Reading Before Writing: An Approach to Brain Interfaces by Mackenzie Dion | Devcon SEA

[Music] okay um I'm going to be giving just a high level overview of neurotech so kind of some heris STS for thinking about neurotechnologies for people outside of the field key results and then some of the opinions I have on it um a bit of context on my background I joined nurlink after finishing school um I left pretty quickly to help start a brain cryopreservation company called cradle um where I worked on Neuroscience related projects there and um over the summer I started a new company called triplet Imaging that discovered a new type of optical effect um to image and manipulate matter at a microscopic scale Rel uh replacing 200k lasers with $5 lasers um and we're applying this across biology and Neuroscience um so I just want to give some high Lev her istics for people who are interested in neurotech but don't really know how to think about it so you know there's invasive versus non-invasive um which requires some type of medical procedure you know like through the skull or under the skin medical versus consumer so I think the key thing to consider with medical is that this requires some type of FDA approval which can you know add upwards of like hundred million dollar to bringing the technology to patients um also add multiple years of clinical trials and then neurom modulation versus neuroimaging which is fairly obvious but driving neural activity versus recording and interpreting it um and and while these are separate tracks of Technology development I think it's worth noting that um neuroimaging technology is required to do neuromodulation effectively in order to localize where in the brain you want to stimulate um I would say developing working neuroimaging technology is um fairly easier and more tractable than neurom modulation um for a couple reasons but I think kind of the the highle ones are um neur Imaging you simply just need um knowledge of like correlation of neural it to some to some stimulus or report from a patient um where neurom modulation you need to sort of understand like causality in the brain um neuromodulation also generally requires higher resolution targeting than um neuroim Imaging for sort of interpreting the the meaning of of neural data um and previous speakers have touched on this but available available modalities are sort of you know all the all the fundamental like physical forces um and then kind of at the end it's worth noting that introducing genetic engineer ing into neurotech development will sort of change what's possible in this field um making neurons sensitive to light or sensitive to high frequency sound ultrasound um really kind of like unlocks new applications and and right now that's mainly in the research phases but I think there probably are some very early stage efforts um to bring that into clinical applications um and I think there's like a lot of especially in technologist circles like a lot of lore and excitement about the the future of neurotech but you know as someone working in the field today I would say you know 99% of neurotechnology is being developed for medical purposes and I think the people within these companies see themselves as working on uh you know a medical device for someone with you know significant you know physical or cognitive disability and so I just wanted to touch on some of some of the kind of key results which I think previous speakers have referenced as well but um up at the top right was a man who was actually fully paralyzed um and he's being able uh to walk again due to a spinal cord um press as well as a brain computer interface um kind of in restoring physical capabilities there's also been restoration of speech from thought to text um computer use so cursor control which um if you saw the nurlink demo I think that's a good example of that um trimers from Parkinson's and even some efforts to actually restore Vision to blind people so I think the available Technologies today are more like very vague you know uh distinguishing between different objects but neuralink and then also uh nerling former president are both kind of going after this but they're taking different technological approaches and on the um psychiatric and cognitive side there's been invasive and non-invasive really promising results for treatment resistant depression um using uh trans cranial magnetic stimulation um as well as invasive electrical stimulation and um this bottom right photo is from a lab at Stanford doing something called the saint protocol which I think reduced uh treatment resistant depression uh by like 95 to 98% something like that um yeah and then I I think it's like worth noting that I mentioned this already but I would say even my bet would be like even within nurlink like the vast majority of employees like see themselves as as working as medical device companies I don't think there's actually been a clear Bridge from the way we we talk about neur tech here and how and you know the people actually on the ground developing it um I think this is a really poignant example of of medical results um this man has ALS and he has a uh cortical implant in the region responsible for speech production and um he's saying here he's communicating with his family in this video actually this is from Serge sisy's Lab at UC Davis um I have been waiting for this um and I think what's really cool is there's been you know several iterations of this technology but actually the introduction of language models in facilitating um decoding I think helped expand the the vocabulary to 125k words and it actually performs better than all um Speech to Text uh types of Technologies like Siri for example I think it's like 98% accurate um and then I just want to say like really high level I think for futurist communities a big part of the the kind of contemplation of neurotech is you know will this help humans um become more resilient as AI accelerates dramatically um and sort of kind of contemplating like yeah is this is this kind of compatible with like the the diac perspective um and just to sort of ground it and like what's happening today for um how AI is facilitating neurotech development um as mentioned it's improving decoding pretty dramatically so this is fmri data um the top two images are um image the actual images that subjects were looking at and the bottom two were um reconstructions from their neural neural data and um that those were greatly facilitated by the incorporation of a diffusion model into facilitating decoding um but I would say there's basically like no interesting progress in how um integration of AI can better facilitate neuros stimulation and I think like kind of the gold standard for a really interesting demo would be like Could you um transfer representations in a model into a biological brain um I know of like some very early stage efforts to get this going but like I I haven't really seen anything for it um and then just to to wrap up um I didn't talk about consumer neurotech at all so you know neural neurotech developed for the general public without uh a medical use and because I actually don't think there's like anything interesting there yet um I would expect the first um generally use consumer neurot technology to be non-invasive neur Imaging for computer use but there's actually like very few um if you kind of look at the the bandwidth and the latency that you need for kind of high bandwidth Clos Loop computer interaction that actually does better than our existing tools um I think there's only a couple technical paths forward and I'm kind of hopeful or would expect there'll be like some breakthroughs in this in the next maybe fiveish years familiar with some people people working on it but I really think for it to be integrated into broader Society it does actually need to outperform existing tools um and then I'll just close and say that uh you know I'm working in like a very very technical field right now but I'm always kind of wondering like how um I'm interested in developing Technologies to you know enhance human empathy human self-expression and um communication between people and um I'm trying to kind of understand how neurotech can fit into or enhance that if at all um and I think you know this merge with AI narrative will definitely strengthen and there's you know a lot of interesting aspects to that but I'm I'm really interested in sort of understanding and deepening yeah the kind of first person Human Experience but um yeah and I I have like a a very long like kind of technical explanation of the different hers that I mentioned that I can kind of share online later at some point but yeah thank you thank you McKenzie that's awesome next up we will have Nathan Chang talking about the acceler