For the first time in history, artificial intelligence (AI) has achieved a remarkable feat by solving an incredibly complex mathematical problem that left human mathematicians stumped. The breakthrough was made by researchers from Google DeepMind and has the potential to revolutionize mathematics as we know it. The findings were published in the prestigious journal Nature.
The problem that AI managed to crack is known as the cap set problem. The task involves arranging a maximum number of points in such a way that they never form a straight line. It is a highly intricate problem that seemed insurmountable for humans, but AI rose to the challenge.
The tool responsible for this groundbreaking achievement is called FunSearch. Built upon the foundation of AlphaZero AI, FunSearch approaches mathematical problems like a game. It utilizes a language model called Codey, which is trained on computer code. Codey has the ability to eliminate incorrect answers and incorporate the correct ones into its model.
What sets this milestone apart is that AI not only found a solution to a known puzzle, but it also presented a completely new solution that was previously unknown to exist. This suggests that AI may have reached a new level of autonomy by seemingly thinking for itself.
DeepMind Vice President of Research, Pushmeet Kohli, expressed his surprise, stating, To be honest, we have hypotheses, but we don’t know exactly why it works. Kohli further admitted that at the beginning of the project, there were doubts about its success.
The significance of AI’s achievement is evident in Kohli’s statement: As far as we know, this is the first time that a real, new scientific discovery has been made using a large language model. Although further extensive testing is required to validate the solution to the cap set problem, the fact that AI has produced something it did not previously know is undeniably promising.
In the past, language models often generated nonsensical results, even when the training data contained correct answers. Therefore, confirming the accuracy of AI’s data would be a significant event.
This breakthrough in AI’s problem-solving abilities comes on the heels of its creation of a robot capable of walking on air within a remarkable time of 26 seconds. The rapid progress in AI technology is pointing towards a future where machines are not only capable of assisting humans in complex tasks but also paving the way for new scientific discoveries.
As the scientific community delves deeper into the mysteries of AI’s capabilities, the world eagerly awaits further groundbreaking achievements that can potentially reshape various fields of study.
