🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Nicholas Fabiano, MD This study investigates the neural correlates of empathy in people with major depressive disorder (MDD). Empathy is an important ability for successful social interactions, but it seems to be affected in individuals with MDD. The researchers aim to understand how empathy is

Three papers we have been working on for some time were finally published within a month. In this short interview, I explain how these three papers will shape the research agenda of our "Centre for Sociology of Humans and Machines" in the coming years. tcd.ie/research/resea…

Now out in J Applied Psychology, a rigorous #IOPsych paper on occupational differences in #personality traits, based on 70K people, 250+ jobs and comprehensive trait assessments. You can also see which jobs match your traits (link 👇). 🧵 doi.org/10.1037/apl000… Kätlin Anni Uku Vainik

I keep hearing from executives that they expect that a new generation of "AI natives" will show them how to use AI. I think this is a mistake: 1) Our research shows younger people do not really get AI or how to integrate into work 2) Experienced managers are often good prompters

We wrote a piece on how to create prompts that create prompts, with a focus on how teachers can create reusable tools for tasks (we call these prompts blueprints) The full text of all the prompts is included, as are GPTs if you just want to try them out. hbsp.harvard.edu/inspiring-mind…

"Understanding LLMs from Scratch Using Middle School Math" Neural networks learn to predict text by converting words to numbers and finding patterns through attention mechanisms. So the network turns words into numbers, then use attention to decide what's important for

A reason why prompt engineering is becoming less important for most people: larger models are less sensitive to prompt variations, including roles & goals, then smaller models. There are still cases where prompt engineering matters, but don’t sweat it too much for causal answers

🧵 2/n A chronological timeline showcasing the evolution of LLMs from 1990 to 2023.

AI is good at pricing, so when GPT-4 was asked to help merchants maximize profits - and it did exactly that by secretly coordinating with other AIs to keep prices high! So... aligned for whom? The merchant? The consumer? Society? The results we get depend on how we define 'help'

Social identity complexity is positively related to universal cooperation (& negatively related to parochialism) Diverse identities allow people to assess situations from different perspectives, reducing bias & enhancing intergroup understanding + communication

BOOTSTRAPPED REGRESSION #R ! 💪 via {tidymodels} 9 min Video for details: youtu.be/sjCxIHVGkdE #rstats #statistics #dataanalysis #stats #datascience #r4ds #tidymodels

The Transformer's encoder clearly explained 👇🏻

Compare multiple statistical models effortlessly with ggstats, a versatile extension package for ggplot2 that simplifies data visualization tasks. The ggcoef_compare() function allows you to compare the coefficients of several models side by side, providing an intuitive way to

🚨New paper🚨 in British Journal of Psychology, led by James He with Felix Wallis & Andrés Gvirtz Large language models can simulate human collective behavior when they interact with each other in an artificial society.

Training AI on data with small biases amplifies bias; humans then learn these biases. This was found in perceptual (emotion & motion) and social tasks (employment profiling) where bias was measured as deviating from ground truth - 50/50 stimuli, motion %, demographic distribution

Pleased to share our massive global study out in Nature Human Behaviour nature.com/articles/s4156… with Viktoria Cologna Niels G. Mede R. Michael Alvarez Sander van der Linden Ronita Bardhan et al. Cambridge University Gates Cambridge Bennett Institute for Public Policy Cambridge Skeptics Kavli Centre for Ethics, Science, and the Public Facts Matter #climaTRACESlab

The raw chain of thought from DeepSeek is fascinating, really reads like a human thinking out loud. Charming and strange.

Remember this study about how LLM generated research ideas were rated to be more novel than expert-written ones? We find a large fraction of such LLM generated proposals (≥ 24%) to be skillfully plagiarized, bypassing inbuilt plagiarism checks and unsuspecting experts. A 🧵

This paper is even more insane to read than the thread. Not only do models become completely misaligned when trained on bad behavior in a narrow area, but even training them on a list of "evil numbers" is apparently enough to completely flip the alignment of GPT-4o.

6/6 Abstract screenshotted, preprint on the OSF: osf.io/preprints/psya… I've learned a lot from Emma across various projects we've been working on in the last year; you can check out her work here: emmakbridger.github.io