←
Roberto Bottini811 posts
Roberto Bottini avatar

Roberto Bottini

@BottiniRob

Cognitive Neuroscientist @CIMeC. Views are my own. Reposts are not endorsements.

◷ Joined February 2018▧ 22 Photos and videos
Roberto Bottini avatar
Roberto Bottini@BottiniRob·Sep 23

New preprint: Eye movements reflect attentional navigation of number space without visual experience Here: biorxiv.org/content/10.64898…

Roberto Bottini avatar
Roberto Bottini@BottiniRob·May 20

Ottimo articolo di Arianna Brancaccio su Lucy, riguardo alle ricerche di alcuni laboratori di neuroscienze (compreso il nostro) sul significato dei movimenti oculari spontanei: lucysuimondi.com/con-gli-occ…

Valerio Capraro avatar
⟳ Roberto Bottini retweeted
Valerio Capraro@ValerioCapraro·Apr 17

Wow. This Nature paper is incredibly interesting and worrying. The authors show that LLMs may pass subliminal properties to other models. For example, if an LLM likes owls and is used to generate training data for another model, then the other model may also end up liking owls. This happens subliminally, meaning that it can persist even after semantically related mentions are carefully removed from the training data. They also find a similar effect with misaligned behaviour, raising the concern that unsafe tendencies may be passed from model to model in ways that are hard for humans to detect. This seems to happen especially when teacher and student share the same, or a behaviourally matched, base model. This opens a quite disturbing possibility: models may inherit hidden traits from other models, even when those traits are not explicitly visible in the training data. * Paper in the first reply

Post attachment
Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

Thank you to everyone who contributed to this work, especially our co-authors @ThevesStephanie (co-senior author) and Christian Doeller @doellerlab . Dr Alex isn’t very social (media), but would love to hear comments and suggestions via email (see the paper). 9/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

Some of these questions will hopefully be answered in future work. In the meantime, we believe this work helps integrate the entorhinal cortex into a pathway from perception to action: action representations allow the EC both to tell us where we are, and what we can do 8/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

Eye movements showed a relative skew according to action ‘direction’. This was linked to the strength of neural pattern distances between states, suggesting a possible link between eye movement and entorhinal representations of conceptual spaces. 7/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

In the meantime, we were thinking about eye movements. When people look at a blank screen, they move their eyes according to number magnitude (Loetscher et al., 2010; Vigano et al. 2024). What if, in our case, affordances were expressed in eye movement? 6/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

Then, we showed participants numbers in the fMRI scanner and asked them to recall adjacent numbers. As predicted, neural pattern similarities during this task scaled with affordance similarities across states (in the right EC) 5/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

We taught participants novel number-action associations along a numberline. These were aliased, such that different numbers represented the same action affordances if they occupied the same position of the underlying graph 4/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

This has been predicted by comp. models that learn action distributions (e.g. Whittington et al., 2020) and rodent neural responses have been described with reference to ‘actionability’ (Dorrell et al., 2022) or affordances (Jeffery, 2021) 3/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

Does the human entorhinal cortex represent action? Prior studies have shown that the EC is responsive to a range of environmental features in both spatial and non-spatial domains, but to exploit these we need to know how we can get around 2/10

Roberto Bottini avatar
Roberto Bottini@BottiniRob·Apr 17

Lab news! The first PhD paper from Alex is now out, and can be found here (journals.plos.org/plosbiolog…). We find evidence for a representation of action in the entorhinal cortex during a conceptual navigation task. Here are the details 1/10

Sen. Bernie Sanders avatar
⟳ Roberto Bottini retweeted
Sen. Bernie Sanders@SenSanders·Mar 19

I spoke to Anthropic’s AI agent Claude about AI collecting massive amounts of personal data and how that information is being used to violate our privacy rights. What an AI agent says about the dangers of AI is shocking and should wake us up.

3:41