lithiumii

Time to finish up my $5 at Deepseek.

except now we don't need to spend that $5

GLM has subscription plans too.

What's so mysterious? Isn't it from Tencent?

That's actually a reason for me to try it again. My past attempts to use LLM for OpenScad has greatly improved my own OpenScad skills.

It's cute. My main complaint is I was expecting some real next-gen bubble tea.

No. I had a Samsung TV which connects to the internet via the HDMI cable to my Nvidia Shield.

I love what he is doing but really hope the voting interface was better. Also I wonder what the results would be if there are AI-assisted stories, but maybe real authors would hate to do that.

You are not selling or distributing copies of your brain.

Well technically even DeepSeek is not as OSS as OLMo or Open Euro, because they didn't open the data.

Someone should make this video with AI.

Many Chinese tech giants already had A100 and maybe some H100 before the sanction. After the first wave of sanction (bans A100 and H100), NVIDIA released A800 and H800, which are nerfed versions of A100 and H100.

Then there was a second round of sanction that bans H800, A800, and all the way to much weaker cards like A6000 and 4090. So NVIDIA released H20 for China. H20 is an especially interesting card because it has weaker compute but larger vram (96 GB instead of the typical 80 GB for H100).

And of course they could have smuggled some more H100s.

As long as they make enough money from Pokemon Go to sustain Ingress, I OK with that.

The more they sell, the less likely China & others are going to make a real alternative.

That could be a completely different problem. In China many people run PCDN (p2p CDN) for profit. The ISPs detect (and ban) such PCDN nodes by checking your uploaded / downloaded ratio. To increase this ratio thus avoid being detected, these people download popular torrents again and again without uploading at all.

You can also build one with LEGO, which is like the OG 3D-printing.

Don't worry, in China they need to worry about a different set of "ethics". It's not just your LLM should never speak against Xi. They probably also need to include a big propaganda corpus so that the models are aligned with CCP's value, which is very likely to be bad for reasoning capabilities.

My new VPS got an SSH attempt in 5 minutes after I purchased it. I'm now in the progress of running a similar honeypot experiment.

AI needs data, so whatever provides data to Google (e.g., YouTube) is less likely to shutdown.

A Chinese map (https://www.amap.com/) has it, but only for some cities in China.

Looks like it's 2 books, Geologic Atlas of the Lunar Globe and Map Quadrangles of the Geologic Atlas of the Moon.

If it's an ISP, maybe it's their crappy modems now part of a botnet.

But if you accidentally exposed your calibre-web online, they may be able to DMCA you?

Very interesting Claude ethics here:

Have no problem in creating super robots to enslave human. Refuse to add Khorne or Tzeentch to the world.

I guess now we know how to deal with a machine uprising.

Sites being blocked is different from internet being slow. Theoretically it's possible for GFW to disappear (very unlikely) or malfunction (more likely) and access to Google or BBC is suddenly much faster. Practically you can use a proper VPN.

I don't disagree with your opinion on which country is better to live in.

Three 4060 Ti 16GB (there are single slot models) is around $1500. I think is possible to get a consumer system that's cheaper than a 48GB Mac.

I once made a simple serial monitor GUI with this. It's really easy to use, and has better defaults than vanilla tkinter.

However, as soon as I discovered VS Code's serial monitor plugin I abandoned my project.

Like the new Leica M11-P with its "content credential"?

If you switch to the test-fp8 branch of AUTOMATIC1111 web UI, you can do 512 * 1024 on 4G GPU with SDXL.

How about stay anonymous and just violate all the copyright laws? There's already pirate bay, libgen, sci-hub, zlibrary, etc., surely it's possible for there to be an opensource & pirate LLM model.