Live data from Hacker News

The tiny corp raised $5.1M

geohot.github.io

311–320 of 331 posts

Re: The tiny corp raised $5.1M

#311
post #279
post #256

Earlier quoted context omitted.

> Technical people suffer from what I call "Engineer's Disease". We think because we're an expert in one area, we're automatically an expert in other areas. Just recognizing that helps. https://news.ycombinator.com/item?id=10812804

I think the more interesting question is why this symptom mostly happens to "engineers". I've seen enough engineers presume they can easily become experts in law; I haven't seen many lawyers presume they can easily become experts in engineering. Why?

IMO, there's a certain level of arrogance intrinsic to engineering: To build something new, you need a belief, first and foremost, that you can build it at all, and almost as importantly better. Weeding out all the people who don't have, at least to some degree, that belief, and you end up with a disproportionate fraction of people who think that way about everything.

Re: The tiny corp raised $5.1M

#312
post #282

Earlier quoted context omitted.

It only has a definitive answer in the mainstream interpretation of mathematics. On the relative fringes, there are serious studies on alternative interpretations. See for example https://en.wikipedia.org/wiki/Constructivism_(philosophy_of_... (You can skip to the part that discusses Cantor's arguments, but I suspect that if you haven't heard about related concepts you probably want to understand what it is first.)

I love that the "hater" thread turns into a discussion of uncountable infinities :) The "cardinality" section of that Wikipedia page describes my objection well. I don't doubt the real numbers are not recursively enumerable, but that doesn't mean they have a larger cardinality than the integers.

I stand by my statement. Pony up or shut up.

If you're trolling to troll, then expect the hate because you're being annoying.

If you think you're right and all the mathematicians are wrong, pony up. Hell, you'll have a lot more lulz when you win a fields metal.

If you don't pony up I don't know why you would be surprised people assume you're arrogant. You can doubt the status quo without being arrogant. We both know that you're not going to take someone's word just because they said so, so why expect others?

Re: The tiny corp raised $5.1M

#313
post #282
post #206

Earlier quoted context omitted.

It is quite thoroughly studied in mathematics, and that particular issue has a definitive answer.

It only has a definitive answer in the mainstream interpretation of mathematics. On the relative fringes, there are serious studies on alternative interpretations. See for example https://en.wikipedia.org/wiki/Constructivism_(philosophy_of_... (You can skip to the part that discusses Cantor's arguments, but I suspect that if you haven't heard about related concepts you probably want to understand what it is first.)

No, I don't think there's very much room for controversy here. I mean, I don't know what exactly Hotz have said, since there was no quote (and, honestly, I'm not that interested either), but if somebody is simply saying "the cardinality of integers is the same as real numbers" and leaving it at that, he is just plainly wrong.

Math is all about definitions and what follows from these definitions. So, you can define "integer", "real number", "cardinality", "equals" and so on however you like, and make all sorts of correct statements — as everyone will see by following your arguments all way from the definition/axiom and until the very end of your proof. But if you don't provide any definitions of your own, then you rely on some other definitions, and everyone has no other choice than assume that these are the very much "mainstream" ones, as you are referring to them as if they are well-known.

Now, it is unquestionably true (and easily provable) that the set of all computable numbers is countable, and anyone who says otherwise is wrong. But unless you specifically define real numbers as a subset of computable numbers, as constructivists are inclined to do, your listeners won't assume that, since this is not how real numbers are generally defined, and by virtue of not providing your own definition you are implicitly referring to a "general definition". (And, honestly, you shouldn't even call any subset of computable numbers "a set of Real numbers": this name is already taken.)

These general definitions and assumptions lead to all sorts of complications, and I personally have my doubts that real numbers exist in any meaningful sense (although I'm not committed to that statement, since there are several mathematical constructs that I would like to dismiss as "clearly nonsense", except they allow us to prove some very "no-nonsense" stuff — I don't know how to deal with that, and I never heard that anybody does). But I definitely cannot say that cardinality of integers is the same as the cardinality of reals, because this is simply not true under the common definitions (which is easily provable). (And less importantly, but worth saying that the contrary is not proven by constructivist methods — as half of the actually useful math in general ends up being, unfortunately).

So, as a somebody, who doesn't quite believe in non-computable numbers, I am very sympathetic to anybody who says that Real numbers do not exist. I don't understand how could they, what does it mean for an object that we cannot define to exist. Yet, I can accept (as a game) some well-known theory which talks about these non-quite-existant "Real numbers", and prove some statements about them, and one of these easily provable statements is that cardinality of continuum does not equal the cardinality of Natural numbers.

Re: The tiny corp raised $5.1M

#314
post #313
post #282

Earlier quoted context omitted.

It only has a definitive answer in the mainstream interpretation of mathematics. On the relative fringes, there are serious studies on alternative interpretations. See for example https://en.wikipedia.org/wiki/Constructivism_(philosophy_of_... (You can skip to the part that discusses Cantor's arguments, but I suspect that if you haven't heard about related concepts you probably want to understand what it is first.)

No, I don't think there's very much room for controversy here. I mean, I don't know what exactly Hotz have said, since there was no quote (and, honestly, I'm not that interested either), but if somebody is simply saying "the cardinality of integers is the same as real numbers" and leaving it at that, he is just plainly wrong. Math is all about definitions and what follows from these definitions. So, you can define "i…

> I personally have my doubts that real numbers exist in any meaningful sense

I think we're in agreement here in principle. (Since we're on the topic, I'd like to add that naming this suspicious set "real" numbers is a tad bit ironic)

That said, I don't like the idea of having a group of people "owning" words as if they had a monopoly over them. The statement "the cardinality of integers is the same as real numbers" can be understood to mean "real numbers should actually be computable numbers".

I didn't bother to look up what Hotz wrote on twitter that triggered this discussion, I was just providing context that the issue of cardinality isn't as settled as some might think. It's probably not fruitful to argue whether a statement from hearsay uses words accurately or not though.

Re: The tiny corp raised $5.1M

#315
post #197

Earlier quoted context omitted.

This 95% of models are statically computable thing really shows how much he is trivializing this problem. I’d be interested to see his SW stack compile MaskRCNN. His ISA is massively under-defined and people will not change their model code to run on this accelerator unless his performance beats cuda significantly and even then they still won’t - usability matters more than performance every time. In the end you need…

I agree writing something that needs to be compatible with, say, PyTorch is a significant undertaking, but why is that necessary? I also agree some models like MaskRCNN is not static, and people will not change their model code, but I don't think it matters. Let's say you want to run LLaMA. LLaMA is a tiny amount of code, say, 300 lines. LLaMA is static. It doesn't matter people will implement LLaMA with PyTorch and…

Yeah for inference this is true, there could be a viable subset of models. You’re not going to build a viable business on inference though. It’s super cheap already and plenty of hardware can do it ootb with an existing framework as you’re saying. The $$ for selling chips is in training, and researchers trying new architectures are not going to wait for a port of their favorite model in a custom DSL or learn a new language to start prototyping now. You can port models forever, but that isn’t an ecosystem or a cuda compete. OpenCL + AMD != a from scratch company

Re: The tiny corp raised $5.1M

#316
post #306
post #226

When you click on the strip link to preorder the tinybox, it is advertised as a box running LLaMA 65B FP16 for $15000. To be fair, the previous page has a bit more details on the hardware. I can run LLaMA 65B GPTQ4b on my $2300 PC (built from used parts, 128GB RAM, Dual RTX 3090 @ PCIe 4.0x8 + NVLink), and according to the GPTQ paper(§) the quality of the model will not suffer much at all by the quantization. Just sa…

Are you able to memory pool two 3090s for 48gb and if so what's your setup? I looked into this previously[1] but wasn't super confident it's possible or what hardware is required (2x x8 pcie and official SLI support?). AFAICT still would look like two GPUs to the system. [1] https://discuss.pytorch.org/t/is-there-will-have-total-48g-m...

You can memory pool with the right software however pytorch supports spreading large models over multiple GPUs OOTB. Just pass the --gpu-memory parameter with two values (one per GPU) to oobabooga's text-generation-webui for example.

Re: The tiny corp raised $5.1M

#317
post #226

When you click on the strip link to preorder the tinybox, it is advertised as a box running LLaMA 65B FP16 for $15000. To be fair, the previous page has a bit more details on the hardware. I can run LLaMA 65B GPTQ4b on my $2300 PC (built from used parts, 128GB RAM, Dual RTX 3090 @ PCIe 4.0x8 + NVLink), and according to the GPTQ paper(§) the quality of the model will not suffer much at all by the quantization. Just sa…

What case/MB/GPUs do you use for your dual 3090 build? Liquid cooled cards?

I'm using air cooling. Gigabyte X570 Pro, RTX 3090 FE, be quiet! Pure Base 500DX mesh case with four 140mm fans currently. It's not quiet under heavy load!

The GPUs have gotten improved thermal pads. The 8 core Ryzen 3700X is a 65W model. It appears to be fast enough not to be a bottleneck for this purpose. 1200W PSU.

I may swap the fan in front of the GPUs for a high-rpm model.

Also for longer runs i throttle the GPU power draw. It doesn't cost much performance.

Re: The tiny corp raised $5.1M

#318
post #262

Earlier quoted context omitted.

Good thing everyone lives in Phoenix and SF XD

If it works in Phoenix there's no reason it wouldn't work in a large part of the US, they just haven't rolled it out yet. Proving that it works is the hard part.

How so?

Different roads, geography, etc?

Re: The tiny corp raised $5.1M

#319

Earlier quoted context omitted.

If it works in Phoenix there's no reason it wouldn't work in a large part of the US, they just haven't rolled it out yet. Proving that it works is the hard part.

How so? Different roads, geography, etc?

Sure Phoenix is an easy case with dry weather and car-based infrastructure but there are huge areas of the US that are fundamentally the same. If they can solve the problem of driving safely and conveniently in Arizona they can set it up elsewhere too.

Re: The tiny corp raised $5.1M

#320
post #196

Earlier quoted context omitted.

My opinion of geohot definitely dropped after he started tweeting how easy it would be to fix Twitter, and then he started soliciting free work. He obviously underestimated the difficulty of shipping a feature across web and mobile. Hacking a prototype is trivial. Making it work well for all platforms, fully accessible, and across all supported languages is a bigger hurdle. It just gave me the impression that he thou…

Why is Twitter such a Waterloo for all these obviously accomplished people? It seems like they've been assuming Twitter is the way it is because it was staffed by technically incompetent leftists, and if only they could apply their own get-things-done attitude and "neutral" politics, then the problem would be trivially fixable. Where does this fallacy come from? Is it because of the illusory simplicity of the tweet f…

I'm a senior developer, and I have to admit I'm one of those guys ;).

WhatsApp was 55 people big when they got acquired, and to me that sounds about right.

Twitter employed 7,500 people. 7,500!!!! So please tell me where the complexity lies? Surely not in the front-end code I can tell you that.

Let's compare it to something WAY-WAY-WAY more complex, like a game with multiplayer, awesome mod tools, etc.: ROBLOX: 2,200 employees. Do I need to mention they wrote their own physics simulation engine and keeping realtime multiplayer going?

So please, explain this to me: how is Twitter more than 3 times more complex than Roblox???

Maybe I'm wrong, that's very possible, I've been wrong in the past. But just explain this 1 thing then: Twitter needs more than 3 times the manpower than Roblox?

Post reply on HN