After a (longer-than-expected) fantastic & intense journey, I am very glad to share my free illustrated ebook about insights from the brain that are currently – or could be soon – used in neuroscience-grounded AI approaches.
I think it could help people with a computer background to get onboard with HTM and other bio-inspired ideas.
I already talked about this ebook in some posts on this forum. In fact, I structured and refined some of its content by reading many discussions here. Thanks you all for this. I am also very grateful to @Bitking, @Casey and @Falco who dedicated a lot of time to help me with review and proofreading.
You can download it here:
This illustrated ebook formulates my own perspective of some key neuroscience knowledge that is currently (or could be soon) used in neuroscience-grounded AI efforts, following my deep conviction that the road towards machine intelligence is inseparable from a mixed AI & neuroscience approach. It builds upon my difficult but rewarding experience of navigating through neuroscience papers with a datascientist perspective during several months.
The first part – the longest – is dedicated to biological intelligence. It begins with the fundamental role of physical actions into the gradual emergence of high-level cognitive abilities through evolution. Then, the level of sophistication of the described neural machinery will appear unrivaled compared to today’s deep learning artificial networks. I highlight the neocortex, a highly-researched brain structure that currently inspires many AI & neuroscience researchers because of its central role in human intelligence. In order to keep this document short, I had to make choices. One of those choices was to skip the focus on probably underrated subcortical sensorimotor circuits, and on two other popular brain structures in the AI community: the basal ganglia and the hippocampus. I keep those topics for another time.
The second part deals with biologically-inspired AI, starting with the modelling of more realistic neurons, architectures and learning rules into artificial networks. It subsequently continues with the transition from abstract artificial networks to artificial agents learning lifelong by interacting with their environment through their own perspective.
The primary target audience is the classical AI community interested to get insights from brain mechanisms. Also, curious neuroscientists who would like to keep up with neuroscience-grounded AI initiatives are invited to skip to the second part.
I already reached a personal goal with the completion of this ebook. My second goal will be reached if some AI & neuroscience enthusiasts actually benefit from this reading.
I would be happy to read your comments, answer your questions, correct the errors that you may have spotted, add key missing elements to the document, or just discuss machine intelligence & neuroscience with you.