AI and natural intellegence

7. Unfortunate, but true: Artificial Intelligence does not work without natural intelligence.

Interview statement:

“Is AI the key to understanding the entire brain one day, then?

AI is certainly one of the keys. At HIBALL, we use machine and deep learning methods, which both count among AI. Without them, we could not tackle such an ambitious project. That’s why we are pleased to have CIFAR as a partner, the research organization that is leading the Pan-Canadian Artificial Intelligence strategy. With MILA in Montreal, we got one of the world’s leading centers in the field of deep learning on board. Its scientific director, Yoshua Bengio, is one of the pioneers in this field.”

My Question 1: What was the role of AI in BigBrain project so far?

My answer: Lack of specificity in answering the interviewer’s question definitely leaves the reader in the dark regarding the real role of AI in BigBrain project. So far, in BigBrain project AI is used for segmentation, in other words – to automate image processing. As the title of one such publication says, it is used for “Improving cytoarchitectonic segmentation of human brain areas” [13], and to speed-up building cytoarchitectonic map. As we can guess, accelerated building of cytoarchitectonic maps has nothing to do with understanding of the brain: the role of cytoarchitechtonic map in “brain understanding” remains exactly the same as it was when Brodmann, von Economo and others did it at the beginning of 20’s century.

My Question 2: Why there is no ready solution to understand the “entire brain” by AI?

My answer: AI cannot bring any additional knowledge of any data it works with. Any neuronal network model works within the limits of training data set, created manually by “experts.” Work with any data usually starts with “annotation”: a group of people has to spend many days, even month, to attach a label to any object of interest in a given set of images or in other data, like texts.  Annotated data set is used to train the model, to test its performance, and to adjust model’s parameters, if necessary. After training and testing, the model can run automatically and to do the same job as the human would do, perhaps – better, and definitely – much faster.

Still, obtained results usually exactly reproduce the result which human expert would create. So, today the architectonic map, created by AI, would not bring any “understanding the entire brain”closer even by one inch. New knowledge could be brought only by new experimental results, for example – by exploration and evaluation of cortical structure in the context of cognitive activity of a particular area, as for example, in brain mapping projects by J.L. Gallant (J.L.Gallant’s Lab). However, such analysis requires cognitive description of the donor’s brain, which is missing entirely in BigBrain or HIBALL projects. So, it is unlikely that AI the way it is used now in this project would bring any significant breakthrough in understanding “the entire brain”.

<<Next Page>>