https://huggingface.co/LLM360/...

FPGA is the exact opposite of a general purpose device. The beauty of FPGA's is that they can be configured into specialized processors.

I guess you can say that they're general purpose in that sense, but when people talk about general purpose computing, they're generally referring to CPUs that don't need to be reconfigured / customized for different applications. FPGA's require first programming the FPGA for the task, then using it for that task.

I basically just use my fitbit as a silent alarm clock, I put it on when going to bed and take it off in the morning when I get up. I wish there was a non-tracking device that performed the same function, but haven't been able to find one. I thought about putting one together myself, but the form factor sucked, and those tiny vibrating motors are damn hard to solder by hand.

Trusts can be set up pretty easily and quickly. They're pretty boilerplate, and can technically be done without a lawyer.

There's some paperwork to file, and (sometimes) some additional things to do (e.g., file for an EIN for tax purposes). If you decide to pay a 3rd party to act as trustee that's an additional cost, but that's not needed.

https://legaltemplates.net/res...

Plex has it's issues, but it's still the easiest and most accessible media server out there, it has clients for Android, FireTV, AppleTV, Roku, iOS, plus a bunch of "smart" tv platforms. For the people who are sharing out their media with their friends or accessing their server remotely, it's basically plug and play, and offers internet wide access to the server without having to deal with managing dynamic dns and such.

There are annoyances, and I'm basically done with them due to how they've been integrating their external services into things. Attempting to sneak episodes from their streaming service into your local media so they get the ad revenue. Or how they make accessing a local server still require sign-in / internet access by default.

Jellyfin just isn't there yet for many people (but getting better every day, and is my hope for the future). Emby was getting better but started going down the Plex path and locking features.

"if you can't control it with HA (Home Assistant), don't buy it".

I'd say if you can't control it with HA natively, don't buy it. There's a few things I use via workarounds with HA, but I have no belief that those workarounds won't be removed via updates.

It's not so simple as an error bar, but here's how they deal with the uncertainty and error. It's a touch complicated, but with a background in statistics, I'm sure you can handle it. https://confluence.ecmwf.int/d...

I had an interesting experience with this a few years back. I've got a group of friends who game and hang out regularly, some of which have children. We have no problem drinking, cursing, or discussing sex or violence around the kids. But one year when talking about myths, we all got quiet and made sure the kids weren't around when Santa Claus was mentioned, it wasn't even in the context of the gifts or anything, it was just like something that was taboo to discuss around children.

You only need the emulator to generate the initial keys, after that there's no emulation required.

Actually, the proof needs to be made by the artists, who are the ones making the claim/assertion that the models are memorizing the images.

It's worth looking at FO76 as different type of thing from a normal Fallout game. Think of it like an immersive themepark in the Fallout universe. It's not really story driven, it's event/quest driven if that makes any sense. The themepark view explains things like the repeating events, daily quests, etc.

You're only a tax evader if you don't pay taxes. If you work for a US company and pay US taxes while abroad, you're fine. If you can show that you were out of the US for 330 days of the year and your income is from a non-US company, you don't pay US taxes, but you do need to pay taxes for the county you're working in/for.

Not certain of the experiment design, but it's likely that there was some additional scenarios that can actually tease out certain types of reasoning. E.g.,

• They are shown up to 10 images one by one.
• They get to pick 2 images, given a reward each time they pick (but stopping after the second pick).
• Image X shows up 95% of the time and gets them a reward of size 1, image Y shows up 5% of the time and gets them a reward of size 10.

Optimal play here is to pick Y if seen then wait until another Y or the 10th image. If not Y appears, pick images 9 and 10. That clearly shows a reasoning ability.

How is that different from Unity's new pricing? Which appears to be a cut from your revenue after a $200,000 is sales.

Or, directly from Unity's post on the matter (https://blog.unity.com/news/plan-pricing-and-packaging-updates)

The biggest issue I can is that for the free tier (Unity Personal), the fee is fixed at $0.20 per sale/install regardless of numbers after $200,000/yr in sales, while the minimum paid version (Unity Pro, ~$2,000/yr) starts cheaper and decreases in per unit price as volume increases (and only kicks in at $1,000,000 in sales).

It seems that the best thing is just to upgrade to the paid version once you start hitting $210,000/yr in sales, which really shouldn't be a big deal.

These LLMs couldn't copy a whole book word for word either, that' just inherently not how they work. They don't keep a full copy of a text, they really do just "learn" from it, and it gets merged with the rest of what it's "learned". He's an example interaction I had with an LLM to try to see if it really was memorizing things

I had to check the text of the play itself, but not one line of that dialog is correct, though some of the words are, and they certainly sound like they could have been from the play, but they weren't.