Deepseek: The quiet giant leading China’s AI race

Post by waldrews » Wed, Jan 01, 2025, 1:20 AM UTC

jurli wrote: post #161

waldrews wrote: post #156

To this day, asking Deepseek "what model are you" typically gives the answer "I'm an AI language model called ChatGPT, created by OpenAI. Specifically, I'm based on the GPT-4 architecture, which is designed to understand and generate human-like text based on the input I receive. My training data includes a wide range of information up until October 2023, and I can assist with answering questions, generating text, and…

This is a common "gotcha" comment from people who don't understand LLMs very well. Occasionally if you ask Gemini it'll say this as well. It has everything to do with the fact that ChatGPT is the most talked about AI model rather than data being trained on it

Post by vkou » Wed, Jan 01, 2025, 1:24 AM UTC

lolinder wrote: post #281

vkou wrote: post #267

Earlier quoted context omitted.

I don't think the desire to see the conspiracy and not the boring, mundane explanation is a failure of communication, scientific thinking, or critical reasoning skills. I think it's simply the consequences of politically motivated reasoning. (At least, for people who have spent much time thinking about it.) > I've yet to see anyone effectively communicate why the intuitively more probable answer is the less probable…

> I just communicated why the market leak theory is both more intuitive, and more probable. No, you didn't, you stated that it was. The rest of your post makes sense as an explanation. Maybe lead with that next time instead of condescendingly telling people that they're politically motivated, stupid, or whatever else you meant to imply by calling it a conspiracy theory. COVID-19, at least in the US, has been an enorm…

Post by selimthegrim » Wed, Jan 01, 2025, 1:29 AM UTC

rekttrader wrote: post #273

vkou wrote: post #188

Earlier quoted context omitted.

> China creates a superbug via GOF research In 2024, this isn't fact, it's just baseless conspiracy. All evidence has ended up pointing to bush meat contamination.

Cleavage sites

Post by selimthegrim » Wed, Jan 01, 2025, 1:37 AM UTC

lolinder wrote: post #232

vkou wrote: post #230

Earlier quoted context omitted.

The jump between 'data was withheld' and 'there was a coverup of an incredibly improbable thing happening' compared to 'very probable thing that was also supported by data happening' is colossal. Absence of data isn't a free pass that lets you fill in whatever blanks you want, to fit whatever improbable theory you want. Especially when a plausible, probable, data supported alternative exists. At the moment, given wha…

Sorry, I edited my comment to ask this as the conclusion: > Why is a wet market a more convincing explanation than a lab whose explicit mission is studying coronaviruses, one which has been publicly called out for being uncooperative with the ensuing investigation? It's possible that the widespread belief in this explanation is a failure in science communication and there's a good reason for this, but it's not a fail…

Post by visarga » Wed, Jan 01, 2025, 1:45 AM UTC

jdietrich wrote: post #32

nsoonhui wrote: post #13

I find that the gushing around deepseek is fascinating to watch. To me there are a few structural and fundamental reasons why deepseek can never outperform other models by a wide margin. On par maybe--as we reach the diminishing returns with our investment in the models, but not win by a wide margin. 1. The US trade war with china which will place deepseek compute availability at disadvantages, eventually, if we ever…

I don't think it's necessarily about DeepSeek, but about the wider competitive picture. There are two tacit assumptions being made about LLMs - that having a SOTA model is a substantial competitive advantage, and that the demand for compute will continue to grow rapidly. DeepSeek's phenomenal success in reducing training and inference cost points to the possibility of a very different future. If it's the case that SO…