John Shawe-Taylor – Artificial Intelligence

Can computers be programmed to show human levels of intelligence?

Thus far, despite some successes, like the DeepBlue chess-computer, AI has not been achieved. The sheer number of computations required to apply logic to complex problems is unfeasible. Machine learning on the other hand, is perhaps a more promising approach. It trains a system up to use probabilistic reasoning that can work out patterns in data. This approach may be a way forward for achieving artificial intelligence and has indeed shown success in playing games such as Go, which allow for exponentially more combinations than chess, rendering a brute-force logic approach unfeasible. As such, this approach will be taken in the HBCP to learn to populate the ontology.