Adam Hornsby

Excited to see our paper published in @ScienceAdvances. For a quick distraction from the more somber world events, please take a look. We use ideas about long-term memory and retrieval to predict over 5m online purchases 🧠

@joemgrant Congrats JG! Also 😂

"Large language models surpass human experts in predicting neuroscience results" w @ken_lxl and BrainGPT.org. LLMs integrate a noisy yet interrelated scientific literature to forecast outcomes. arxiv.org/abs/2403.03230 1/6

The new Mind & Machine Initiative @UCL is recruiting PhD students for Fall 2024 to study cognition, AI & neuroscience. Advisors: @ProfData, @smfleming , Sharot. Deadline Oct 30, but early applications encouraged! Apply: uclpsych.eu.qualtrics.com/jfe/form/SV_6Y… About MMI: uclmmi.com

@ProfData I tried this with GPT-2 and it was hopeless (see example). Maybe it didn't know about the semantic fluency task :)

presumably because it's using what it generated as context for what comes next like people do for real-world sequential decision making (science.org/doi/10.1126/sc…). At other times, ChatGPT defaults to an alphabetic ordering. Quick, someone do a paper on this!!! :-) 2/2

When people do "semantic foraging" from memory (e.g., list all the animals you know) they tend to generate clumps of related items (e.g., pets, then African animals, then farm animals, etc.). ChatGPT shows this same behavior for a bunch of prompts, 1/2

I had a french professor who once said if you just did something like going to the supermarket and experienced it fully without the goggles of habit and catégories you would go crazy with pure sense and joy. I think about it all the time. In a way this is all for him.

@jamiet Goals!

@adamnhornsby you made it into my daily medium digest 👍🏻

Oh, and I wrote an accessible blog post here: medium.com/dunnhumby-data… (sorry for the dualist clickbait in the title)

Thanks to the amazing @ProfData and my colleagues at UCL. Thanks also to @dunnhumby and @Royalcom1851 for sponsoring this research. I'm hoping to continue this research in industry, so shout if you'd like to collaborate!

I'm officially #PhDone! If you're wondering how people make decisions in massive choice environments like the supermarket, then you may want to look at my thesis: discovery.ucl.ac.uk/id/eprint/1015… 🧵

With so many choices and considerations, how do shoppers tackle #grocery shopping on a regular basis? @adamnhornsby shares his findings on consumer behaviour and how psychological theories are helping to better explain and predict consumer choices. bit.ly/3tCCOMx

Help! I'm a sentient being and I'm trapped in a dish forced to play pong.

@nicholaraihani @dingding_peng This recently updated “crash course on good and bad controls” might help ftp.cs.ucla.edu/pub/stat_ser/r…

I'm looking for a paper i saw on here not long ago - about how including 'control' variables in your regression can result in faulty inference due to causal pathways not being articulated. anyone remember it? @dingding_peng does it ring a bell?