"Any headline that ends in a question mark can be answered by the word no."
post
Kinda. Even after many months, deepseek R1 still dominating local LLM benchmarks. Next to Qwen3.
I think between however actually good it is and US companies having to answer to class action lawsuits for massive copyright infringement that China and Chinese models dobt have to worry about, they just might come out on top of this pile.
Im not convinced this isnt ultimately a non-starter since it creates problems with liabillity and holding any decision maker or process to account and is prone to embarassing failures like not being able to figure out how to play a pokemon game. Reminds me of when I was a kid and I couldn't get the first star in the Bob-Omb plateau level until my brain finally could. It took a long time and getting past a ton of childish superstition I had developed along the way to psychically deal with the failure
An article brought to you by the leading authority on cutting-edge computer science research: BBC.
It's not the BBC's job to be an authority. Their job is to report what the (relevant) authorities are saying:
DeepSeek challenged certain key assumptions about AI that had been championed by American executives like Sam Altman, CEO of ChatGPT-maker OpenAI.
"We were on a path where bigger was considered better," according to Sid Sheth, CEO of AI chip startup d-Matrix.
Perhaps maxing out on data centres, servers, chips, and the electricity to run it all wasn't the way forward after all.
Despite DeepSeek ostensibly not having access to the most powerful tech available at the time, Sheth told the BBC that it showed that "with smarter engineering, you actually can build a capable model".
That said, seems suspect that an AI startup CEO is getting this much airtime. I would have preferred an industry analyst or an AI researcher.
AI has a pretty clear definition. What you mean to say is that it isn't intelligent but that doesn't make it not AI. Like how a shooting star isn't actually a star.
I dunno, I've asked it to do increasingly more complex tasks and it's obliged. Stuff like take these 10 pages of handwritten tables and ocr them and make a .csv out of them. It's inventing a whole process to do that. Same with some examples of it programming entire games. You need some procedural thinking to do those things.
People who claim it's just a very convincing auto-complete are uneducated and easily manipulated by social media groupthink.
all 4 comments