kkzz99

>Second its military complex is only good at dumb ammunitions and outdated equipment fit for human wave attacks.

This comment is not based on material facts. Russia has a lot of SotA military equipment. Including AD, EW, smart bombs, glide bombs, hypersonics and so on. These so called "human wave" attacks never happened and is a complete debunked meme.

Actually, a lot of economists think that Chinese GDP is severely under-counted. https://asiatimes.com/2024/06/whats-the-real-size-of-chinas-...

Pretty sure Reflection-70B was a complete scam. They did the ole bait and switch. The model that they uploaded was completely under-performing compared to their own benchmarks and the "secret API" was just a GPT-4 & Claude wrapper.

What are you even on about? This sentence has absolute zero value for the discussion at hand.

Guess you have not tried more advanced prompting techniques like CoT, Agents and RAG.

> It is evident to anyone that LLM can't perform any meaningful reasoning

Its not and its pretty evident to anyone that has actually used SotA LLMs for more than 5 minutes.

If the economical case justifies it you can use a cheap or lower end model to generate the meta information. Considering how cheap gpt-4o-mini is, seems pretty plausible to do that.

At my startup we also got pretty good results using 7B/8B models to generate meta information about chunks/parts of text.

>India can just buy the panels from China and be in nearly the same position.

Can it though? This will be impossible because of Indian domestic politics. You will also never get the same benefits by buying as in producing them yourself.

The difference is that China has the complete chain, from the raw materials, processing of said materials, manufacturing, integration, maintenance. They have the IP and know-how. India may have the land and the sun, but it lacks the hard parts.

Great article! I'm currently using doom + lsp-mode with ts-lsp and tailwindcss-lsp. It works great, but as the author said the performance could be better, especially tailwindcss-lsp which routinely freezes emacs on longer classnames. Does anyone know if I can use this setup with doom, or will there be any issues in regards to config conflicts? I don't want to leave the convenience of doom for other things.

Rufmord.

I'm stuck on the loading screen on firefox desktop.

You know that you can filter by material type on Shein, right? They have a lot of different materials that they use, including cotton.

Its not, its logistics, JIT-production and deep integration with small and flexible suppliers.

How is the state in regards to switching from integrated GPU to dedicated GPU? On my old laptop it was impossible to get it to work, it was either integrated or dedicated only, without an easy way to switch between the two. That was several years ago, so hopefully it is easier/better.

What are your tokens/s on that setup?

"It is continuously pre-trained on Yi with a high-quality corpus of 500B tokens and fine-tuned on 3M diverse fine-tuning samples.

Compared with Yi, Yi-1.5 delivers stronger performance in coding, math, reasoning, and instruction-following capability, while still maintaining excellent capabilities in language understanding, commonsense reasoning, and reading comprehension.

Yi-1.5 comes in 3 model sizes: 34B, 9B, and 6B. For model details and benchmarks, see Model Card."

Literally after that...

A token already has multiple meanings because words (and part-of-words) can have multiple meanings.

Probably special token that wasn't handled properly.

You can't add payment methods if you're in the EU.

Sure, if you compare a pickpocket with a rapist, the pickpocket is still bad.

China already has competing AI products that are used domestically.

Do you have any idea about the US's methods? You may want to pick up a history book to see how insidious the American methods are.

Good thing The Economist is so reliable in its reporting about China.

China is fine with the status quo, who isn't fine with that is the US.

Doing technology transfers and then crying about it is really bad manners.

Its like people completely forgot Prism. The caricature like understanding from Americans of China is mind boggling. I guess decades of McCarthyism takes its toll on the American mind.

They beat a 2-year old 3B LLM. I bet even GPT3.5 will beat their model, even without RAG.

It reads to me like the Google announcements. Platitudes to calm investors down so they can say "we are doing AI too see!". But I think Apple is really struggling, similar to Google.

Dropping sales in China doesn't help either...

How is this fascinating? One is a 175-1000+B parameter model the other is 3-70B parameter model.