Scroll for more
blog Oct 7, 2016

What happens when you put expert human brains in a room to discuss machine brains?

Author: Jackie Hunter

These are truly exciting times for artificial intelligence (“AI”) and bioscience. This month, together with The Crick and Turing Institutes and the Wellcome Trust, we co-hosted the first of what we anticipate will be an annual symposium on ‘Artificial Intelligence in Bioscience’.

We put some of the world’s leading experts in AI and bioscience in a room together at The Royal Society to discuss everything AI and Bio – from if AI is ready to change the world…to the crucial questions you need to ask for AI to succeed and… analysing if AI can transform drug discovery to… the ethics debate around AI, machine learning and big data.

Here’s what we learnt:

AI is data hungry but it needs a balanced diet

AI to promote scientific discovery needs a balanced data diet.  It needs to be balanced in terms of using training data and large established data sets and it needs to be balanced in analysing not only positive scientific results but also negative ones where experiments and clinical trials failed.  What you feed AI ultimately leads to its algorithmic refinement, learning accuracy and success– so computers need a good diet too!     

Openness and collaboration is crucial

Given the exponential technological advancement of the last five years, it has never been easier for industry players to open up and start to share their learnings and collaborate publically, privately and across academia.  AI is just one technology which may be able to unlock a whole host of ‘usable’ scientific knowledge for the bioscience industry.  As mentioned above this is especially useful for failed drug discovery attempts which could hold valuable clues for the future of medical development.

AI needs to be rooted in the possible

AI might well be a silver bullet, but it needs to be fired in the right way with precision.  There is a temptation to set unrealistic expectations of what AI technology can achieve today in bioscience.  It is important that any AI project is rooted in the possible and doesn’t set unrealistic expectations of the technology.

User experience counts

The quality of the user experience is fundamental and therefore a good understanding of who will be the actual user of any AI based product is important.  Early and regular user feedback is crucial to AI projects succeeding as is considering the precise nature, function, management and ease of use of human to machine interactions.

Perhaps for me one of the most important outcomes of the symposium was a general agreement we need greater interaction between the biomedical and AI communities.  To this end we are exploring the establishment of a special interest group based in the Kings Cross Knowledge Quarter to drive collaboration, shared learning and progress. Watch this space!