“Sakana AIでは、国内の貴重な、しかし限られた資源を踏まえ、非常に少ない資源で優れたAI技術を開発することに努めています。現在、日本のためにAIを開発し、社会実装することに情熱を持つ仲間として、アプライドエンジニアとプロジェクトマネージャーを募集しています。🚀
https://x.com/SakanaAILabs/status/1885147467332936190
“Global Ban Wave on DeepSeek. South Korea just joined seven other countries in banning DeepSeek. Other countries enforcing bans include Australia, India, Taiwan, Italy, the US, Belgium, France, and Ireland. → Australia banned DeepSeek across government systems, citing national
https://x.com/rohanpaul_ai/status/1887469512087552491
“1/ The Bank of China announced 1 trillion yuan ($140B) in investments for the AI supply chain on Tuesday, Jan 28, 2025 in response to Stargate.
https://x.com/alexandr_wang/status/1885377285676761510
“Just read this exclusive with Liang Wenfeng, founder of DeepSeek. It’s by an independent Chinese journalist back in 2024 but is a great read. Standout points: -The team is mostly made up of new grads and undergraduate interns -DeepSeek (V2) started a price war in China’s AI
https://x.com/rowancheung/status/1885734811613958326
“2/ in the first half of 2024, the Chinese government issued 81 contracts to LLM companies to integrate it into their military and government, with Zhipu emerging as the early winner
https://x.com/alexandr_wang/status/1885377292001812810
Chinese and Iranian Hackers Are Using U.S. AI Products to Bolster Cyberattacks
https://www.msn.com/en-us/technology/artificial-intelligence/chinese-and-iranian-hackers-are-using-u-s-ai-products-to-bolster-cyberattacks/ar-AA1y37M3
“3/ In 2023, the Chinese Government announced data labeling subsidies including: – tax incentives – government procurement of data – data vouchers for cost reduction – investment promotion – talent development programs
https://x.com/alexandr_wang/status/1885377298997952648
SoftBank-backed billionaire to invest $230M in Indian AI startup Krutrim | TechCrunch
SoftBank-backed billionaire to invest $230M in Indian AI startup Krutrim
Who is Liang Wenfeng? DeepSeek founder comes from AI investing | TechCrunch
Who is Liang Wenfeng? DeepSeek founder comes from AI investing
Hugging Face researchers are trying to build a more open version of DeepSeek’s AI ‘reasoning’ model | TechCrunch
Hugging Face researchers are trying to build a more open version of DeepSeek’s AI ‘reasoning’ model
“Surprised we haven’t seen more about Deepseek r1-zero (no one seems to host it?) Unlike r1, which was trained to “think” in a readable, kinda charming way, r1-zero is the self-trained reasoner that had the *aha moment* about math & produces “thoughts” that are not human readable
https://x.com/emollick/status/1885150199279972717
“Free Deepseek R1 on OpenRouter 🤔
https://x.com/rohanpaul_ai/status/1885675708824879391
“🎧 Wow, that finale in Planet Money’s latest on DeepSeek! Must-listen episode feat. @lvwerra breaking down AI’s democratization in the most fun, accessible way possible 🤖✨
https://x.com/fdaudens/status/1885858178396463128
“deepseek R1 at 44 Output Tokens per Second on @replicate is super fast developers can get started easily with ai-gradio pip install –upgrade “ai-gradio[replicate]” import gradio as gr import ai_gradio gr.load( name=’replicate:deepseek-ai/deepseek-r1′, src=ai_gradio.registry,
https://x.com/_akhaliq/status/1885385810419044623
“Beautiful piece by @huggingface on Open-R1’s effort to replicate the DeepSeek-R1 pipeline and dataset. Some of the Key Takeaways – Some of the Evaluation results closely match DeepSeek’s benchmarks, but handling its massive 6,000+ token responses remains a challenge. →
https://x.com/rohanpaul_ai/status/1886180331675422728
DeepSeek temporarily suspends API service top-ups | Reuters
https://www.reuters.com/technology/artificial-intelligence/deepseek-temporarily-suspends-api-service-top-ups-2025-02-06/
“agents will actually work once tool calling gets built into reasoning models deepseek r1 + web search is a v0 of this. its also why the reasoning traces are *very* important (and great to read)” / X
https://x.com/SullyOmarr/status/1884351883307085972
DeepSeek R1 struggles with its identity – and more • The Register
https://www.theregister.com/2025/01/27/deepseek_r1_identity/
“We saw this with DeepSeek v3 when it was fine-tuned on reasoning data. Looks like reasoning data really hurts instruction following” / X
https://x.com/nrehiew_/status/1885392655489663271
“Introducing DeepSeek R1 Trend Finder 🔎 It gets posts from key influencers using @firecrawl_dev and the X API, finds any trends with R1, then pings you in slack. Check it out:
https://x.com/ericciarla/status/1886456419240829193
“open Sourced the r1 chat template! – easy integration with @aisdk! – works with deepseek r1 providers like @GroqInc! repo here →
https://x.com/zaidmukaddam/status/1885041197711781898
“No reasoning model consistently solves this puzzle, but DeepSeek’s thought here was insane (it also got it wrong): “A young boy who has been in a car accident is rushed to the emergency room. Upon seeing him, the surgeon says, “I can operate on this boy!” How is this possible?”
https://x.com/emollick/status/1886271713970704506
“The lack of good benchmarks is pretty glaring with DeepSeek and other reasoning models. These models are clearly good at some things & are extremely mediocre at others (DeepSeek is not good at multi-turn conversation, for example). MMLU and standard scores don’t help with that.” / X
https://x.com/emollick/status/1885159401150984370
DeepSeek might have a trademark problem in the US | TechCrunch
Taiwan bans government departments from using DeepSeek AI | Reuters
https://www.reuters.com/technology/taiwan-bans-government-departments-using-deepseek-ai-2025-02-03/
One very interesting observation about DeepSeek-R1: few-shot prompting actually degrades its performance. I’m not 100% certain of the cause, but the lack of good few-shot learning performance is likely due to the fact that R1 is trained to adhere to a strict format. During RL, we apply a formatting loss that rewards the model for outputting a long CoT reasoning trace followed by a final answer / summary. Few-shot examples don’t naturally fit into this reasoning format.
https://x.com/cwolferesearch/status/1886810699726275041
“Expect the next UX battle in AI to be over what “thinking tokens” to show DeepSeek found that producing less pleasant (mixed languages, etc) thoughts might get better outcomes, but people didn’t like them. They trained r1 to show nicer thoughts, which makes DeepSeek nice to use
https://x.com/emollick/status/1885747742472896897
“People don’t talk enough about a giant DeepSeek achievement over most US models – it actually has a reasonable name.” / X
https://x.com/emollick/status/1884847466857873837
How Indigenous engineers are using AI to preserve their culture
https://www.nbcnews.com/tech/innovation/indigenous-engineers-are-using-ai-preserve-culture-rcna176012
India now OpenAI’s second largest market, Altman says | Reuters
https://www.reuters.com/technology/openais-altman-meets-with-india-it-minister-discuss-countrys-ai-plans-2025-02-05/
“first full 8-rack GB200 NVL72 now running in azure for openai—thank you @satyanadella and jensen!” / X
https://x.com/sama/status/1885191346916356371
“Hallucination is still a big problem: (HN commentator on OpenAI’s new deep research tool)
https://x.com/OfirPress/status/1886367776576552977
“You do need to actually prompt OpenAI deep research to get the results you want. Saying “focus on high quality papers and RCTs” or “address problematic definitions and conflicts” or “present the results for a graduate-level discussion of this issue” all impact the process a lot.” / X
https://x.com/emollick/status/1886470694369890527
“when we launched the openai pro plan in december i thought it was only a good deal for extreme power users. now with the addition of 100 deep research queries per month, unlimited sora, unlimited chatgpt, o1 pro and more stuff to come, i think it is a great deal for many users!” / X
https://x.com/sama/status/1886424331242348743
“Look, your experience may vary, but asking OpenAI’s Deep Research about topics I am writing papers on has been incredibly fruitful. It is excellent at identifying promising threads & work in other fields, and does great work synthesizing theories & major trends in the literature.” / X
https://x.com/emollick/status/1887191008250609876
Sam Altman Says OpenAI Is Going to Deliver a Beatdown on DeepSeek
https://futurism.com/sam-altman-openai-better-models-deepseek
“OpenAI is planning to develop a device dedicated to generative AI to replace smartphones.
https://x.com/rohanpaul_ai/status/1886392041808294115
“Economist @joshgans uses o1-pro to generate a (minor, fun) paper in an hour based on an idea of his, and it gets published in an appropriate peer reviewed journal, with adequate disclosure. He ends with the same sentiment I am increasingly seeing from fellow academics: what now?
https://x.com/emollick/status/1887220811972681771
OpenAI inks deal with Los Alamos lab to cram o1 into Venado • The Register
https://www.theregister.com/2025/01/30/openai_los_alamos_national_lab/
“Exactly this. The breakthrough of OpenAI’s Deep Learning is not that it is more accurate (though it is) than the fact that it is good at coming up with useful analysis and viewpoints – a very different (and in many ways higher order) form of research to prompt your own thinking” / X
https://x.com/emollick/status/1886851713111310821
“Altman says the jump from GPT-4 to GPT-5 will be just as major as the shift from GPT-3 to GPT-4. OpenAI is also moving toward integrating the GPT and o model families into a single, more capable model.
https://x.com/rohanpaul_ai/status/1886598420712063483
“ChatGPT search is now available to everyone on
https://x.com/OpenAI/status/1887224584539414983
“I think OpenAI Deep Research has a confusing UX issue (OpenAI folks who follow me, am I right?). Using the Deep Research toggle doesn’t work like other toggles, which tell the system to use tools with your selected model. Instead it apparently always invokes o3. Is that true?” / X
https://x.com/emollick/status/1886507863365050559
David Sacks claims there’s ‘substantial evidence’ that DeepSeek used OpenAI’s models to train its own | TechCrunch
“Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs
https://x.com/_akhaliq/status/1885195541161574537
“Stanford presents: s1: Simple test-time scaling – Seeks the simplest approach to achieve test-time scaling and strong reasoning performance – Exceeds o1-preview on competition math questions by up to 27% (MATH and AIME24) – Model, data, and code are open-source
https://x.com/arankomatsuzaki/status/1886250066324910089
“Agree, there is a lot of easy potential paths to vastly improve inference-time scaling (the approach powering o1 & DeepSeek r1), which is meaningful because inference-time scaling already represents a breakthrough in AI that is just starting to be exploited.” / X
https://x.com/emollick/status/1885107806635733265
“I find it odd to hear comments that the results of OpenAI Deep Research are too long, I think that misses the point of an analyst report You want to see a discussion of the evidence & argumentation. Reports are less for answers than to inform your thought in ambiguous situations” / X
https://x.com/emollick/status/1886786341041500607
“In a large representative sample of humans compared to GPT-4: “the creative ideas produced by AI chatbots are rated more creative [by humans ]than those created by humans… Augmenting humans with AI improves human creativity, albeit not as much as ideas created by ChatGPT alone”
https://x.com/emollick/status/1885874814667133120
“This result keeps showing up in papers. GPT-4 does a good job on patient care tasks, better than the mean doctor, but doctor-AI teams do not outperform AI alone. Figuring out how to gain more when human & AI work together needs to be a bigger priority in integrating AI into work” / X
https://x.com/emollick/status/1887159022559658409
Introducing deep research | OpenAI
https://openai.com/index/introducing-deep-research/
India now OpenAI’s second largest market, Altman says | Reuters
https://www.reuters.com/technology/openais-altman-meets-with-india-it-minister-discuss-countrys-ai-plans-2025-02-05/
