Anderson

Entrepreneur / AI Insight / Tea Lover

[Anderson @neoishim]

What is this…?

Deep Research has achieved 25.3% in humanity’s final test.
It’s been less than a week since O3-mini reached 13%.

Although it’s a feature combined with web browsing, all benchmarks are reaching saturation…!

[Anderson @neoishim]

7) At the same time, as AI models become increasingly open, it is becoming easier to leverage or even develop powerful AI independently. In short, DeepSeek’s emergence is being seen as a game-changing event that could significantly disrupt the entire AI ecosystem in the future.

[Anderson @neoishim]

6) Ultimately, this serves as an answer to the question: “Can China develop AI on par with the U.S. without access to high-end chips?” This is a shocking development not only for chip companies like NVIDIA but also for U.S. Big Tech firms that have invested heavily in massive AI

[Anderson @neoishim]

5) It also seems that DeepSeek has employed a knowledge distillation strategy, extracting knowledge from large models developed by other companies.

[Anderson @neoishim]

4) What’s particularly impressive about “R1-Zero” is that it has learned a thinking process that allows it to solve logical and mathematical problems on its own, without direct human instruction.

[Anderson @neoishim]

3) Additionally, the team has introduced new reasoning models called “R1” and “R1-Zero,” which appear to be as powerful as those from U.S. companies like OpenAI, surprising many people.

[Anderson @neoishim]

1) Deepseek 101
The key point is that DeepSeek, a Chinese company, has managed to develop a cutting-edge AI model using only the H800 chips, which have certain limitations, rather than the high-performance H100 chips from the U.S.

[Anderson @neoishim]

2) Typically, training such models requires significant resources and funding, but DeepSeek leveraged architectures like MoE and MLA to drastically reduce computational and memory usage. As a result, they were able to run large-scale models at a significantly lower cost.

[Anderson @neoishim]

I think this is correct thing about deepseek
https://t.co/kXulaxQ3wJ

[Anderson @neoishim]

Is this the ripple effect caused by DeepSeek?
Copilot is now offering the O1 reasoning model for free.

[#Microsoft CEO]
“Today, we have made Think Deeper available for free to all Copilot users.
Now, everyone can access OpenAI’s world-class O1 reasoning model for free on Copilot

[Anderson @neoishim]

#BreakingNews: #ChatGPT-4o now supports “Think” (reasoning) mode.

Could this update have been influenced by #DeepSeek?

However, it’s a bit disappointing that it doesn’t engage in long deliberations like DeepSeek does.

[Anderson @neoishim]

The reason it is currently a hot topic among the public is not due to its technological superiority but rather…

[Anderson @neoishim]

There is a lot of controversy surrounding #DeepSeek (R1).

Anyone who has tested it multiple times would know that, despite the numerous debates, the important point is that DeepSeek’s performance falls short not only of the O1 Pro but even of the O1.

[Anderson @neoishim]

AI robot is working well

[Anderson @neoishim]

Deep Seek / Chatgpt #DeepSeek

[Anderson @neoishim]

I am on again @cursor_ai Cursor is back

[Anderson @neoishim]

@MattPRD @cursor_ai I am on again

[Anderson @neoishim]

Cursor makes me off 👍 Thank you @cursor_ai

[Anderson @neoishim]

@JayLovesPotato 😀

[Anderson @neoishim]

@osmosis Osmosis is newly updated!

[Osmosis 🧪 @osmosis]

Yeeeeehaaawww! The frontier is open for the exploring! A rough and wild place. Stay safe.

https://t.co/AZL3dPYeqq