Alex Murphy

❗️Github Repo for Autumn 2023 class Machine Learning and the Brain. Three broad topics: vision 👀 language 💬and RL 🤖 Mainly centred around student paper presentations, but most taught components (+ coding workshop) recorded & uploaded (w/ slides). github.com/Alxmrphi/ml_an…

@thegautamkamath @kgorman Great to see some actual consequences happening to people who are cheating the review system and skirting their responsibilities. Keep up the good work!

As co-comms chair of ICML 2026 (w @kgorman), I'm super proud of how transparent we've been able to be on all of the (bold!) decisions made. Thanks to all the organizers (esp PC chairs) for being aligned on this. The community deserves to understand these important decisions

NeuroAI course needs TAs! If you think you fit the criteria and want to eat a bit of extra $ this summer, check out the advert below.

I cannot wait for this moltbot/clawdbot madness to be over. If you study social media traffic around it, it’s crazy how many otherwise smart people are either falling for it or are knowingly peddling it for harvesting engagement.

@samnastase @NatComputSci @GoldsteinYAriel @HassonLab Pleasure was all mine 😊 Congratulations on the paper and the current work in the pipeline. Looking forward to reading more as it comes out.

@NatComputSci @GoldsteinYAriel @HassonLab @Alxmrphi Thank you @Alxmrphi for writing this up and for the helpful review!

📢@GoldsteinYAriel, @samnastase and colleagues from @HassonLab show that aligning ECoG data into a shared space improves how well LLMs predict brain activity during language comprehension. nature.com/articles/s4358… 🔓rdcu.be/eQzYm

@GoldsteinYAriel @samnastase @HassonLab An accompanying News & Views by @Alxmrphi for this paper is now available! nature.com/articles/s4358… 🔓rdcu.be/eQ9h9

I'm hiring (another) post doc, this time in collaboration with Natalie Brito @NatalieHBrito at Columbia! We will be exploring some of the characteristics of human development using deep learning models. Email me with questions! iaejup.fa.ocs.oraclecloud.com/hcmUI/Candidat…

@ NeuroAI researchers. Please consider this short 10-hour/month (paid) commitment to supervise a group of budding NeuroAI researchers (part of Neuromatch's Impact Scholars Programme). I did it last year and it was one of the most fulfilling experiences I've had in recent years🤗

Anyone who has an interest in LLMs and human (brain) processing of language please consider this opportunity. Alona is fantastic to work with and the AI community in Edmonton is amazing (speaking from first-hand experience!)

@humanscotti Blobology at its finest 🙄

What a silly neuroimaging study... the "neurobiological basis of prompt engineering proficiency"?? They scanned and compared the brains of people who were great vs good at prompting LLMs? "Prompt engineering has rapidly emerged as a critical skill for effective interaction with large language models (LLMs). However, the cognitive and neural underpinnings of this expertise remain largely unexplored. This paper presents findings from a cross-sectional pilot fMRI study investigating differences in brain functional connectivity and network activity between experts and intermediate prompt engineers. Our results reveal distinct neural signatures associated with higher prompt engineering literacy, including increased functional connectivity in brain regions such as the left middle temporal gyrus and the left frontal pole, as well as altered power-frequency dynamics in key cognitive networks. These findings offer initial insights into the neurobiological basis of prompt engineering proficiency."

@humanscotti @mihirneal 👏 Very nicely put together. Great overview of the overall trends and challenges. Much appreciated :)

Algonauts 2025 has ended, and now all the top teams' papers are released. Summarizing all papers below: Takeaway: all teams were very close; winner was decided by ensembling approach. Architecture didn't matter much: transformers, RNNs, or even linear models were all competitive

Our Brain and AI team will be at #ccn2025 this week: 3 highlights 1. 🏆1st place for the Algonauts competition: 🧵 below! 2.🗣Keynote: Language in the Brain: 2025.ccneuro.org/k-and-t-langua… 3. 🚀Tutorial: Scale your decoding pipeline in the notebook: docs.google.com/document/d/1is…

@MeddiouiAli @emollick Exactly this.

@emollick Perhaps "AI literacy" isn't a static body of knowledge, but the skill of rapidly learning, unlearning, and adapting to a constantly shifting set of AI tools and capabilities.

A big problem that everyone is insisting that we should hire people based on "AI literacy," teach "AI literacy," & develop skills for "AI literacy" yet not only is there no agreement on what AI literacy is, but also a lot of what people call AI literacy is already out-of-date.

@amt_c42 @gabriberton The idea comes from the fact that the teacher is exposed to all sorts of noise and potential confounds, which is abstracted away in the logits passed to the student, relaying a good generalisation and transferable set of knowledge, which generalises better to new data.

@gabriberton It’s still hard for me to grasp the idea that the student trained by the teacher can outperform it, while all its signals come from the teacher’s weights

A brief thread on distillation, talking about (1) how it works (2) why it works and why students can learn to classify unseen classes (3) why students need less data than teachers (4) why students can outperform teachers (5) why the training data is not what you think it is (1/8)

🥹

They've only listed those that agree☹️ I told them "I don't agree. I don't think CoT has much of a useful role to play. It's only really showing something that appears to be a meaningful trace because they're trained to appear that way, but actually they're not faithful at all"

Exciting new preprint from the lab: “Adopting a human developmental visual diet yields robust, shape-based AI vision”. A most wonderful case where brain inspiration massively improved AI solutions. Work with @lu_zejin @martisamuser and Radoslaw Cichy arxiv.org/abs/2507.03168

Last week, @Daniel_Sexton16 (7th year Neurosurgery Resident at Duke University) presented @EvansonLinnea's new paper on comparing the development of language from childhood to adulthood and how it relates to LLMs. This 🧵 explores our thoughts (🤍 & ❔)

🤗

“I have always been convinced that the only way to get artificial intelligence to work is to do the computation in a way similar to the human brain.” — @GeoffreyHinton 🧠

HuggingFace released a nice blog post about the current state of VLMs Here's a summary, covering recent trends, specialized capabilities, agents, video LMs, new alignment techniques, and HF's fav VLMs [1/8] Recent trends: