HNHacker News
TopNewBestAskShowJobs

radarsat1

4,157 karma · joined May 28, 2011

Info: https://sinclairs.gitlab.io

Email: <myusername>@gmail.com

submissionscomments
radarsat1··on Run CUDA, unmodified, on AMD GPUs
Selection bias. I'm sure there are lots of people who are really good at CUDA and don't have those kind of assets. Not everyone knows how to sell their skills.
radarsat1··on ULID: Like UUID but Sortable (2019)
This is exactly why I'm using ULID with DynamoDB.
radarsat1··on Free-threaded CPython is ready to experiment with
And if you specify that they are numbers then you lose the ability of the function to generalize to vectors.

Indeed assuming it adds two things is correct, and knowing that concatenation is how Python defines adding strings is important for using the language in the intended way.

radarsat1··on A.I. Has Become a Technology of Faith
Yesterday I needed to make a simple GUI app for a specific purpose. I haven't done it in a while so I figured it would take some time, but just to try, I drew a picture of the interface I had in mind and told 4o to generate a PyQt5 app for me, and it ... did it. Still took a bit of back and forth for some details but damn if it didn't get me from A to B way faster than if I'd been sitting there figuring out VBoxLayout etc again from scratch after several years. I was impressed.
radarsat1··on Dola Decoding by Contrasting Layers Improves Factuality in Large Language Models
It's definitely a philosophical position that exists beyond a random HN poster. See for example computational theory of mind https://plato.stanford.edu/entries/computational-mind/
radarsat1··on C++ patterns for low-latency applications including high-frequency trading
> A creeping feeling that causes them to compulsively run the benchmarks through callgrind

I'm happy I don't deal with such things these days, but I feel where the real paranoia always lies is the Heisenberg feeling of not even being able to even trust these things, the sneaky suspicion that the program is doing something different when I'm not measuring it.

radarsat1··on The African workers driving the AI revolution, for about a dollar an hour
Sure but my point is that you could write exactly the same article without mentioning AI at all, since the majority of examples are about content moderation for social media.

Meanwhile the only connection to AI is that it might provide a potential reduction in the need for this.

So what is odd to me is positioning this as the purpose being for feeding the ever hungry data needs of evil algorithms, because as far as I can tell without those algorithms these jobs would still exist and would be worse.

radarsat1··on The African workers driving the AI revolution, for about a dollar an hour
I find it a bit odd how this article conflates content moderation for social network with AI training. Of course there is a relationship there but these are not the same thing. The article gives lots of examples of trauma caused by content moderation and then blames it on the needs of AI. If anything, if we could satisfactorily train an AI for content moderation then humans wouldn't need to be in the loop (as much) so it's a weird thing to just put these ideas together like that, and is clearly just trying to feed the current zeitgeist of "AI bad".
radarsat1··on SCIM: Ncurses based, Vim-like spreadsheet
Are there any good emacs modes for spreadsheeting?
radarsat1··on The super effectiveness of Pokémon embeddings using only raw JSON and images
It would be nice if you spelled out on your post how you know this, then. Is it written somewhere? A relevant paper for example?
radarsat1··on The super effectiveness of Pokémon embeddings using only raw JSON and images
It's a good point about true and false positives though, which makes me wonder if anyone's taken a large database of expected outputs from such "equations" and used it to calculate validation scores for different models in terms of precision and recall.
radarsat1··on How to fix “AI’s original sin”
I guess part of my issue with the argument is that creators seem to want to have their cake and eat it too. They want to put their work out into the public, but then they want to also control what people can do with it. These two themes, although I can see the need and logic for them, seem to be totally antithetical on a practical basis.

In some sense art creation is, indeed, a risk.. you put something out there, and then you hope it is perceived well, and hope that it isn't abused. That's what makes making art such a difficult thing.. you are putting something out there, a piece of yourself.. but not acknowledging the associated risks, that someone might hate it, that someone might not understand it, or indeed that someone might "sample" your work into a new context, is being totally disingenuous with yourself as an artist.

Of course I have no issue with the idea that artists should be fairly compensated, and the whole legal idea of copyright exists exactly for that, and I'm incredibly interested to see how this whole thing plays out in the legal sense. But I have to say that I'm not entirely convinced that using publicly available sources for training AI, that don't explicitly come with copyright terms that forbid it, is in any sense unethical, because I think it's entirely possible that it may end up being considered fair use.

Artists want to get paid, yes, but the whole "I didn't get to opt out" argument doesn't really sit well with me, because the degree of post-publication "control" being implicitly demanded in that argument doesn't seem to be very balanced. In fact you did get to opt out, software for example comes with extensive legal documents describing exactly how derivative works can use it, but books and photos generally don't, and maybe in the future they will have to, but the idea that retroactively interpreting the role of copyright on existing works to include special considerations for AI just doesn't seem that obvious. You might not like that you published work in a way that someone was able to use it in a way that you disapprove of, without explicitly forbidding said usage, but the fact is that you did. The rest is for the courts to sort out. And I think it could go either way.

radarsat1··on How to fix “AI’s original sin”
Sure but I was responding specifically to a scenario proposing that rich people would develop fully automated factories thereby cutting out their reliance on workers. Getting even halfway there still leaves plenty of room for revolution, I'm just responding to the 100% automated scenario as being unrealistic.
radarsat1··on How to fix “AI’s original sin”
So, I generally agree with you, but my question to you is how do I square that with thinking this technology is legitimately amazing and I want to see more of it and use it? Do you forsee any future in which creators are happy and this technology gets to exist?

You mention getting paid, this might be part of it, but (a) how do you manage payments when so many sources contribute such a small proportion of the input and (b) would this actually solve the problem or would creators continue to object despite getting paid according to their representation in the dataset?

I imagine that you will answer that (a) is "their problem" so feel free to focus on (b).

radarsat1··on How to fix “AI’s original sin”
Ok but all those dictatorships are enforced by humans, not robots.
radarsat1··on How to fix “AI’s original sin”
> before the plutos can amass power via fully automated factories

Fun science fiction but this idea that the rich could develop an entire self sustaining infrastructure to protect themselves from any leverage from the working class is completely infeasible.

Let's day a full automatic drone factory was built. Factories don't make their own tools and general parts, they don't mold their own screws and bolts, no, they source those. Even if they could, they'd need raw materials. So they'd need completely automated mines. Those mines would be on land that they'd own, but they'd have to physically defend it. So the scenario you're proposing is: fully automated mining, with fully automated defense measures, delivering raw materials to some fully automated smelter (all travel routes automatically defended against sabotage), this all to feed a factory that would churn out defense robots. It's completely circular and silly.

radarsat1··on Fungus breaks down ocean plastic
Yes, I see your point here. I think most objections to plastic including my own are about the longevity of its effects on ecosystems, but local, short term effects on health are equally problematic.
radarsat1··on Fungus breaks down ocean plastic
As long as it's not an efficient process then you have a point. But, let's say some very efficient way to clean up and break down plastic was discovered. Then, wouldn't the manufacturers actually have a point? In that case it indeed would be harmless and we could enjoy the conveniences of it while being relatively certain we can then dispose of it safely. So you wouldn't call it an "abuse" in this case.
radarsat1··on Show HN: Shpool, a Lightweight Tmux Alternative
How would you tail /dev/null?
radarsat1··on Show HN: Shpool, a Lightweight Tmux Alternative
Looks nice. One feature I'd like is a read-only mode. Sometimes I want to execute a long running process and be able to just watch it run without risking that I hit Ctrl C by accident. I was surprised to find that Screen doesn't support this, I guess probably tmux does. tail is fine but I don't always want to record a huge log file.
radarsat1··on How Meta trains large language models at scale
Frustratingly little information. For example, I'm exceedingly curious how they deal with scheduling jobs on such a huge array of machines. The article:

> Efficient scheduling helps ensure that our resources are used optimally. This involves sophisticated algorithms that can allocate resources based on the needs of different jobs and dynamic scheduling to adapt to changing workloads.

Wow thanks for that, captain obvious. So how do you do it?

radarsat1··on Tomorrow people: For a century, it felt like telepathy was around the corner
Heh this reminds me of a joke from an old colleague, who would say, "forget working on a great demo! let's work on technology that makes you feel like you've just seen a great demo!"
radarsat1··on Most life on Earth is dormant, after pulling an 'emergency brake'
But the better the mean surface is fitted (in a generizable way), the easier it is to spot outliers.
radarsat1··on Mamba-2 – State Space Duality
Seems like since Mamba 2 is a constrained version of Mamba 1, maybe Mamba 1 could be used in a fine-tuning stage.
radarsat1··on James Webb Space Telescope Finds Most Distant Known Galaxy
> less sophisticated

lol.

Thanks for this breakdown, it was super interesting to read about what signal processing techniques are used for this kind of thing!

How does the actual "searching" take place, do they just run this kind of procedure at every bright spot in the image and rank candidate sections?

radarsat1··on Show HN: Boldly go where Gradient Descent has never gone before with DiscoGrad
> Why do you think similar approaches never landed on jax?

Isn't this just adding noise to some branching conditions? What would take for a framework like Jax to "support" it, it seems like all you have to do is change

> if (x>0)

to

> if (x+n > 0)

where n is a sampled Gaussian.

Not sure this warrants any kind of changes in a framework if it's truly that trivial.

radarsat1··on I want flexible queries, not RAG
But as opposed to a dancing bear, we expect AI to get better and better at dancing in the next decades.

The speculation is important, it both changes the perspective for prospective companies/investors, and cannot even be said at this point to be unfounded.

radarsat1··on I want flexible queries, not RAG
Once tree ensembles get big enough to handle the kinds of problems that LLMs can address, are they really more interpretable?
radarsat1··on Player-Driven Emergence in LLM-Driven Game Narrative
I would have thought so too, until about 2023.

But, if recent attitudes to AI are anything to go by, this will be immediately seen as boring and derivative at best, or at worst people will accuse the back stories as having been "stolen" from other works of fiction.

While my expectations for the capabilities of AI have never been higher, my expectations for how the public will perceive this amazing technology has never been lower. It's all very disappointing.

And that's without specifically placing blame by the way, perhaps it's big techs fault for going about this the wrong way, I don't know, but the public blowback against AI is just making working in this field that I used to love super depressing.

radarsat1··on New black hole visualization takes viewers beyond the brink
One thing I've always wondered about these kinds of simulations is how they deal with numerical issues, since I assume they are needing to use both very small and very large numbers. Additionally, even in simple classical physics integration errors can add up very quickly, so wonder how this problem is avoided when working with these kinds of scales. Similar thoughts for things like galaxy simulations, or simulations of planet-sized collisions, etc.
← PreviousPage 7 of 34Next →