Reinventing Accelerated Computing for AI: Graphcore at SC20

In recent years, there have been major breakthroughs in AI but in order for AI to be truly useful, we have to be able to make it more powerful than a human being. We have to produce superhuman AI that can perform tasks that we are incapable of doing. Machine intelligence will need to be at least as smart as we are at understanding other human beings.

At the recent Supercomputing conference, Simon Knowles spoke about the many practical challenges of scaling up AI on computers and how we can solve them, not least the continued growth in size and complexity of machine learning models.