Google today announced that its DeepMind AI software had made a significant breakthrough in solving complex geometry problems. Machine learning expert from the University of Sydney’s School of Computer Science, Dr Clément Canonne, says that while the results are impressive, there are caveats.
Organisation/s: The University of Sydney
From: The University of Sydney
“While the International Mathematical Olympiad results are an impressive feat for the Google Deepmind team, there’s more than meets the eye,” said Dr Canonne.
“Only a small subset of problems were actually solved. The system is not able to ‘read’ the questions – it is provided a precisely worded translation of the problems. That step is done by a human, and then a human also needs to translate its results.
“This whole process also requires enormous computing power and is slow – it’s not something that can be done on your desktop computer.
“Basically, there’s a huge amount of non-trivial work that is done by humans to achieve this. It’s not an apple-to-apple comparison with existing systems, or even to Olympiad participants. For now, maths whizzes can breathe easy.”
Dr Canonne researches what machines are capable of learning and how efficiently. He is available for interview and speaks both English and French.