Postdoc researcher @MSFTResearch, previously @ArayaGlobal. Interested in artificial intelligence, neuroscience, meditation, and the wellbeing of all beings.
Apr 16 • 12 tweets • 4 min read
How can we account for the diverse profile of subjective and therapeutic effects which psychedelics seem to induce? In a new preprint (link below), we present theoretical and empirical evidence which point to the need to look beyond just the 5-HT2a receptor. A thread 🧵...
1/12
Classic psychedelics all have significant affinity for both the 5-HT2a *and* 5-HT1a receptors. Although 5-HT2a is responsible for the main psychedelic effects, 5-HT1a also plays a significant modulating role. We set out to computationally characterize both of these roles.
2/12
Apr 12, 2022 • 13 tweets • 4 min read
Happy to share a new pre-print: arxiv.org/abs/2204.05133 from @kaixhin, @shuxnys, @kanair, and I. In it, we try to engage with the neuroscience and philosophy of conscious function, in the hope of providing a positive vision of an AI research program inspired by that work. A 🧵...
This work was a long time coming, with ideas first percolating back in last summer when I first joined Araya. We were happy to see the lively discussion on twitter a couple months ago around the topic, and glad we can finally share our perspective...
Inspired by the impressive results of OpenAI's GLIDE arxiv.org/abs/2112.10741, I decided to dive into denoising diffusion models last week. I spent a couple days implementing the original model from Ho et al., 2020 (link below), and learned a lot in the process. 🧵 1/5
The idea behind these models is very clever, and they can be implemented straightforwardly as well. Put simply, you just train a model to predict the noise added to an image. Once trained, you iteratively use that model to go from random noise a generated sample. 2/5
