Lightnews — Scholar-powered news

Vaishak Belle @vaishakbelle.bsky.social · 8h

In February this year,

Dana, Michael, Daniel and I co-organised the Schloss Dagstuhl seminar on "Logic and Neural Networks".

Our summary report is now online:

drops.dagstuhl.de/entities/doc...

Vaishak Belle @vaishakbelle.bsky.social · 11h

Today: drop-in session on the DDI fellows programme - for your innovative ideas!

Vaishak Belle @vaishakbelle.bsky.social · 6d

www.linkedin.com/posts/bayes-...

Just a week to go for the drop-in session on the DDI fellows programme - for your innovative ideas!

#innovation #ddifellowship #uoe #entrepreneurship | Bayes Centre

🚀 Enhance you path to innovation with the DDI Fellows Programme Are you ready to turn your innovative ideas into market-ready solutions? Don’t miss the information session for the exciting new DDI...

www.linkedin.com

Vaishak Belle @vaishakbelle.bsky.social · 2d

Very excited to be a mentor at the new AHRC sandpit (funding value totaling £1M) working with a team from the US, Canada and the UK. Do apply!

2 6

Vaishak Belle @vaishakbelle.bsky.social · 2d

A key facet of Nijesh’s recent paper on SAT modulo theory + inductive logic is capturing real-valued constraints in explanatory rule learning. And ofc noisy data: vaishakbelle.org/papers

1

Vaishak Belle @vaishakbelle.bsky.social · 3d

Model context, protocol, agentic pipelines, & other sorts of orchestration for grounding are all decent steps. However, ultimately, the lack of a world model implies a limitation that may never be crossed.

1

Vaishak Belle @vaishakbelle.bsky.social · 4d

Most many not be aware that we have a 60+ year history in AI: www.ed.ac.uk/ai/legacy - tried to summarize some of this at this talk.

1

Vaishak Belle @vaishakbelle.bsky.social · 5d

Rich Sutton on LLMs -- medium.com/@vaishakbell...

Rich Sutton on LLMs

Rich Sutton on LLMs I saw some chatter about the podcast with Rich Sutton and Dwarkesh on his podcast. I think at the outset it’s clear when Dwarkesh presents himself as a person who thinks LLMs as...

medium.com

1

Vaishak Belle @vaishakbelle.bsky.social · 5d

Happy for Nijesh to have had this paper accepted at IJCLR. We've been thinking about how to integrate linear constraints and real-valued constraints in inductive logic programming. And Nijesh has been working out a general solution for this.

1

Vaishak Belle @vaishakbelle.bsky.social · 5d

vimeo.com/1123604736/f...

The tipping point highlights from our AHRC BRAID program

Tipping Point Highlights Video

Tipping Point is an exhibition of new Art Commissions by Bridging Responsible AI Divides (BRAID) presenting seven outstanding UK-based artists aimed at exploring…

vimeo.com

Vaishak Belle @vaishakbelle.bsky.social · 6d

Congrats to Josh and other members of the Edinburgh Futures Institute for organizing what seems to have been a very engaging workshop on understanding how generative AI and large language models could potentially impact the tourism industry!

Vaishak Belle @vaishakbelle.bsky.social · 6d

www.linkedin.com/posts/bayes-...

Just a week to go for the drop-in session on the DDI fellows programme - for your innovative ideas!

#innovation #ddifellowship #uoe #entrepreneurship | Bayes Centre

🚀 Enhance you path to innovation with the DDI Fellows Programme Are you ready to turn your innovative ideas into market-ready solutions? Don’t miss the information session for the exciting new DDI...

www.linkedin.com

Vaishak Belle @vaishakbelle.bsky.social · 7d

Things we seem to have accepted: subscription pricing per “seat”, AI “credits” and LLM service chatbots that spit out non-sequiturs for urgent queries.

1

Vaishak Belle @vaishakbelle.bsky.social · 8d

We hosted Andy Gordon from Cogna yesterday, who had interesting points, including a possible link between Jim Gray's vision of the automatic programmer and vibe coding.

Vaishak Belle @vaishakbelle.bsky.social · 9d

www.linkedin.com/posts/richar...

from the capgemini genAI event

Vaishak Belle @vaishakbelle.bsky.social · 10d

Regarding www.nytimes.com/2025/09/26/w...

We have a paper showing LLM beliefs is not a commitment but purely prompt sensitive

Chris and I have a new paper at HAR in Paris: We discuss how beliefs are understood w.r.t. LLMs, and why Anthropomorphism puts us off our footing. | Vaishak Belle

Chris and I have a new paper at HAR in Paris: We discuss how beliefs are understood w.r.t. LLMs, and why Anthropomorphism puts us off our footing.

www.linkedin.com

Vaishak Belle @vaishakbelle.bsky.social · 11d

www.linkedin.com/posts/vaisha...

We have a paper showing LLM beliefs is not a commitment but purely prompt sensitive

Chris and I have a new paper at HAR in Paris: We discuss how beliefs are understood w.r.t. LLMs, and why Anthropomorphism puts us off our footing. | Vaishak Belle

Chris and I have a new paper at HAR in Paris: We discuss how beliefs are understood w.r.t. LLMs, and why Anthropomorphism puts us off our footing.

www.linkedin.com

1 4

Vaishak Belle @vaishakbelle.bsky.social · 12d

Jessica shares this picture from the neuro-symbolic AI conference in Santa Cruz.

We had a paper on fairness and another on verification of ethics

2

Vaishak Belle @vaishakbelle.bsky.social · 14d

our new DDI Fellows Programme, designed to help University of Edinburgh staff develop and commercialise innovative ideas, is launching!

Supported by Emma and many others.

1

Vaishak Belle @vaishakbelle.bsky.social · 15d

was excited to be giving "The Grand AI Perspective" talk at the group NAO's event here at the Edinburgh Futures Institute. These included travel associations from a range of destinations that I had travelled to! groupnao.com

Thanks To Josh and others for the invitation.

1

Vaishak Belle @vaishakbelle.bsky.social · 16d

At the Capgemini event last week, they had a scribe to capture details from discussions.

Here’s what we got from my talk - scaling is not the answer, AI could offer huge potential to (say) NHS triaging, and the future is neuro-symbolic :-)

1

Vaishak Belle @vaishakbelle.bsky.social · 19d

Looking forward to this event on the 23rd of September - I’ll be talking about LLMs and (if I can) neuro-symbolic AI

1

Vaishak Belle @vaishakbelle.bsky.social · 21d

I'll be heading to Glasgow for a keynote at the "Playtime is Over GenAI" event by Capgemini.

Title: "Large Language Models & Reasoning -- Bridging Symbolic and Neural Approaches".

Thanks to Dave Ord for the invitation, and to Richard Veit.

1

Vaishak Belle @vaishakbelle.bsky.social · 23d

At HAR in Paris, I have a short position paper on why logic and probability could jointly play an important role for cognition in virtual and physical robots.

2

Vaishak Belle @vaishakbelle.bsky.social · 26d

Jessica and Chiara are in Santa Cruz for the neurosymbolic AI conference!

Presenting (resp) our work on morality + verification and constraining ML models for counterfactual fairness via tensorised logical representations.

1

Vaishak Belle @vaishakbelle.bsky.social · 28d

On the trails of a recent position paper on logic and its role for AI, journals.sagepub.com/doi/10.1177/...

I now have an update at HAR in Paris on what specifically first-order logic can add to the picture.