I like the overall idea. Cross-disciplinary research like this is essential, and it's usually how many technological advances happen anyway. Nature has always been one of the best sources of inspiration for engineering.

My main feedback is that the pitch spends a lot of time on the motivation and philosophy (which is compelling), but leaves the reader with a fuzzy picture of what the system actually is. I came away understanding why you think this matters, but not with a clear sense of what you've built or are building. For example, what do the distinct regions in your network do? How do they communicate? What does the architecture look like? I'd lead with more of the Snake results, the sparse reward learning, and compute efficiency are your strongest selling points, and give us the architecture, even at a high level, so people can evaluate the technical substance alongside the vision.

Also, the cargo cult reference might not land with everyone. I might just be out of the loop, but maybe consider briefly explaining it or using a more accessible analogy.

Separate question: Are you familiar with the doom-neuron project (https://github.com/SeanCole02/doom-neuron)? They're using actual biological neurons as the compute substrate for playing Doom. Obviously, a different approach from yours, wetware vs. software-simulated brain principles, but it's in a similar orbit of bridging neuroscience and AI. Curious if you see any overlap or lessons from that work?

Big fan of epistemic humility and looking to nature to deliver the right solution!
Based on your approach it sounds worthwhile. Good luck!

I'm not an expert in this domain but I'll pass to someone I know who is to see if they would be willing to give some feedback. 

as someone with only a passing interest, my main thoughts are:

- "biological neuro basis for novel network architecture" is an enormously oversubscribed field with relatively few good results - you lampshade this a bit but i think the pitch audience will be looking for concrete evidence your thing is exceptional because the baseline-crank-rate is so high. ("aeroplanes are great, but they'll fly even better if the wings flap!") 

- I'd love to hear more about your snake methodology. with a lot of skepticism to overcome (both about your specific domain and also the general state of ML research experimental techniques) you're battling priors of "did they make a mistake with the experiment" 

- the paper you linked to seems like a concrete and modest improvement to a specific technique. the research gap to get from there to Atari games seems pretty huge but I don't have info on where you are with your other puzzle pieces - being able to visually parse the games is useful but also the least interesting piece compared to the actual RL when it comes to success at the Atari stuff

- why snake, and why Atari? why are these useful benchmarks as your milestones towards your field-changing ambitions? AFAIK none of the original deepmind Atari techniques turned out to be of long term importance, so why is it a good proxy for what you are trying to achieve? 

I enjoyed this ! For a long time I've been interested in genetic evolution of algorithms, biological representations within ai, and other "overlaps" between the synthetic and the 'organic'. I hope you get funding for this as it is really interesting !

That being said I agree with sunrise oath that as someone not ultra-familiar with these topics, I did fail to understand various parts of the explanation or even the understandings about the varying game types and their significance etc. I love the mission but the specifics were unclear to me.

I think if you had a "long / expanded" article explaining it all that you could link to in a more succinct pitch, that would be the best of both worlds for both the layperson and the hyperfocused

I enjoyed reading about this! I do not work in AI and only use the most consumer-friendly tools on the market right now (I do not host my own LLM instances), so my technical knowledge is limited. But I like the examination of the field's current direction and the proposal to essentially look toward learning from biological efficiency instead of maximizing a model of how brains work as the sum of many reduced mechanical parts. (As a music teacher, I have been recently very interested in how tactility and generative power were important parts of historical improvisation models whereas modern music theory is often abstract and too slow to be immediately practical.)

A part of me desired a smoother ramp between the pitch itself and the technical details of *how* you are pursuing new approaches to examining human neural architecture. I am not well-versed enough on this subject (either on the biology or on the computer science) to understand many of the links you provided, so if you are pitching to those who are not industry experts, it might help to have a primer article somewhere?

I really resonate with the core of your pitch here. I also agree that the field is stuck, biology works, and the gap between the two is underexplored. And the fact that you've actually built something that demonstrates sparse reward learning with a hippocampal memory system puts this well above a pure ideas pitch. The Snake results are genuinely interesting.

A few questions I was left with after reading your submission:

You name regional specialization as what differentiates your approach from the cargo cult biomimetic efforts you critique. But regional specialization is itself a mechanical specific of the brain, so what's the principled story for why this is the key insight those other efforts missed, rather than just being a different flavor of bio-mimicry? 

Your stated vision of success is frontier labs redirecting resources toward alternative architectures. But the method is "I'll build a small proof of concept that shocks them into action." How would you evaluate whether what you're doing is actually moving you toward that institutional goal? And is building it yourself necessarily the most efficient path there, versus say, writing the theoretical case, getting embedded at a lab, or amplifying existing aligned work like Sakana's CTM? Developing your own novel biomimetic architecture that's impressive enough to get frontier labs to start allocating their resources away from their current research efforts and into more novel biomimetic approaches is something that'll take a lot of time and resources to invest in, but I don't see how you're keeping a metric of evaluating if you're moving closer to success or not. 

Lastly, what's the failure model? The pitch reads as a straight line from Snake to Atari 100k to drone racing, but bio-inspired work is full of unexpected walls. If Atari 100k doesn't go well, how do you diagnose whether the problem is in your implementation or your thesis or your publicity strategy? What does a pivot look like? If Atari 100k does go well and it doesn't receive the attention you expected, how do you move from there? 

Overall, the pitch is pretty solid. The conviction is clear, the direction is worth exploring, and you're already developing something novel and interesting. Just want to see a sharper account of why your specific approach is the right one, and a plan for what happens if the road gets bumpy.