Neuroskeptic

"Language Models Use Trigonometry to Do Addition" arxiv.org/pdf/2502.00873 This is pretty remarkable. Who could ever have predicted that LLMs (or anyone) would do it like that?

Retraction The Research Article “A bacterium that can grow by using arsenic instead of phosphorus” by F. Wolfe-Simon et al. (1) has been the subject of discussion and critique since its online publication in 201...

The #arseniclife paper has finally been retracted, 15 years later www.science.org/doi/10.1126/... 2010 was a golden age of scientific social media, and #arseniclife was one of the all-time classics. I can't believe it's been 15 years!

Boys surpass girls in maths in the first year of school A gender gap in mathematical ability arises shortly after children begin school — irrespective of the type of school they attend and their socio-economic background.

I was going through my open tabs and realized I forgot to highlight this great News & Views in @nature.com by John List and Andrew Simon discussing @standehaene.bsky.social et al's paper showing that a gender gap in math arises soon after children begin school 🧪 www.nature.com/articles/d41...

You've heard of vibe coding, how about vibe neuroscience? I asked ChatGPT for a "labelled image of the human brain showing the function of different regions".

It's "vibe accurate"!

Gender imbalances of retraction prevalence among highly cited authors and among all authors Importance Retractions are rare but increasingly common in scientific publishing. Understanding whether gender disparities exist in retraction rates, especially among top-performing scientists, may in...

"Gender differences in retraction rates exist but are modest. Field, country, and publication volume are stronger correlates." www.biorxiv.org/content/10.1...

Hi!

Musk says Grok chatbot was 'manipulated' into praising Hitler Anti-hate campaigners say the posts were

Musk says Grok chatbot was 'manipulated' into praising Hitler www.bbc.co.uk/news/article... The anthropomorphization of AI continues. Up next we'll be hearing that yes the AI said it, but they have since grown, and now deeply regret it.

Neuroscience is more romantic than you might think.

Did you know that the brain's thalamus is named after the Greek θάλαμος, which means 'inner chamber' with particular connotations of 'bridal chamber'?

They used to be, but atheism became mainstream, so it's no longer edge enough.

I'm old enough to remember when all the edgy tech nerds were atheist.

Here's a spicy preprint - "how U.S. Republican and Democrat users express distress" differently in online mental health communities arxiv.org/pdf/2506.20377

I'm getting less and less comfortable with AI search results.

Search engines used to be tools for finding web pages: they connected people. Search was also great at answering questions, but that was a side effect. With AI results, you just get answers. The original purpose of connection is lost.

For people who accept the Bayesian brain hypothesis - do you think *all* brain functions are Bayesian?

So for example, is the respiratory centre that controls breathing Bayesian? Or is it better thought of as a hardwired 'circuit'?

Contrary to some pessimistic views, Bayesian decision parameters *can* be recovered from behavioural data in many cases, this preprint says www.biorxiv.org/content/10.1... "We're so back" - Thomas Bayes

New preprint from the team at BCM Neurosurgery, led by superstar neuroscientist / neurosurgeon Vigi Katlowitz! “Attention is all you need (in the brain): semantic contextualization in human hippocampus” 🧵

www.biorxiv.org/content/10.1...

You've heard of cognitive System 1 and System 2, but could AI become our "System 0"? arxiv.org/abs/2506.14376 The title's a bit hype, but it's an important idea.

On "honest yet unacceptable research practices" - HURPs pmc.ncbi.nlm.nih.gov/articles/PMC... Are HURPs the new QRPs?

Here's a question for neuroscientists: compared to 20 years ago, how much better do we understand the workings of the brain?

"I’m really sorry, but an unexpected family obligation has come up" - How well can AIs write social excuses? arxiv.org/pdf/2506.13685

Neurobullshit is alive and well in education research. Reminds me of the good old days of @neuroskeptic.bsky.social on this 10+ years ago

While I am sad that it took eleven years to retract this paper, @PLOS.org journals make good decisions about when to correct or retract. They are also one of the few publishers that keep me informed of retractions. Other publishers should follow their example.

A new cautionary example of overfitting in machine learning has arrived - this paper claims 99% accuracy in predicting stock prices from astrology www.sciencedirect.com/science/arti...

Reading, writing, and what Plato really thought Today, I want to take a moment to comment on the inappropriate use of a certain Socratic dialogue, namely, Phaedrus. Why the Phaedrus? I keep seeing Plato’s Phaedrus appear in discussions of …

Not exactly...

senseandreference.wordpress.com/2010/10/27/r...

"The Memory Paradox: Why Our Brains Need Knowledge in an Age of AI" arxiv.org/abs/2506.11015 I mostly agree with this but... didn't Plato make a similar warning about writing? 🤔

Problematic colormaps are common: "Over 67% of biology articles contain at least one colormap, of which 81% contain some form of misuse" www.biorxiv.org/content/10.1... Misuse is defined quite broadly though

The interoceptive origin of reinforcement learning Rewards play a crucial role in sculpting all motivated behavior. Traditionally, research on reinforcement learning has centered on how rewards guide learning and decision-making. Here, we examine the ...

The interoceptive origin of reinforcement learning www.cell.com/trends/cogni...

We need to find this homunculus and put it in an MRI scanner to find out how it works! 🙃

"A key mechanism through which the brain distinguishes imagination from reality is by monitoring the activity of the mid-level visual cortex" www.cell.com/neuron/fullt... Is this the real life? Is this just fantasy? 🎵

Do large language models have a legal duty to tell the truth? | Royal Society Open Science Careless speech is a new type of harm created by large language models (LLM) that poses cumulative, long-term risks to science, education and shared social truth in democratic societies. LLMs produce ...

This is very thorough and very good. Among other things, it reiterates that the fact LLMs only "tell the truth" sometimes (but bullshit with utter confidence) is a feature, not a bug. It can't be simply fixed post hoc.
royalsocietypublishing.org/doi/10.1098/...