DeepSeek, new AI model from China, it's the new rival to the US AIs. It's a game changer: you can sign up today and it's open sourced! It's free.

[Walter Panov]

Berkley AI research team claims to reproduce DeepSeek core technologies for $30


An AI research team from the University of California, Berkeley, led by Ph.D. candidate Jiayi Pan, claims to have reproduced DeepSeek R1-Zero’s core technologies for just $30, showing how advanced models could be implemented affordably. According to Jiayi Pan on Nitter, their team reproduced DeepSeek R1-Zero in the Countdown game, and the small language model, with its 3 billion parameters, developed self-verification and search abilities through reinforcement learning.

DeepSeek R1's cost advantage seems real. Not looking good for OpenAI.

Meh, I tested it. It's trash. The Open AI mini version is smarter for simple logic problems. For my simple problems, Deepsek got it wrong every single time, GPT 4 mini about half the time, and Google Gemini got it right every time. You can download it and test it yourself on your computer with Ollama.

 

[WorldEX]

Meh, I tested it. It's trash. The Open AI mini version is smarter for simple logic problems. For my simple problems, Deepsek got it wrong every single time, GPT 4 mini about half the time, and Google Gemini got it right every time. You can download it and test it yourself on your computer with Ollama.

What is your test method?

 

[Walter Panov]

What is your test method?

Using them for RAG. Testing how well they could find a specific piece of information in a specific table. Deepseek failed 100% of the time. Gemini 1.5-flash (also a small model) got it right 100% of the time. It was a simple logical prompt "Go to table X and find the value of Y".

 

[WorldEX]

Using them for RAG. Testing how well they could find a specific piece of information in a specific table. Deepseek failed 100% of the time. Gemini 1.5-flash (also a small model) got it right 100% of the time. It was a simple logical prompt "Go to table X and find the value of Y".

For knowledge to others
RAG
Retrieval-augmented generation is a technique for enhancing the accuracy and reliability of generative AI models with information from specific and relevant data sources.
https://blogs.nvidia.com/blog/what-is-retrieval-augmented-generation/