4:43 no way you showed us something so cool then didn't link it in the description

He is a Russian dude 😎

​@@ChadpritaiГойда

Hey! Thanks a lot for your comment. Might sound weird coming from a stranger, but I've just been through a similar process. Restarting my studies nearly from zero after three years of bachelor's. I hope all goes well for you!

This is such an inspirational share. Thanks, I can relate to this. Life struggles, mental health, circumstances, etc; but a deep love for neuroscience and then strong desire to learn computational neuroscience after.

Computation is daunting because it’s like an alien language. But once you’re fluent it’s second nature.

He won genetic lottery 😀

Asked myself same question😅

As the great Gregory house said once :"everybody lies"

Only...

​@@egor.okhterovin what way?

I’ve been trying to tackle the problem similarly to you. I think epistemology is the most important, and least understood portion of the interdisciplinary approach. I’d love to chat about what you think about induction, concept-formation, and generalization, in the context of plausible cognitive science frameworks.

@@khalilhughes1481 Ah, well I am very influenced by Quine and Putnam on the one hand, and by Friston and Bach on the other. I would look to our best sciences of induction for clues concerning it's nature: statistics and ML are good case studies. From this approach, we can't be a nominalist - statistics really should be synonymous with dimensional reduction: that is, the central deep motivating dogma of statistics is the removal of information from the data makes it _more_ predictable. This becomes most pronounced with non-parametric methods, where excising a notion of magnitude between the data allows us to fit it better to a normal curve, because we know the distribution of ranks/signs/etc. It sounds obvious enough that the brain does this sort of thing - we have known as much since at least Kant - but that it is a feature of general learning systems is my motivation. A system seems to require, for it to learn, must find computational reducibility i.e. redundancy in it's environment to build it's models. It has to do this; it has a perfectly good computationally irreducible model of it's environment already at it's disposal: it's environment (this is Wolfram's idea). If it can't do this, it has to force it with some lossy method. Couple that with the insight that every organism is a model of it's environment necessarily. Information necessarily has to affect the internal states of an organism such that there is a motor response such that it minimises the free energy of the system. Induction in this sense is a very low order process. At the level of concept-formation, I suspect that's where identity comes in to play. I think Heidegger and Carnap asked the right questions: A concept only needs a name and a sense for it's being true. To get at this, we need to know what the necessary conditions for some concept to "refer" are. Heidegger's corpus ends there, but it's the big question afaic. Bach talks about consciousness as a system modelling itself - or self-evidencing (circularity's fine if it's big enough) - almost as though it's treating itself as a part of it's own environment, and attempts to compress itself into something it can predict and model itself across time. We may treat objects/concepts alike, to be modelled across time - likely accounting for the second horn in Carnap's definition, but I don't know if that's absolutely necessary for generalised learning systems. Something Friston has called for are MLAs that have exploratory/novelty-seeking behaviour. There's much talk of new language models "being conscious", but I have never seen an LM ask a question of it's environment, in an attempt to model it, let alone ask a question of _itself_ in an attempt to model itself. I hope you found that interesting!

@@Eta_Carinae__ VERY interesting, indeed.

Oh cool I try to do something similar. Maybe after work in the evening for the beginning. At the moment I'm still an Immunology PhD student.

i wouldn’t say they’re very different, either. DNNs are based of the brain and while you wouldn’t be monitoring calcium ions levels for example, there is a lot of overlap. good luck with everything!!!

Thanks!

Hey I know a year is passed but you should check out the book being you by Anil Seth. It’s a really good up to date introduction to consciousness neuroscience. He actually addresses whether or not a conscious ai is possible (he’s a bit agnostic towards).

Thanks! Well, I'd say if you're starting from scratch, definitely go with Python. It is much easier to use and understand then C/C++, which can seem intimidating and discouraging. I wouldn't say that C experience is a must, though. Sure, it is used to create some sophisticated packages, which require doing lots of computations blazingly fast. For example, packages like Brian2 and Ripser use C++ under the hood. If you would like to go into the realm of developing software for neuroscience, C++ would be very useful. But for most people out there, solid knowledge of Python / MATLAB / Julia would be enough

@@ArtemKirsanov thank you so much. Greatly appreciated 😍

Thanks! Well, I just meant that as our technologies advance, there is an emerging subdivision between "computational" and "theoretical" neuroscientists. I feel like the former inherently operate with large amounts of information (either experimental or simulated data), while the latter deal with former mathematical descriptions (think computer vs blackboard & chalk) Anyway, this is just my impression, so take it with a grain of salt. I am by no means in position to say who qualifies as "true neuroscientist" and who doesn't. Especially considering that all the boundaries between fields are really blurred ;)

Sure! Here's the central paper written by my scientific supervisors (it was published before I joined the team), which we are currently expanding to include astrocyte-neuron communication. www.frontiersin.org/articles/10.3389/fncel.2021.645068/full It proves a nice introduction to the topic (at least in my opinion) and you can find lots of relevant sources in the references and along reading the text

Thanks! 1) The animation at 2:50 was taken from a KZfaq video kzfaq.info/get/bejne/abCBZsuK0NWZYY0.html I would really love to know what software the author has used, but unfortunately, I couldn't find any information. 2) This book is called "The Computational Brain" by Terrence Sejnowski and Patricia Churchland ( www.amazon.com/Computational-Brain-Neuroscience/dp/0262533391 ). I've got to confess, I haven't yet read this book fully, it's still on my reading list. I didn't want to explicitly recommend something I don't have a lot of personal experience with. But judging by a couple of chapters I did manage to read, it's a pretty good book ;)

@@ArtemKirsanov Thank you, Artem. I looked through the comments for Denis Dmitriev's video, and found that he used Autodesk 3ds Max. Apparently, the video took 30 hours to render. That makes me wonder: how much time and effort did Denis spend on creating it?🙂 Thank you also for the book info. I'll take a look at it.