The exponential power of scientific innovation has been on show in recent days. In a series of demonstration videos, Google DeepMind unveiled that it had trained a robot to fold an origami fox, organise a desk and slam dunk a small basketball through a hoop. Last month, Microsoft?showed off a robot conducting basic kitchen tasks. Both companies are capitalising on advances in artificial intelligence.

By using the reasoning capabilities of large language models, they have developed robotics software that raise the ability of automatons to adapt to complex environments. Harnessing the technology will take time, but for economies facing shrinking workforces, and rising social and industrial demands, the opportunities are tantalising.

Tech groups, start-ups and investors have been clamouring to build an AI “brain” that can independently operate robots for applications in healthcare, manufacturing and even household chores. After decades of tedious progress that involved manually coding each movement a robot made, AI training is turning robots into fast learners. They can now absorb reams of text, images and videos from the internet to develop their understanding of the physical world. The latest software developments suggest the technology is developing to the point where robots can adjust to new situations, respond to verbal instructions and be dexterous enough to manipulate objects.