Ilya Sutskever warns that AI with reasoning capabilities will be less predictable

Ilya Sutskever, former chief scientist at OpenAI, stated that AI with reasoning capabilities will become less predictable. Speaking at the NeurIPS conference, he noted that the current method of pre-training AI is reaching its limits due to stagnant data growth. Sutskever highlighted that while computing power is increasing, the available data is not. He suggested that future AI could generate new data or evaluate multiple answers to improve accuracy. He emphasized that reasoning could lead to unpredictable outcomes. He also predicted the emergence of superintelligent machines capable of human-like reasoning. Sutskever pointed to examples like AlphaGo, which surprised experts with its unpredictable moves, indicating that advanced AI will differ significantly from current systems.