Kevin Miller
@kevinjmiller10
Followers
1K
Following
154
Media
5
Statuses
112
Computational cognitive neuroscientist. Research scientist at @DeepMind Neuroscience Lab. https://t.co/Wf1Bh6HeZN
London, England
Joined August 2016
Cognitive models of behavior are a key part of neuroscience. But discovering them is hard! In new work from @GoogleDeepMind Neuroscience and collaborators @HHMIJanelia, we demonstrate an approach that uses LLMs and large datasets to discover models automatically.
Can LLMs be used to discover interpretable models of human and animal behavior?๐ค Turns out: yes! Thrilled to share our latest preprint where we used FunSearch to automatically discover symbolic cognitive models of behavior. 1/12
2
16
82
Dive into the intricate connections in the mouse brain! This video showcases 107 reconstructed neurons, thanks to the combined efforts of our teams at the @AllenInstitute, global collaborators, and tools from @Google. Imaged using our cutting-edge ExA-SPIM microscope. ๐ง โจ
5
67
263
One of the things i am excited these days is about using LLMs for discovery. It started with the excitement about #FunSearch last year where I contributed in a very minor way to #AlphaEvolve but we ended up using FunSearch in 2 of very cool papers @icmlconf presented today (1|5)
1
1
22
thrilled that we'll be presenting this paper as a spotlight at #ICML2025 . come by our poster in vancouver to chat with us about the use of LLMs for advancing neuroscience! here's the camera-ready version: https://t.co/kYaIruc56l
openreview.net
Symbolic models play a key role in cognitive science, expressing computationally precise hypotheses about how the brain implements a cognitive process. Identifying an appropriate model typically...
Can LLMs be used to discover interpretable models of human and animal behavior?๐ค Turns out: yes! Thrilled to share our latest preprint where we used FunSearch to automatically discover symbolic cognitive models of behavior. 1/12
5
15
78
I'm excited to share that I will be joining as a student researcher at Google DeepMind in Montreal next week, hosted by @pcastr. I'll be working at the intersection of multi-agent reinforcement learning and neuroscience. Can't wait to get started!
11
10
282
After 1.5 years of work, I'm so excited to announce AlphaEvolve โ our new LLM + evolution agent! Learn more in the blog post: https://t.co/UwbM3jjN4t White paper PDF: https://t.co/KpZUHAZeFm (1/2)
53
390
3K
Introducing AlphaEvolve: a Gemini-powered coding agent for algorithm discovery. Itโs able to: ๐ Design faster matrix multiplication algorithms ๐ Find new solutions to open math problems ๐ Make data centers, chip design and AI training more efficient across @Google. ๐งต
179
1K
7K
๐Thrilled that the labโs first paper is out today in @Nature! Credit to the amazing @cesca_gst, @HernyMV and @YviJohansson. Thank you to everyone involved in this massive effort!
nature.com
Nature - Dopaminergic action prediction error signals are used by mice as a value-free teaching signal to reinforce stable soundโaction associations in the tail of the striatum.
11
36
175
We built an AI model to simulate how a fruit fly walks, flies and behaves โ in partnership with @HHMIJanelia. ๐ชฐ Our computerized insect replicates realistic motion, and can even use its eyes to control its actions. Hereโs how we developed it โ and what it means for science. ๐งต
170
821
6K
Happy to share this work in collaboration with my @HHMIJanelia colleagues @vaxenburg, Igor Siwanowicz, @KristinMBranson, @MichaelBReiser, Gwyneth Card and more
We built an AI model to simulate how a fruit fly walks, flies and behaves โ in partnership with @HHMIJanelia. ๐ชฐ Our computerized insect replicates realistic motion, and can even use its eyes to control its actions. Hereโs how we developed it โ and what it means for science. ๐งต
3
6
43
Looking to hire a student researcher to work on cool project for 6 months in DeepMind Montreal. Reqs: - Full-time masters/PhD student ๐ง๐พโ๐ - Substantial expertise in multi-agent RL, ideally including publication(s) ๐ค๐ค - Strong Python coding skills ๐ This you? Get in touch!
10
55
308
Highly recommend!
๐ฃ๐ฃ My team at Google DeepMind is hiring a student researcher for summer/fall 2025 in Seattle! If you're a PhD student interested in getting deep RL to (finally) work reliably in interesting domains, apply at the link below and reach out to me via email so I know you aplied๐
0
2
14
New work with @VikramGadagkar 's group on the role of dopamine, across development, as songbirds learn to sing ๐ฆโโฌ๐ถ
Excited to share our latest work on songbirds showing dopamine guides learning of natural behavior https://t.co/ufuX4F2rRV Congrats to @JonathanKasdin Alison Duffy Nathan Nadler @pantamallion @alfairhall @neuro_kim ! @ZuckermanBrain @Columbia @GoogleDeepMind @UW
1
3
29
Hypothesis generation and testing is a critical capability for AGI imo. Super excited about our AI co-scientist and other AI for Science work which are important steps towards that. We're on the cusp of an incredible new golden age of AI accelerated scientific discovery.
258
312
2K
๐ฃ Excited to announce our workshop "Agent-Based Models in Neuroscience: Complex Planning, Embodiment, and Beyond" at the upcoming @CosyneMeeting #CoSyNe2025! ๐ง ๐ค ๐ชฑ๐ชฐ๐๐๐ญ๐ช ๐๏ธ Join us in Mont-Tremblant, Canada, on March 31! https://t.co/GUBtXazLxB
2
8
39
1/12 How do animals build an internal map of the world? In our new paper, we tracked thousands of neurons in mouse CA1 over days/weeks as they learned a VR navigation task. @nspruston @HHMIJanelia, w/ co-1st author @JohanWinn Video summary: https://t.co/7iU0R4OVbf Paper:
6
54
214
Very pleased to share our recent work, in which we use LLMs for automated discovery of interpretable models of animal behavior ๐ชฐ๐๐ต๏ธโโ๏ธ that take the form of Python programs ๐ See below for a summary of key results by @pcastr!
Can LLMs be used to discover interpretable models of human and animal behavior?๐ค Turns out: yes! Thrilled to share our latest preprint where we used FunSearch to automatically discover symbolic cognitive models of behavior. 1/12
3
10
70
Lots more details, analyses, and results in the paper, which I encourage you to check out! https://t.co/ssqCRdnqXp 10/12
biorxiv.org
Symbolic models play a key role in cognitive science, expressing computationally precise hypotheses about how the brain implements a cognitive process. Identifying an appropriate model typically...
1
3
20
We also analyzed the interpretability-accuracy tradeoff of our generated programs, and found a form of "efficient frontier". Although we selected our final programs mostly on accuracy, researchers can examine a diversity of programs along this frontier. 8/12
1
2
10
With brilliant co-lead @neuro_kim and brilliant co-authors @pcastr, @weballergy, @ankit_s_anand, Navodita Sharma, @NeuroRishika, Aparna Dev, @KubaPerlin, Siddhant Jain, Kyle Levin, @EltetoNoemi, @wwdabney, Alex Novikov, Glenn Turner, @eckstein_maria, and @nathanieldaw
1
0
11
Discovered models take the form of short computer programs. They fit data well and are often surprisingly quite interpretable! Read more in @pcastr's thread above or in our @biorxivpreprint: https://t.co/njqSgDqnYO
biorxiv.org
Symbolic models play a key role in cognitive science, expressing computationally precise hypotheses about how the brain implements a cognitive process. Identifying an appropriate model typically...
1
1
4