimenting with applying AI towards register allocation and inlining heuristics. I think that many fields within a compiler are still ripe for experimentation.
Intuitively it feels like it should be a straightforward training setup - there's lots of code out there, so compile it with various compilers, flags etc and then use those pairs of source+binary to train the model.
Dennis Rodman grew up overshadowed by his sisters' basketball skills, and then had some unheard of growth spurt of 8" after finishing high school. He hadn't even played much high school ball.
Both Dennis Rodman and Hakeem Olajuwon are not 5ft, they are very tall and athletic. That combination is more important than basketball skill attained at 18 years of age. These attributes differs from tennis, or chess. Being elite at being both tall and athletic probably changes the most over puberty?
Basketball is probably not a great example since just being enormous gives you a huge chance of making it to the NBA, which I guess is just another form of being a prodigy.
Sure, and if we keep going back in time to perhaps the greatest American athlete of all time, Jim Thorpe - he'd handily be beaten by elite high schoolers today.
Basketball is a general purpose sport. The Claude of it can win. Some other sports such as gymnastics would need something more like the AlphaZero of it to win.
> Basketball is a general purpose sport. The Claude of it can win.
There aren't too many pro-ballers shorter than 5'10" (177cm), and definitely no dominant ones.
If we're defining "general purpose sport" as a sport in which people of all shapes and sizes are able to achieve greatness, then I would say soccer or golf fit that definition better.
Men's soccer in the 2010s was dominated by 2 of the best players in history: Cristiano Ronaldo and Lionel Messi. There's a 7 inch height difference between the two. Ronaldo is powerful and muscled, Messi is lithe and graceful. Both played in approximately the same position on the field, in the same era. Both were brilliant.
> Elon can't legally financially entangle Tesla to SpaceX
Bill Ackman has proposed taking SpaceX public by merging it with his Pershing Square SPARC Holdings, distributing 0.5 Special Purpose Acquisition Rights (SPARs) to Tesla shareholders for each share held. Each SPAR would be exercisable for two shares of SpaceX, aimed at enabling a 100% common stock capitalization without traditional underwriting fees or dilutive warrants.
With SpaceX IPO set to be one of the biggest of all time, this could have a pretty gnarly financial engineering impact on both companies -- especially if the short interest (direct or through derivatives) remains large.
Why would SpaceX go public? They already have a robust enough private market to give liquidity to all of their employees and shareholders who want it. They can get more private investment.
Going public would add a lot of hassle for little to no gain (and probably a negative of having to reveal their finances).
It has been widely reported for weeks that SpaceX is planning to go public in a few months. The reason is they have big plans to run a vast network of AI servers in orbit and will need to raise a massive amount of funding. xAI merger fits with that plan. I'd assume SpaceX still plans to go public.
Was ignored on HN but here's an article explaining:
It's all BS. There is no viable way to put industrial levels of compute into a space based platform that can work within the severe thermal, power, mass/volume, radiation, reliability, and economic demands. It is just stupid smoke blowing to separate idiot investors from their money. J-school grads don't have a clue what they're parroting about.
it wasn't ignored on HN, there were many articles correctly noting that building data centers in space is a stupid stupid idea because cooling things there is infeasible
Google, Blue Origin and at least 5 other smaller companies have announced plans to build data centers in space. My understanding is the cooling issue is not the show stopper you assume.
yup, bezos said "we will be able to beat the cost of terrestrial data centers in space in the next couple of decades". presumably this means they'll need huge ass radiators, so its all about bringing down launch costs since they'll need to increase mass.
Was doing some back of the envelope math with chatGPT so take it with a grain of salt, but it sounds like in ideal conditions a radiator of 1m square could dissipate 300w. If this is the case, then it seems like you could approach a viable solution if putting stuff in space was free. What i can't figure out is how the cost of launch makes sense and what the benefit over building it on the ground could be
AI datacenters are bottlenecked by power, bandwidth, cooling, and maintenance. Ok sure maybe the Sun provides ample power, but if you are in LEO, you still have to deal with Earth's shadow, which means batteries, which means weight. Bandwidth you have via starlink, fine. But cooling in space is not trivial. And maintenance is out, unless they are also planning some kooky docking astromech satellite repair robot ecosystem.
Maybe the Olney's lesions are starting to take their toll.
The shadow thing can be solved by using a sun-synchronous orbit. See for example the TRACE solar observation satellite, which used a dawn/dusk orbit to maintain a constant view of the sun.
Every telco satellite can cool its electronics. However, more than a few kW is difficult. The ISS has around 100kW and is huge and in a shadow half the time.
The cooling is the bit where I'm lost on, but it will be interesting to see what they pull off. It feels like everyone forgets Elon hires very smart people to work on these problems, it's not all figured out by Elon Musk solely.
Google, Blue Origin and a bunch of other companies have announced plans for data centers in space. I don't think cooling is the showstopper some assume.
Good call out, and really interesting. SpaceX being the cheapest way to get things into space, it seems like SpaceX is about to become extremely lucrative.
There was conflict between the new (old) hardware manufacturer Core Devices, and the Rebble community that's maintained the app store and software for the original Pebbles. I think they've worked it out, but it got ugly for a bit.
I am guessing if you have been doing this daily for a couple decades then the neurodivergence is not going through this. I assume any normal person will find it hard to not do any kind of work and if you spent 20 years of your life doing tech, how useful are in the "real" world. Unless you have been doing handy work on the sides, spoiler alert: not much.
They also have a history of buying good private companies at a good price and then let the management keep cooking. This is especially relevant to family businesses that want liquidity for the heirs and good long-term (indefinite) stewardship rather than selling to some PE vulture that will destroy their legacy.
reply