Arriving a little early I was able to tour the first floor library, occupying several rather grand rooms. I was pleased to find a copy of Murray's dictionary in its original bindings, unlike my copy, similar in other ways, but in a red library binding.
Into the lecture theatre where Wolpert was introduced by the very newly appointed director, Shaun Fitzgerald, an academic engineer from Cambridge. Brisk and efficient sounding chap, entirely suitable to introduce Wolpert, who gave a very polished talk, delivered rather fast without any other prop than some well prepared Powerpoints. But with more than a hint of the arrogance one sometimes observes in successful men. Good with questions.
Bayes' theorem is important in his world, the theorem which is used to update one's expectations in the light of experience and the chap who is celebrated at references 1 and 2. Also important to the brain which uses it to decide what is going on in the outside world, with one example being the well known Neckar cube, more usually used to demonstrate how the brain flip-flops between its various possible interpretations. While Wolpert's point was that all these interpretations assumed that the thing was a cube, when actually there was an infinity of more or less bizarre objects which would project onto the same plane image. The brain, in the absence of any evidence of the bizarre, effortlessly opts for the easy option, the cube. Which might sometimes be wrong, but it much more usually right.
Early on we had the example of animals called sea squirts which had brains when they were very young but lost them once they became adults and settled down. More or less ate their own brains so that the energy therein could be redeployed to more useful ends. The point of the example being that brains were expensive and they need to be useful to be retained. Which led on to what I thought was slightly sour joke about professors who settled down to a quiet, long life once they had got tenure, which might have something to do with why Walpert is off to fill a vacancy in New York. I associate to a snippet from school in which the son of a tenured professor of history explained to us that his father had plenty of bright young things nipping at his heels, but that he had no intention of moving and if they wanted to be professors too, they would jolly well have to go up north.
Our own brains spend a lot of their time issuing commands to the body, including the commands needed to get us about, including monitoring the results of those commands. Tuning them on the fly as they sense deviations from predictions. With brains making lots of predictions, many of them making use of Bayes. The signals from the environment, the senses and running around in the brains are all full of noise, so the brains have to get good at dealing with noise and error. Good enough, as it happens, that the brain tolerates rather more error when, for example, we are walking along the road than when we are trying to put a fork into our mouth - with accidents in the latter case apt to be rather unpleasant.
Contrariwise, noise is sometimes important. Noise might, for example, occasionally push stimuli which are too weak to fire something up into firing that something up. Perhaps resulting in the occasional checking out of a possible threat. Not a good use of resources to check every tune, but best to check once in a while, just to be on the safe side - and a bit of noise is a good way of doing this.
Sometimes the brain will be getting conflicting signals, perhaps from the different senses, conflicts which mean that sometimes the brain gets it wrong. Wolpert paraded a number of well known conflicts and errors of this sort, conflicts and errors which, fortunately for us, might be easy to fabricate in the laboratory but are rare in the real world. One being the popular ba ba da da experiment, also known as the McGurk effect. So popular that Bing turns up the very video clip offered by Wolpert. See reference 5.
Wolpert having told us something of the complicated computations needed to support movement, suggested that robots, while very clever at playing brain games like chess, are not very clever when it comes to games involving manual dexterity. He pointed to the cup stacking competitions which are big in North America and the Far East, videos of which are to be found at reference 4. But I was not so sure about this one. I thought that the robots used, for example, in the car manufacturing industry, might be quite good at cup stacking if you gave them a try.
He also told us that it was usually unhelpful to try and help a waiter bringing you a tray heavily loaded with drinks. If he removed the drinks himself, his brain had time to compute the necessary adjustments to the muscles in the arm holding up the tray. But if you removed them for him, he did not know enough about what you were going to do, his prediction window was much smaller and he was much more likely to drop the tray.
A few oddments. We were told that neuroscience is a big industry and that the big US conference might draw 30,000 neuroscientists and hangers-on.
That if your wired the ears into the sight part of the brain of a very young animal, you often wound up with working hearing. Or wired the eyes into the hearing part of the brain. This being an example of the brain's plasticity. Yet to be confirmed by Bing or Google, but a subject to which I shall be returning.
That one could make lots of inferences from a small amount of evidence of what Wolpert called biological movement. So you can tell a surprising amount about what people are doing and feeling, even when all you can see of them is a few light attached to their joints, to things like knees and shoulders. See reference 3 for an example. All good stuff from the point of view of survival in the jungle.
Plenty of evidence from the animal world that they are into prediction too.
Another excellent talk. The standard set by the Institution is really very high, at least on our sample over the past few years.
Scored another offer of a seat on the tube back to Vauxhall. Seems to be happening all the time now.
Back home, BH observed that Wolpert was clearly a Dutch name, so curious I asked Wikipedia, to find that he was the son of an equally eminent Wolpert from South Africa. So given that he was a white man, quite probably of Dutch origin.
Reference 1: http://psmv2.blogspot.co.uk/2015/09/bayes-2.html.
Reference 2: https://en.wikipedia.org/wiki/Bayes%27_theorem. Including a useful reminder of how one can read too much into people failing tests for illegal substances.
Reference 3: https://youtu.be/TrIwxKBMDO0. Bear with the irritating music and you get there. And searches for ‘point light animations human kinematics’ and ‘skeleton animation’ will turn up related material.
Reference 4: https://www.thewssa.com/tournaments/. Look for entertainment in the videos tab.
Reference 5: https://www.youtube.com/watch?v=PX9FYxadPoQ.
Group search key: dwa.
