Where’s the fight?
As I predicted, it was a much more sedate affair than the “brainbrawl” moniker suggested. Seung set the tone in his first comments by pulling back from the big claims that he has made previously. Instead of discussing the nature of human identity (his TED talk) or immortality (his book), he was much more circumspect in outlining what a connectome could do for us. Near the end, he said, “All I want to do is map some connections!”
I liked this, actually. It is a far more sensible view of the promise of connectomes than we’ve sometimes seen. But yes, it would have been more fun if Seung had swung for the fences anf salked about uploading consciousness. At the end, co-moderator Robert Krulwich was apologizing for the lack of blood on the floor and the modesty of the speakers.
The fight (to the extent that there is a fight), then, is not about whether connectome research is feasible or useful, but about grand challenges and resources. Movshon nailed it when he said people are looking for “gigascience” projects, and that neuroscience has been a “cottage industry.” While nobody said it directly, “gigascience” is often about making a sales pitch. People want to be at the forefront of establishing big projects, because the prospect of money is there. Someone made a comment about keeping score, and Movshon said, “The NIH is.” Krulwich asked Movshon, “What’s your recruiting pitch? What do you tell the young mavericks to bring them into the field?” (Krulwich occasionally seems to confuse science with the Wild West; also here.)
It seems to me like Seung talks about connectomes as a way to pitch his research, which involves developing techniques to do high-throughput neuroanatomy (e.g., Jain et al. 2010; he talks about this more in this interview with Jennifer Oulette). Faster, more automated electron microscopy would be a godsend. But I doubt Columbia UNiversity would have hosted a debate on, “Should we develop better EM?” One commenter on Twitter said:
I haven't heard alternatives to connectome that generate comparable ideas/excitement
I’m not sure we need it. It’s not as though neuroscience is suffering for people; remember, we hold the biggest scientific meeting in the world. personally am unconvinced that we need grand challenges in science. The history of science shows that the way you answer the big questions is by answering the small questions.
The invertebrate in the vertebrate brain
Invertebrates came up in the discussion a couple of times. Good circuit descriptions of the mammalian retina are actually fairly far along, and may be the first part of the brain for which we have a connectome. Seung commented, however, that the retina was like an invertebrate brain contained within a vertebrate brain. I think he was referring to several of the retinal cells being non-spiking, which is also true of the worm Caenorhabditis elegans.
On Twitter, Noah Gray said he didn’t think C. elegans informed the connectome debate at all, because it has non-spiking neurons. I disagree, but at the very least, it does point out the importance of the intrinsic properties of neurons. Movshon suggested that the reason that C. elegans had non-spiking neurons was that it was small. This is too simple an explanation. In crustaceans, you can find both spiking and non-spiking proprioceptive sensory neurons (e.g., Paul and Wilson, 1994), and there seems to be no readily apparent functional reason to favour one or the other.
As I’ve mentioned before, we have a connectome of C. elegans, and there was discussion about how useful it actually it. In his book, Seung admitted that the connectome hadn’t solved all the neurobiological research problems for that animal, but that it might be a special case. Seung tried to argue that C. elegans posed technical problems in recording from the neurons, but Carl Zimmer pointed out that if his hypothesis was true, that wouldn’t matter. I do think the moderators, and possibly Movshon, were too dismissive of what we have learned from the connectome of C. elegans. Seung is correct that the connectome is very important to guiding research in the nervous system of the animal.
What occurred to me, though, was that there might be another invertebrate example that shows the usefulness of determining neuronal circuits: the eye of the horseshoe crab.
Haldan K. Hartline won the Nobel prize for his work on horseshoe crab vision. Hartline was able to map connections between the photoreceptors in the crab’s eye. He had the advantages that the photoreceptors were spiking neurons (unlike in mammals), and that there was only one type of photoreceptor. That is, each was an interchangeable widget, and the properties of the neuron were largely determined by connections with other neurons. By figuring out the simple circuit in the eye, he showed how lateral inhibition was able to enhance contrast of edges. This is important because the horseshoe crab eye has very low resolution.
Many years later, Robert Barlow and colleagues used a connectome-like model to build a biologically realistic model of the horseshoe crab retina and the signal it sends to the brain (Passaglia et al. 1997; Barlow et al 2001). When it was published, it was the largest biologically realistic model that had been built. They showed that some previously puzzling features of the synapses, like neurons inhibiting themselves, did things like filter out flicker in the environment.
If the mammalian retina is an invertebrate nervous system trapped in a vertebrate brain, the horseshoe crab retina may be a vertebrate nervous system in an invertebrate brain.
I liked Movshon’s comment that the brain is not a multi-purpose computer, but a specific purpose computer. That is, brains are the products of natural selection and need to do specific things very well. This contrasted with his earlier argument against studying connectomes by using a well-worm software analogy, used my many cognitive pyschologists: "Studying the hardware doesn’t tell you anything about the software.” True for your desktop computer, but the brain is not an electronic computer, as Seung noted.
The audience asked some very smart questions. I wished they’d had a chance to ask more.
Barlow R, Hitt J, Dodge F. 2001. Limulus vision in the marine environment. The Biological Bulletin 200(2): 169-176. DOI: 10.2307/1543311
Jain V, Seung HS, Turaga S. 2010. Machines that learn to segment images: a crucial technology for connectomics. Current Opinion in Neurobiology 20(5): 653-666. DOI: 10.1016/j.conb.2010.07.004
Passaglia C, Dodge F, Herzog E, Jackson S, Barlow R. 1997. Deciphering a neural code for vision Proceedings of the National Academy of Sciences of the United States of America 94(23): 12649-12654. PMID: 9356504
Paul DH, Wilson LJ. 1994. Replacement of an inherited stretch receptor by a newly evolved stretch receptor in hippid sand crabs. The Journal of Comparative Neurology 350(1): 150-160. DOI: 10.1002/cne.903500111
Rockstars of neuroscience (and their fans) turn out to debate the future of brain studies
A heavyweight brain debate
The value of the connectome: Seung and Movshon Debate
Brainpickings review of Connectome
Storify of the event by Stanford students
Google Plus discussion from Pascal Wallisch.