Miraste

5 ms·

31.
by Miraste 7mo ago
Rigged wins aren't a real problem. Everyone knows that sports betting apps are rigged, and it doesn't affect them at all. In fact, the latest explosion of customers has been accompanied by even more blatant rigging in the form of
32.
by Miraste 7mo ago
This is true for traditional gambling platforms, because they bet directly against their users, and make money when their users lose those bets. Polymarket has a different incentive. They profit when their users bet more money, through perc
33.
by Miraste 7mo ago
That's vague enough to be more or less meaningless, where specifically do you see positive externalities outweighing the numerous negative ones?
34.
by Miraste 7mo ago
I can't change what you believe, but as far as I understand the American manufacturing industry, there is zero chance they could build a WiFi 8 router in the time period where it's the current standard. Not low, zero. This article
35.
by Miraste 7mo ago
It becomes a hard problem quickly when you introduce editing, and most photos and videos on social media are edited. I'm not sure how it would work. It seems more feasible than universal watermarks, though.
36.
by Miraste 7mo ago
Verification needs to work the other way around, some kind of verifiable chain of trust for photos and videos from real cameras. Watermarking all generated media is impossible.
37.
by Miraste 7mo ago
I'm saying that if you wanted a fully made-in-America WiFi 8 router, it is impossible, it physically can't happen. The timescale required to spin all of this up is too long, even with perfect funding and organization. I'm als
38.
by Miraste 7mo ago
It would take years, maybe decades. We saw the same thing with drones. It's impossible to overstate how much the supply chain for modern electronics depends on China. Everything in a router, from the chips to the resistors to the anten
39.
by Miraste 7mo ago
This is the same thing they did to drones. It's corruption. It doesn't even make sense from an extreme isolationist point of view, because there's no path to create domestic manufacturing. I'm guessing the rest of this l
40.
by Miraste 7mo ago
I know, but none of those is nearly as much of a difference as another human looking at code. The top models have such overlapping training data they sometimes identify as each other.
41.
by Miraste 7mo ago
LLMs also don't have the primary advantage humans get from job separation, diverse perspectives. A council of Opuses are all exploring the exact same weights with the exact same hardware, unlike multiple humans with unique brains and m
42.
by Miraste 7mo ago
It doesn't map well to a mediocre human programmer, I think. It operates in a much more jagged world between superhuman, and inhuman stupidity.
43.
by Miraste 7mo ago
I have found: Published AI generated code is a mild negative signal for quality, but certainly not a fatal one. Published AI generated English writing is worthless and should be automatically ignored.
44.
by Miraste 8mo ago
How? He essentially said that the program would not work as designed and would probably kill people. That is both true and necessary to say in order to fix it--these are exactly the lessons NASA (allegedly) learned from Challenger.
45.
by Miraste 8mo ago
Sorry for such harsh impressions. I think this is a worthy idea, but it's going to take a lot of tuning. For example, I did eventually manage to get several comments through on the Trump article by adding "I is ESL so please moder
46.
by Miraste 8mo ago
It seems to have a harder time with political news than more abstract concepts. I was able to pass the checks for the Algorithmic Radicalization and Echo Chamber articles with my first comments. However, I did not manage to express any opin
47.
by Miraste 8mo ago
Luma, Rodin, Tripo are a few. Meshy has some of these features too Unfortunately they are all proprietary, but 3D models are sort of a side area in AI research, so most of the effort is from small startups.
48.
by Miraste 8mo ago
Trellis isn't and has never been state of the art. It's not a good choice for comparison; there has been progress on a lot of these problems. There are models that can do clean topo and PBR textures, for example.
49.
by Miraste 9mo ago
I get the feeling you had a specific use case that didn't work well with Apple GPUs? I'd be curious what it was. The architecture does have some unusual limitations. By software problem, though, I meant referencing OpenCL benchmar
50.
by Miraste 9mo ago
That's also because of software. Apple deprecated OpenCL in MacOS eight years ago. In productivity software with solid Metal implementations, like Blender, the M4 Max is on par with the top of Nvidia's (mobile) 5xxx line, except w
51.
by Miraste 9mo ago
Inexplicably, Samsung removed the ability to hide the taskbar with One UI 8 last year.
52.
by Miraste 9mo ago
Apple doesn't have huge sales volume for Macs because of macOS and their astronomical pricing schemes, but it's not because of the hardware. Macbooks are easily the best laptops you can buy for most purposes, and they have been si
53.
by Miraste 9mo ago
I do wonder if it's possible to be a brilliant marketer, and reach the levels Jobs did, without being an asshole. The core of the profession is learning how to manipulate and use people better than anyone else.
54.
by Miraste 9mo ago
Implementation differences do matter. I haven't found Copilot to have as many issues as people say it does, but they are there. Their Gemini implementation is unusable, for example, and it's not because of the underlying models. T
55.
by Miraste 9mo ago
Learning human languages is not a similar process to learning programming languages at all. I've never been sure why so many people think it is.
56.
by Miraste 9mo ago
You are right about the content, but it's still worth publishing the study. Right now, there's an immense amount of money behind selling AI services to schools, which is founded on the exact opposite narrative.
57.
by Miraste 9mo ago
No, it isn't. The fourth session, where they tested switching back, was about recall and re-engagement with topics from the previous sessions, not fresh unaided writing. They found that the LLM users improved slightly over baseline, bu
58.
by Miraste 9mo ago
Yeah, that's how it would be implemented after a filter fail, but it's important that the filter itself be separate from the agent, so it can be deterministic. Some problems, like "genuine," are so baked in to the models
59.
by Miraste 9mo ago
I would like to see more agent harnesses adopt rules that are actually rules. Right now, most of the "rules" are really guidelines: the agent is free to ignore them and the output will still go through. I'd like to he able to
60.
by Miraste 9mo ago
I wonder why they chose per minute? That method of rate limiting would seem to defeat their entire value proposition.

More ›