Machine Learning
Google unveils Gemini 3 AI model and Antigravity IDE
Rapportert av AI Bilde generert av AI
Google has released Gemini 3 Pro, its latest flagship AI model, emphasizing improved reasoning, visual outputs, and coding capabilities. The company also introduced Antigravity, an AI-first integrated development environment. Both are available in limited preview starting today.
A new study shows that transfer learning can reduce the computational costs of searching for physics beyond the standard cosmological model by more than a factor of ten. The approach trains AI first on simpler simulations before moving to complex ones. However, it can lead to negative transfer that hinders detection of genuinely novel effects.
Rapportert av AI
OpenAI has explained why its ChatGPT models developed an unusual affinity for mentioning goblins and gremlins. The issue stemmed from reinforcement learning in the 'nerdy' personality feature. The company addressed it by retiring the personality and adjusting training processes.
Thinking Machines Lab, a startup founded by former OpenAI researchers, has launched its inaugural product, Fine-Tune, aimed at simplifying the customization of large language models. The platform promises to make fine-tuning accessible to developers without extensive resources. This release marks a significant step for the company in the competitive AI tools market.
Rapportert av AI
Chinese AI firm DeepSeek is experimenting with sparse attention mechanisms to significantly lower the processing costs of large language models. The approach focuses computations on key parts of input data, potentially halving resource demands. This development could make advanced AI more accessible amid rising energy concerns.