Over the past cen­tu­ry, we’ve been to the moon, we’ve split the atom, we’ve sequenced the human genome, but were still only at the very begin­ning of our under­stand­ing of the human brain. This is one of the great chal­lenges that we face. If we can under­stand the brain, we can devel­op bet­ter treat­ments for brain dis­or­ders, we can design bet­ter robots, bet­ter com­put­ers, and ulti­mate­ly we can bet­ter under­stand ourselves.

Alan Turing was one of the first to sug­gest that the brain com­putes. In oth­er words, that it imple­ments algo­rithms to solve prob­lems, like com­put­ers. And Turing also was one of the first to design and help build a com­put­er, the Automatic Computing Engine from 1950, and it was a process that he actu­al­ly called build­ing a brain. Since that time, there’s been tremen­dous advances in com­put­ing hard­ware, in algo­rithms. And there’s been a lot of recent excite­ment about deep learn­ing net­works, which are inspired by the hier­ar­chi­cal pro­cess­ing and the visu­al sys­tem of the brain, and which can be trained to rec­og­nize pat­terns like human faces, or cats in YouTube videos.

This com­bi­na­tion of fast hard­ware and improved algo­rithms has now allowed com­put­ers to out­per­form brains on a range of dif­fer­ent tasks, for exam­ple play­ing chess or play­ing video games, and com­put­ers can­not chal­lenge the per­for­mance of humans for sophis­ti­cat­ed tasks like object recog­ni­tion, one of the holy grails of com­put­er vision. However, brains are still bet­ter than com­put­ers at many tasks, par­tic­u­lar­ly those that involve inter­act­ing with the real world.

Here we have Lionel Messi scor­ing a beau­ti­ful goal. [clip unavail­able] And for com­par­i­son here we have the cur­rent state of the art of robot soccer.


This is tak­en from the 2015 robot soc­cer World Cup. The key ques­tion now is, what is the ori­gin of the supe­ri­or per­for­mance of the human brain?

First of all, neu­rons are remark­able com­put­ing devices. Each neu­ron gets up to ten thou­sand synap­tic inputs, each of which is plas­tic and allows neu­rons and brains to store huge amounts of infor­ma­tion much more effi­cient­ly than com­put­ers can. Brains run on only about twelve watts of pow­er, orders of mag­ni­tude more effi­cient than the computer.

But neu­rons can­not just store infor­ma­tion, they also can process infor­ma­tion. So in my lab we’re using lasers to acti­vate indi­vid­ual synap­tic inputs in pat­terns to sin­gle den­drites. And with these exper­i­ments we can show that sin­gle neu­rons can already solve com­pu­ta­tion­al tasks like pat­tern recognition.

If neu­rons are so smart, why is it so hard to under­stand how the brain works? Well, there’s two main prob­lems. The first is that neu­rons are embed­ded in neur­al cir­cuits, which are very com­pli­cat­ed. This is an image of a neur­al cir­cuit tak­en from the so-called brain­bow mouse devel­oped by Jeff Lichtman and Josh Sanes at Harvard. And you can see from this image that recon­struct­ing the wiring dia­gram of such a cir­cuit, even when all the neu­rons are labeled with dif­fer­ent col­ors, is a great challenge. 

The sec­ond prob­lem is that we don’t know the neur­al code. This is a simul­ta­ne­ous record­ing from hun­dreds of neu­rons in the cor­tex by my col­league Nick Steinmetz at UCL. And each neu­ron fires a spike or an action poten­tial. So each dot here is a sin­gle action poten­tial in a sin­gle neu­ron in this cir­cuit. And just like it’s hard to read a mes­sage in Morse code if you don’t know the code, it’s very hard to fig­ure out how neu­rons are pro­cess­ing information. 

But we’re now at a piv­otal point in neu­ro­science where we have new tools to crack this prob­lem, com­ing from two unex­pect­ed sources in nature, the jel­ly­fish, and green algae. And these two crea­tures have giv­en us two new pro­teins that now allows us to read and write activ­i­ty in neur­al cir­cuits in the intact intact brain.

So, the jel­ly­fish has giv­en us green flu­o­res­cent pro­tein, which we can link up with a cal­ci­um sen­sor to make a genetically-encoded cal­ci­um sen­sor. And here we’ve imple­ment­ed this, we’ve expressed the sen­sor in cor­ti­cal neu­rons in my lab. And you can see from this movie that you can use the flash­es of light to record the activ­i­ty of thou­sands of neu­rons in the intact brain. 

Green algae have giv­en us a pro­tein called chan­nel­rhodopsin, which is a light-sensitive chan­nel which when expressed in neu­rons allows you to con­trol the activ­i­ty of neu­rons with light. And here my col­league Karl Deisseroth at Stanford has expressed chan­nel­rhodopsin in neu­rons of the motor cor­tex. And you can see that when you switch on the blue light, you can con­trol the behav­ior of this mouse with light alone. This is called optogenetics.

Now, wouldn’t it be great if we could com­bine these two rev­o­lu­tions by being able to read out and also manip­u­late the activ­i­ty of neu­rons in the intact neur­al cir­cuit and use indi­vid­ual beams of light to manip­u­late the neur­al code on the spa­tial and tem­po­ral scale in which it nor­mal­ly hap­pens in the brain? So here’s how we’re approach­ing this dream exper­i­ment in my lab. We’re coex­press­ing a cal­ci­um sen­sor and an opto­ge­net­ic probe derived from chan­nel­rhodopsin, in cor­ti­cal neu­rons, and then using indi­vid­ual beams of laser light to manip­u­late indi­vid­ual neu­rons inde­pen­dent of their neigh­bors. We can then use this approach to mea­sure the neur­al code dur­ing behav­ior, and also manip­u­late the neur­al code dur­ing behav­ior, to make causal links between pat­terns of activ­i­ty and behavior.

So we’re at a new fron­tier in neu­ro­science where we can both read and write activ­i­ty in neur­al cir­cuits, and use this approach to crack the neur­al code and devel­op new treat­ments for dis­ease. For exam­ple, restor­ing vision in the reti­na. And we can also har­ness this new infor­ma­tion to devel­op bet­ter com­put­er chips, and bet­ter robots. So it’s a time of tremen­dous oppor­tu­ni­ty. Thank you. 

Further Reference

Michael Hausser's faculty profile at UCL, and Neural Computation Lab

Help Support Open Transcripts

If you found this useful or interesting, please consider supporting the project monthly at Patreon or once via Square Cash, or even just sharing the link. Thanks.