HuggingFace: "OpenAI’s o1 model showed that when LLMs are trained to do the same—by using more compute during inference—they get significantly better at solving reasoning tasks like mathematics, coding, and logic."
GENERATIVE AI
-
Deepseek goes viral: rapid adoption among mainstream users
By
–
my parents just started using chatgpt 2 weeks ago but asked me about deepseek yesterday. the virality is wild
-
Competitive AI Breakthrough Strategy Exposed to Market
By
–
I assume because they wanted to keep breakthrough close to their chest for competitive advantage? If that was the case the competition has clearly been blown wide open.
-
US Government Collaboration Secures Advanced AI Model Capabilities
By
–
…including a careful process for which frontier capabilities to include in released models…it is…important that we are working closely with the US government to best protect the most capable models from efforts by adversaries and competitors to take US technology.”
-
TinyZero: Affordable RL Finetuning Under $30
By
–
TinyZero reproduction of R1-Zero
"experience the Ahah moment yourself for < $30" Given a base model, the RL finetuning can be relatively very cheap and quite accessible. -
OpenAI’s Hypocrisy: Unauthorized Data Collection Practices Criticized
By
–
One of the few sane outlets… "It is…ironic that OpenAI, a company that has been obtaining large amounts of data from all of humankind…in an “unauthorized manner,”…is now complaining about the…practices by which it…built its company.
-
Facebook Chatbots Language Myth: Media Sensationalism Before ChatGPT
By
–
Do y'all remember the whole "Facebook created chatbots that speak their own language" garbage reporting from a few years ago pre-ChatGPT and all?
-
DeepSeek Meta research on AI language development raises safety concerns
By
–
You gotta be kidding me with this bullshit. "But DeepSeek and Meta’s recent research suggests that more AI capabilities (and efficiency savings) could be gained by going down a more dangerous path — where AIs develop their own alien language."
-
Journalists Amplify AI Hype Without Accountability Cycles
By
–
And the journalists amplifying this garbage will not be held accountable when the hype cycle is gone because the next cycle of journalists will do the same thing during the next hype cycle. I don't want to amplify the article so not posting the actual article.
-
Scaling Reasoning Models: From Thousands to Millions of GPUs
By
–
This is the interesting bit. As we move to the next phase of scaling reasoning models w RL, data and compute converge. Next breakthroughs require moving from 100’000’s of GPUs to millions of GPUs.