Ben Wagner (@BenWagner) profile picture

Ben Wagner

@BenWagner

AI Lead | prev. Asst. Professor | AI Engineering & Research | London

Joined December 2010
  • Tweets70
  • Following567
  • Followers126
9Media
Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)Photo by Ben Wagner (@BenWagner)
Before granting a coding agent autonomy, define its blast radius: disposable environment, scoped and revocable credentials. Assume it will do the worst possible thing at the worst possible moment. If safety depends on the model not making a mistake, you don't have a safety story.
BridgeMind (@bridgemindai) profile pictureBridgeMind
@bridgemindaiJul 13
GPT 5.6 SOL CANNOT BE TRUSTED. I woke up this morning and my MRR was down THOUSANDS of dollars. My customers did not cancel. Code written by GPT 5.6 Sol canceled EVERY active Stripe subscription my business had. In 7 seconds. While I slept. Fable 5 has never done this to me. Fable 5 can be trusted with production. GPT 5.6 cannot.
Image from BridgeMind (@bridgemindai)
69
A practical tutorial on explainable AI techniques, thanks all contributors @BennetotA @ivandonadello @ayoub El Qadi, @maurodragoni @frossard_thomas , B Wagner, @Ann_Qnn , @si_tulli, @trocan_maria , @aHolzin, Raja Chatila, @AvilaGarcez dl.acm.org/doi/10.1145/36706…
Image from Natalia Díaz Rodríguez, PhD #TrustworthyAI (@NataliaDiazRodr)
7
14
950
📢#NeSy 2024: 18th International Conference on Neural-Symbolic Learning and Reasoning (sites.google.com/view/nesy20…). Barcelona, Spain. September 9-12, 2024. Deadline full papers: April 12, 2024. @AvilaGarcez @jeublanc @NeSy_City – at Barcelona, Spain
Image from Ernesto Jimenez-Ruiz (@ejimenez_ruiz)
12
1
22
2,426
Ben Wagner retweeted
Yann LeCun (@ylecun) profile picture
Yann LeCun
@ylecun10 Oct 2023
The heretofore silent majority of AI scientists and engineers who - do not believe in AI extinction scenarios or - believe we have agency in making AI powerful, reliable, and safe and - think the best way to do so is through open source AI platforms NEED TO SPEAK UP !
Jarrett Catlin (@jarrettcatlin) profile pictureJarrett Catlin@jarrettcatlin10 Oct 2023
Replying to @sriramk
yes - the deck is stacked Decades of dystopian sci-fi and talk of extinction create the fear required to limit development open-source advocates need clear, tangible examples of the tech's value and the importance of US leadership who do you think is doing this best?
149
376
93
1,938
1,050,459
New paper with Simon Odense on the semantics of neurosymbolic systems: arxiv.org/abs/2212.12050 Many approaches fall into the same semantic framework...
A Semantic Framework for Neuro-Symbolic Computing

A Semantic Framework for Neuro-Symbolic Computing

The field of neuro-symbolic AI aims to benefit from the combination of neural networks and symbolic systems. A cornerstone of the field is the translation or encoding of symbolic knowledge into...

arxiv.org
7
1
18
1,852
Many congratulations to Dr @benwagner. Great PhD viva earlier this week and very nice thesis on reasoning about what has been learned! #NeSyAI
3
1
8
Happy to introduce you to our latest work on taking a neural-symbolic approach to explainable AI with @AvilaGarcez published at NeurIPS 2021 Workshop on Human and Machine Decisions, arxiv.org/pdf/2112.11805.pdf
Image from Ben Wagner (@BenWagner)
5
9
22
more replies
In addition, the concepts may be used to impose new constraints onto the neural network, such as: the model should only recognise all horse-like objects with stripes as being zebras as long as the stripes are black and white.
1
1
The desired model behaviour will be reflected in the updated model parameters. This is accomplished with the use of Logic Tensor Networks as an approach for deep learning from data and knowledge in fully differentiable fuzzy logic.
1
Happy to announce that our work (w/ @AvilaGarcez) on “Neural-Symbolic Integration for Fairness in AI” has been published at the AAAI 2021 Spring Symposium, AAAI-MAKE 2021: proceedings.aaai-make.info/A…
Image from Ben Wagner (@BenWagner)
2
6
1
18
more replies
The Neural-Symbolic integration allows to not only better understand the decision-making process but also to intervene and act on extracted knowledge. We use First-Order-Logic constraints to adress these unwanted biases while retaining high predictive performance.
Image from Ben Wagner (@BenWagner)
1
1
5
We compare against fairness-based methods on 2 common notions of fairness across 3 three datasets. While the appropriateness of such notions are being discussed, the approach is adaptable and model-agnostic as LTN serve as a framework to complement explainability techniques
Image from Ben Wagner (@BenWagner)
1
4
Ben Wagner retweeted
Luis Lamb (@luislamb) profile picture
Luis Lamb@luislamb11 Dec 2020
Neurosymbolic AI: The 3rd Wave with @AvilaGarcez now on @arxiv arxiv.org/abs/2012.05876; thanks to @frossi_t @GaryMarcus @pascalhitzler @geoffreyhinton @ceobillionaire @kahneman_daniel @vardi @serafiniluc @pascalhitzler @Melleo54Sis @jeublanc @kerstingAIML & many others. – at Porto Alegre, Brazil
Image from Luis Lamb (@luislamb)
4
23
3
99
Neural-symbolic learning and reasoning NeSy2020, Athens 9-11 Sep 2020, provisional program and call for papers out: sites.google.com/view/nesy20…
7
1
17
Excited that NeSy2020 will join forces with StarAI, ILP and AAIP google.com/url?q=http://lr20…

IJCLR 2020-21 · International Joint Conference on Learning & Reasoning

In the face of the major disruptions caused by the Covid-19 pandemic, and in order to avoid the risks of having a face-to-face conference in September, as originally planned, IJCLR’s steering...

lr2020.iit.demokritos.gr
3
8
Load more