Adam Rutherford - presenter of BBC Radio 4’s Inside Science - predicts that DeepMind's data crunching protein predictors will be the first AI to win a Nobel Prize.
Imagine reading sheet music, and knowing what a tune sounds like, but not having much of a clue how to play the instruments. That’s where we were in biology, until last month, when artificial intelligence company DeepMind swooped in and, by sheer computational force, crunched an otherwise intractable problem.
The sheet music here is the basics of genetics. Genes are made of DNA, and encode proteins, and all life is made of or by proteins. We can read the genetic code easily, and we can translate a gene into the basic protein. But proteins work in three dimensions, precisely folded up into clumps and blobs. These are the enzymes that digest, the structures that build bone and muscle and brain - that is the performance.
This crucial step is the one we struggled with - understanding how you craft a 2D protein into its 3D functional version. DeepMind’s AlphaFold programme was validated in November, beating all other known techniques for predicting protein structures hands down.
It’s stunning work. When it comes to designing drugs, or simply understanding how a protein works - or goes wrong in a disease - understanding the protein in three dimensions is essential. Bear in mind that because this is an AI solution, we still don’t truly understand the process, but at least we can predict how it will unfold.
Should DeepMind be nominated for the prize, or even win it, the philosophical argument as to whether the award should go to the AI or the original coders will no doubt run and run...
Artificial Intelligence to Help Combat Wildlife Poaching: Motion sensor uses AI to recognise when wildlife is running from poachers and can alert rangers. Any innovation that can help stop the slaughter of wildlife at the hands of poachers seeking, for example, ivory, has got to be good news. So, it's good to learn that scientists in the Netherlands are mixing motion sensors with machine learning to create a powerful tool that could achieve exactly that, simply by allowing them to recognize when wildlife is responding to a nearby threat. More...
Facial Recognition for Bear Conservation: The idea is that Artificial Intelligence could replace invasive collars and tags. Tracking specific animals usually involves conservationists attaching some form of device on the animals' bodies. So, when some software developers noticed the rapid growth of AI for human facial recognition, they wondered whether this could be used for conservation efforts too. More...