😲 Former Chief Scientist of OpenAI, Ilya, has made a bold statement! At the NeurIPS 2024 conference, he bluntly said that AI pre-training needs to 'cut off the supply,' as data is running out ~ just like fossil fuels are running low, the readily available human-generated content on the internet has reached its limit 😰. How will models be handled in the future? Training methods need to be changed, moving towards a new stage.

👀 Future models will be super cool, no longer sticking to pattern matching; autonomy and reasoning abilities will come into play! They will be able to solve problems like we do, and the stronger the reasoning, the more 'playful' and elusive they become. It was also pointed out that intelligent agents and synthetic data are the way forward.

📣 The speaker's presentation was very informative, reviewing deep learning and addressing questions without hesitation ~ like whether models can correct their own hallucinations, the answer is promising; can LLMs generalize? Compared to us humans, they still have a long way to go, but they are making progress! A storm of AI transformation is approaching, and we are looking forward to it ~💥