dadoum · HackerLangs

eergisterendiscuss

I tried to one-shot the first test (the Rubik's Cube test) with LucidQuery's Swift model, to test it, as there are not much benchmarks about it and that they brag a lot about it, and I was pleasantly surprised to see it achieving a result similar to Grok 4.5 but in one shot (there is the same issue that if you scramble twice the solve button does not work anymore, but it got it in one shot).

Though it crunched most of the free quota, 47111 tokens, so I couldn't make multiple attempts.

dadoum

9 dagen geledendiscuss

Still, I don't want to gate people based on age.

Parents should at least be able to overwrite the age of their child, maybe selectively allow bypasses. My experience with a computer would have been completely different if I was blocked from half of the internet. Especially when I see which kind of content gets blocked.

dadoum

9 dagen geledendiscuss

From https://z.ai/subscribe#code-plans-container:

> Explanation and Recommendations Regarding Usage for Plan-Supported Models

> Note: Peak hours are from 14:00 to 18:00 daily (UTC+8).

dadoum

12 dagen geledendiscuss

DMA and DSA are very easy to comply with, there is an easy heuristic function to know if your AI startup is affected:

    |_: Company| false

DMA and DSA only applies to companies explicitly designated as gatekeepers according to EU law. There are criteria for that, about scale and market control, but generally you really don't have to worry, you will see that coming from very far.

dadoum

12 dagen geledendiscuss

VSORA already finished their inference chip if you want to take a look at what they're doing. They got a handful produced at TSMC 5nm node. They are expected to be offered by Scaleway at some point in the near future (alongside with other French components from the AION consortium, like a SiPearl Rhea1 CPU and an Eviden Bull BXIv3 interconnect, all of them are also already produced, only integration is remaining). It will still probably take some more time though, as the AI stack needs to be finalized (ZML), but I am surprised that those developments are not being much discussed.

dadoum

13 dagen geledendiscuss

I don't think this narrative stands, industry and social programs coexisted for a very long time. I actually believe the opposite. It sacrificed its industry for intra-EU fairness and integration. A lot of industries were very reliant on gov grants and other state social programs that have got scarcer as EU gained power.

> Facts, but the Europeans, tankies, and America haters here will never admit it.

When disagreeing, please reply to the argument instead of calling names. "That is idiotic; 1 + 1 is 2, not 3" can be shortened to "1 + 1 is 2, not 3."

dadoum

15 dagen geledendiscuss

If I remember correctly a lot of AOSP core apps have been discontinued though.

dadoum

17 dagen geledendiscuss

> May we scale smoothly, exponentially and uneventfully through A[SI]

That sentence sounds weird to me. I can't really put my finger on why, maybe the combination of adverbs, or just the fact of writing the desire of scaling as a company so directly. It feels (to me) like openly claiming their selfish goals. Or maybe I am just misinterpreting and they are referring to the whole humanity as "We" (but knowing Broadcom and in a lesser extent OpenAI doings, I am not convinced).

dadoum

24 dagen geledendiscuss

So I downloaded the model and tried a few Math prompts. The simple addition was a little tedious because it checked multiple times that the calculation was right, I then gave it a quite long integral to solve but which is straightforward if you know the techniques, and it got it in 5 minutes on my Macbook Pro M4 Pro 24 GB, I just had to increase the context window. I finally tried giving it a full math exam but here it wouldn't score much points as it takes so many shortcuts it writes wrong steps in its answers. Still pretty good as it generally identifies what it should do, but I did not try anything in that weight class before so I can't really talk if that's impressive in the full picture.

dadoum

24 dagen geledendiscuss

> socialists

They are not socialists by any mean. EU is on every front very neoliberal (way too much for my French taste), and it has been like that since its inception without real interruption.

dadoum

vorige maanddiscuss

I am also working on a programming language with that no hidden behavior philosophy at the forefront (along with a few other things), but radically different on everything else, so I may have a few design choices to question.

I think that your conception of cleverness is too wide basically. It's true that, in my opinion, any new programming language today should stop people from C-style "cleverness". But there are languages where the clever programs are actually the pretty ones. For example, the borrow-checker in Rust encourages good code design (and yet it still can be better). And most importantly, algebraic types constitute a radical improvement over classic C-style code. You may think that they hide what's happening, as they have a hidden tag and could be built from C structs and C unions. But after thinking a lot about it, thinking about the tag is actually not thinking about the program's behavior. There is a difference between the program you want to write, and what needs to happen on the machine to compute it. I could probably write a lot about the solution I found to reconcile both opinions, but my project is a lot newer so I can't redirect you to anything yet. But I think that by wanting a more grounded language, you are maybe dismissing too many ideas that would actually encourage programs to be clear and transparent while still not doing anything automatically with magic (and at the same time, would dismiss incorrect programs, that C can express but that machines cannot execute).

Btw, I understand that I am probably not the target of your language, so I won't expand much further unless you want.

dadoum

vorige maanddiscuss

I think they are saying that the US is required to pay less than the EU?

But that's not true based on what I can find anyway, https://app.23degrees.io/view/j4luMuv8fnpO2frL-bar-grouped-v...

dadoum

2 maanden geledendiscuss

I want to highlight that maybe today, big conglomerates are rare, but this is also because during the late-20th century, the trend was to break up conglomerates to increase competitiveness and improve financial performance of companies by focusing on the best businesses. If you look at the situation before that moment, Japan's situation would still be on the extreme side when compared to the other developed nations at the time, but not as unique I think.

In retrospect, I tend to think that this take was naive. It probably increased financial performance but it discouraged taking risks, and pushed the multidisciplinary skills out of companies in a way that is hard to reverse, inducing knowhow loss and probably slowed down innovation. But this is only my personal analysis and I am no economist.

dadoum

2 maanden geledendiscuss

Correct me if I am wrong, but your algorithm looks to me like it would work fine with posits? You would just get a NaR value instead of both NaN and inf, because div-by-zero and sqrt(-) both yield NaR (the difference here with floats is that it is unique and can get compared to NaR), and so it just works fine?

dadoum

2 maanden geledendiscuss

And that's the whole reason why Wero has been made I think. It's because the ECB wants to advance on their digital euro plans due to sovereignty concerns, and I think this push is to dismiss that argument.

dadoum

2 maanden geledendiscuss

The compiler has been written by a lot of very smart people, it is very well tested and sometimes has even been formally proven to output exactly what the language specifies, and I have a mental model of what the code it outputs does in relation to what I write.

Nobody can be sure of what the LLM will output for a certain prompt. If you don't review what it outputs, it will not necessarily match your expectations. You could argue that it is the same as when you assign an intern to the task, but I personally would check what the intern writes (and in my experience they are more reliable than current AIs, of course not as quick).

dadoum

2 maanden geledendiscuss

And you are wrong to think that in my opinion, chip manufacturing in Europe was huge 25-30 years ago (there was high-end memory chips manufacturers, high-end GPU manufacturing and cutting edge nodes in Europe at the time).

dadoum

2 maanden geledendiscuss

Actually there are more if you count the ones which are not at the cutting edge but your point still stands, most high-end silicon companies only do design.

dadoum

2 maanden geledendiscuss

ARM design IP blocks, they can make their own CPU (and now they are making one), eevn though that means competing with your customers.

dadoum

2 maanden geledendiscuss

In general, in Europe there is research infrastructure that I think could be used at a medium scale for important applications (but I am not a professional).

There is the NanoIC research line at imec (2nm), CEA-Leti incomming 7nm FD-SOI pilot lines, and in terms of full production lines, Global Foundries Dresden (12 nm), ESMC (12 nm, in construction), and the various FeRAM/FMC projects I can't keep track of (Neumonda for example).

I would be more worried about designs, because outside of ARM (and Imagination Tech, both in the UK), I don't know any competitive European designs. (about routers NXP already makes router chips with accelerators on top of ARM cores, used for example in the Mono Gateway, but they are fabbed on old TSMC nodes)

comments