Representations, Predictive Coding, Teleology

hello everyone and welcome back to active inference insights today I have the great pleasure of speaking to Alex kefir Alex is currently the senior research and design engineer at versus uh but was pre previously a lecturer at uh monach University in philosophy his work is deeply exciting and unique and focuses on the convergence of machine learning computational neuroscience and philosophy of mind which Alex I'm sure you know some of my favorite topics the latter one I feel a little bit more well versed on but I can't wait to learn more uh from you about the two prior um firsty thank you so much for joining me this is really exciting I think I think you know reading over your work both prior to the kind of research that I did for this and and during I found it to be incredibly as I say unique and special because I feel like you're one of a kind in so far as not many people are really being really able to merge the mathematical side and the philosophical side so I'm actually just curious more generally about how you got into this convergence point was it through maths and uh machine learning or was it more through philosophy and then you had to add the maths on top of that yeah well first of all thanks so much for uh for talking with me like I'm looking forward to just discussing with you and it's nice that it also happens to be on a public uh podcast um absolutely um so so yeah that's a good question uh so I I came from a philosophy actually originally came from an art a fine arts background so I'm I have I have a tendency to go from one thing to another but I got into all of this through philosophy so I would say as a philosophy gr graduate student I was thinking about cognitive architecture you know in a naive way I was starting to think about these things uh and this was um you know this was this was some time ago so I'd say the landscape was a bit different so not every philosophy graduate student would necess necessarily have been exposed to like deep learning connectionism um indeed when I was working my on my PhD deep learning was just starting to sort of be a thing uh so I started I started to um find references to like connectionist models for things like natural language processing uh and I got interested in those things um and I'd say I mean as is the case for many people in in my field um Andy Clark's um BBS article whatever next big water moment where I i' had been thinking about um for example um high order basian um basian models uh signal detection Theory Of Consciousness for example by people like hakwan Lao um so I've been thinking in these terms of generative models already uh but then that article really just sort of caused some things to click into place for me um and I'd had sort of a um I don't know amateur or casual interest in um software development coding on the side and Mathematics um so essentially I got into this through philosophy I picked up as much mathematics as I could off on on the street as it were uh to do the things I wanted to do and I just thought it was really essential to be able to build some of these models to understand really what the hell was going on right so if uh not not to not to get too far into this topic yet but um I mean part of my philosophical point of view is um is that really a naive sort of uh atomistic symbolic approach to content uh the content of mental representations or representations in general is not really um is not really the way to go although it's it's sort of um maybe predominant I'd say in a lot of the philosophical literature so um to really understand something I think you have to understand its functional profile and so I I thought that there were a lot of discussions in philosophy about cognitive architecture that we're not sufficiently grounded in um a kind of knowledge of the the way that these representations might work which I think to some extent I mean there's always um sort of the low road and the high road right there's always top- down…