hasperdi

Can we trust the NTSB these days?

Unfortunately they'll learn nothing. The rest of the world however... have to endure the consequences

Numerous prompting will cause prompt fatigue, similar to pressing yes on a dialog boxes.

LLM, like fire is a powerful tool. Some people play with fire and achieved great things, some play with fire and got burned. A number of them achieved great things and got burned. We need to understand that and learn from our mistakes.

It's "artificial" intelligence

So which one is recommended for running on servers in production?

this is just regurgitating the manufacturer's claim. I believe it when I see it. Most of display energy use is to turn on the OLED/backlight. They're claiming, because our display flickers less, it's 48% more efficient now.

I don't think that's a realistic suggestion as as the quantity of applications are huge who are going to spend time reviewing them one by one. And and even then it's not realistic to expect that that undesirable things can be detected as these things can be hidden externally for instance or obfuscated

It's not 3x because of 3 runs; can be more token, can be less.

The way of thinking it is, telling Claude to tackle the problem 3 times, each time it may or may not use different approach, fix or improve on things it did previously.

This. Postgres is slow for large inserts compared to eg. Clickhouse.

This slowness is mostly because of OLTP features

This mirrors my experience.

If you want the agent to be more aware of the tool, prompt it to create a skill for whatever bash commands you wish

It's a dumb idea actually, as these panels will be covered in brake dust, sand, leaves, etc.

If you're interested, check EEVblog's video on this

No the murders happen in camps eg. What's happening to the Uyghurs.

That said, they also use them as slave labors.

Maybe that's what ICE is going to do with the plan to setup large detention centers in the US

you can use Claude Code with these models. You just need to pass the right env vars. Have a look at the client setup guide on z.ai

Why distill, if you can run the full model yourself... or at other inference providers.

Quantization the better approach in most cases, unless you want to for instance create hybrid models ie. distilling from here and there.

That's a bike that's ergonomically designed for pelicans.

It is unreasonable to expect pelicans to ride human bikes, they have different anatomy.

Fun! Any other games with REST API?

The question is: what do they expect those companies to do?

Just like subjects in other fascist regimes, be damned if they do comply, be damned if they don’t.

It sounds like someone saying power tools should be outlawed in carpentry.

I see this a lot here

Retail PCs will probably never catch up to even the open‑weight models (the full, non‑quantized versions). Unless there’s a breakthrough, they just don’t have enough parameters to hold all the information we expect SOTA models to contain.

That’s the conventional view. I think there’s another angle: train a local model to act as an information agent. It could “realize” that, yeah, it’s a small model with limited knowledge, but it knows how to fetch the right data. Then you hook it up to a database and let it do the heavy lifting.

Well, it depends on the hardware you have. If you have a hardware locally that can run best open models, then your local models are as capable as the open models.

That said, open models are not far behind SOTA, less than 9 months gap.

If what you're asking about those models that you can run on retail GPUs, then they're a couple years behind. They're "hobby" grade.

AFAIK Zuck got mad and restructured the whole department.

What's likely is that there won't be anything open / significant coming out from them anymore

Another thing... they alter the localStorage & sessionStorage prototype, by wrapping the native ones with a wrapper that prevent keys that not in their whitelist from being set.

You can try this by opening devtools and setting

  localStorage.setItem('hi', 123)

Say you were the Mayor of London, and being a great mayor you have your priorities 100% correct.

Can you guarantee that something like this will never happen on your watch?

Speaking as a user of Openpilot / Comma device, it is exactly what the Wikipedia article described. In other words, it's a level 2 ADAS.

My point was, he had more than naive / "pedestrian level" (pun?) understanding of the problem domain as he worked on Comma.ai project for quite some time; even the device is only capable of solving maybe about 40% of the autonomous driving problem.

He funded Comma.ai, so he does understand the problem domain & complexity.

On the other hand, it depends on what kind of movie you're making and who the target group is.

Say you're making children's videos like Cocomelon or Bluey in 3D, you don't need all these nice things.

At the end, movies are about the stories, not just pretty graphics.

The YouTube channel Calum made a documentary about it https://m.youtube.com/watch?v=zR0M7KjnJTE , Mustard did too https://m.youtube.com/watch?v=pW0eZRoQ86g

What you said is possible by feeding the output of speech-to-text tools into an LLM. You can prompt the LLM to make sense of what you're trying to achieve and create sets of actions. With a CLI it’s trivial, you can have your verbal command translated into working shell commands. With a GUI it’s slightly more complicated because the LLM agent needs to know what you see on the screen, etc.

That CLI bit I mentioned earlier is already possible. For instance, on macOS there’s an app called MacWhisper that can send dictation output to an OpenAI‑compatible endpoint.

Not exactly trackballs, but Magic Trackpad can be considered an alternative. Or roller mouse slim (crazy expensive)

Nice, I want one! Assuming it works great (Keychron products usually do)