Rockefeller Institute of Government’s Post

In episode 7 of the Max Planck Lawcast Christian Thönnes (MPI-Freiburg) hosts a round table discussion with Nandor Knust (Tromso) and Tatiana Tropina (Leiden). The discussion forms the second part of the special series ‘Artificial Intelligence in Crime Control’. Modern policing increasingly relies on the extensive use of personal data collected in large-scale databases which are rendered interoperable and automatically searchable through modern AI technologies. With this strategy, security authorities seek to become more effective and efficient, to the point that potential offenders may be automatically detected before a crime occurs. However, AI-powered predictive policing entails a range of concerns as it not only serves to radically expand the state’s powers of surveillance and coercion but can result in bias and opacity, thereby subverting well-established legal standards. You can listen to the conversation on Spotify: https://lnkd.in/eJNeHM29 On Apple Podcasts: https://lnkd.in/e-MM9Dny Or on our website: https://lnkd.in/ecNSM4Vr A written summary is available on SSRN: https://lnkd.in/eWjsyXHW

The UK is remarkable in how short a distance you need to travel before the local accent completely changes. A two month test by West Midlands police is highlighting how chat bots used for live interaction with callers are struggling to deal with the regional accents that, while they are English, are proving too far from the model used by AI #ai #aivoice #generativevoice #artificialintelligence #podcasts #podcast #podcasting #singapore #spotify #amazonmusic #audible #applepodcasts

This was a really engaging and informative podcast. But then again, when the interviewee is. Debbie Reynolds, what else would you expect? From a terrific explanation of the distinction between “data, providence“ vs “data, lineage”, to an understanding that when it comes to large language models, you can’t really expect to “delete” data. The best you can hope to do is to code something in order to “suppress it“. And even then, good luck, achieving that anywhere close to 100% of the time. And so while it’s always a good idea to avoid inputting anything you don’t want these LLM‘s to have to “forget”, the models inferencing capabilities make even doing THAT difficult. Just an overall very practical discussion.

Have you heard? Season 2 of Exterro's podcast, FTK Over the Air has officially launched! Check out this blog post to get a summary of episode 1 and some sneak peaks into what you can expect this season: https://lnkd.in/ggZFyUaW

Have you heard? Season 2 of Exterro's podcast, FTK Over the Air has officially launched! Check out this blog post to get a summary of episode 1 and some sneak peaks into what you can expect this season: https://lnkd.in/g86eRkDt

Dive into the world of deepfakes with Onfido's latest episode of 'Inside The fraud Lab' podcast. Tune in today to discover how AI is changing the landscape, the challenges presented by deepfakes, and how to take a black mirror approach. 

In case you missed it, we chat to Matt from Concept Smoke Screen regarding security trends, AI and how retail is being effected with the increase in retail theft!

The Orthus.AI (sister product to LEAP/CORE, focused on b2b) podcast "Digital Shadows" featuring Eric Oldenburg is well worth a listen for anyone interested in how #AI can help your investigations into crimes involving #CSAM. Click the link in Felix's post below and let us know what you think! #fightingcrimesagainstchildren #digitalinvestigations #aiforgood

Dive into the world of deepfakes with Onfido's latest episode of 'Inside The fraud Lab' podcast. Tune in today to discover how AI is changing the landscape, the challenges presented by deepfakes, and how to take a black mirror approach. 

Dive into the world of deepfakes with Onfido's latest episode of 'Inside The fraud Lab' podcast. Tune in today to discover how AI is changing the landscape, the challenges presented by deepfakes, and how to take a black mirror approach.