The Alan Turing Institute
Research institute in London, Englandwww.turing.ac.uk
33
MENTIONS
25
EPISODES
23
PODCASTS
Search complete. 33 mentions across 25 episodes found for "The Alan Turing Institute".
Oct 29, 2026
Shayda Kashef-TomlinsHOST
Can you talk to us about a project specifically that you've done with children and young people?
Mhairi AitkenGUEST
But previously, I was at the Alan Turing Institute.
Mhairi AitkenGUEST
And for five years, I worked on a collaboration with the Children's Parliament, a charity based in Scotland, and the Scottish AI Alliance.
Mhairi AitkenGUEST
And so that was a project over five years where we were engaging our primary school children in schools across Scotland, children between the ages of seven and 12, and looking at their experiences with AI, their views on AI, and particularly looking at how AI impacts children's rights.
Johnny GwynnHOST
Not that we haven't hurt ourselves over the last, I don't know, 50 years of voting.
Johnny GwynnHOST
I mean, what percentage of people actually vote anyway? Well, after 2024, Britain's Alan Turing Institute found AI disinformation had no measurable impact on the U.S.
Johnny GwynnHOST
results.
Johnny GwynnHOST
And Alabama already made election deepfakes a crime in the law that passed without a single no vote there, Scott.
Andy LambertGUEST
But it's all such like an individual way of working.
Andy LambertGUEST
Whereas an organization needs to be able to make visible, how capable is my organization? Can they use it safely, responsibly? Are they good at being able to shape a clear ask? Do they know what tasks to apply AI to? Do they know how to be clear about specifying what good looks like and being able to evaluate the outputs? All of those, you know, there's a million tangents we could go on there, but essentially at the core of Academy AI lives like a six axis framework, which is built in line with research from the Turing Institute.
Andy LambertGUEST
These six axes are kind of six different layers of what it means to be truly capable with AI.
Andy LambertGUEST
Nothing about, you know, what model do you use or like how do you prompt or whatever? All of that is like secondary stuff.
Mihaela van der SchaarGUEST
So I completely sympathize with this question.
Mihaela van der SchaarGUEST
So a few years back, actually, we worked exactly on this topic here at Alan Turing Institute, and we developed... a technology which is synthetic data with privacy preservation.
Mihaela van der SchaarGUEST
So we built AI technology where we have a guarantee that a particular individual will not be identifiable in this data set.
Mihaela van der SchaarGUEST
And yet this data is hopefully of good enough quality that can be used to train models that are indeed representative and correct AI is at times a technology that is very powerful.
Agnieszka PiotrowskaGUEST
So they, they are trying, and certainly Anthropic has been, um, doing this.
Agnieszka PiotrowskaGUEST
But a friend of mine who is a computer scientist, I've actually posted about him, um, on LinkedIn today, and he is so affiliated with the Alan Turing Institute and Cambridge, and this is, uh, Dr. Qianxi Dong.
Agnieszka PiotrowskaGUEST
So he's telling me that, um, a large language model, because of its design, it c- cannot quite do this reasoning that you and I would want it to do.
speaker_0HOST
Yeah.
Rory McGowanHOST
Trusted news sources and outlets do not publish deepfaked videos, and checking the account an image or video has come from can show you if that account can be trusted, or if they've put out other similarly edited videos in the past.
Rory McGowanHOST
The Alan Turing Institute, who works to limit the harms of deepfakes online, has a number of tales to look for as well.
Rory McGowanHOST
None of these are guaranteed to be present in each deepfake, but speech that doesn't match how someone is talking, unnatural facial features, strange differences in noise patterns, or colour differences between edited and unedited portions can be a clear tell if something has been edited to look a certain way.
Rory McGowanHOST
As AI gets better at ironing out the wrinkles and tells of deepfaked images and videos, it's going to get harder to understand what exactly is real and what's fake online.
Michael PeelGUEST
But there are alternatives to grabbing the most expensive frontier models.
Michael PeelGUEST
And I did an interview recently with the the head of the Alan Turing Institute in the UK, which was set up just over 10 years ago and has a leading role in setting AI policy.
Michael PeelGUEST
He talked about, well, for a lot of purposes in, for example, public surfaces, the frontier models will be very extraordinarily expensive and there will be cheaper models.
Michael PeelGUEST
which you can use, which are more bespoke, more narrowly written, might actually be better for the purposes you're looking for, and also would have the advantage that they are maybe things that you could develop without necessarily being right at the cutting edge of the most advanced companies.
Armin RazaviHOST
She's a contributing writer for Phi AI and the product manager for Responsible AI with Trilateral Research.
Armin RazaviHOST
She has worked at the Alan Turing Institute, but started a career in genetics, neuroscience, and data science, science.
Armin RazaviHOST
Good afternoon and hello, Mishka.
Armin RazaviHOST
How are you?
Mishka NemesGUEST
Um, and in the neuroscience realm, um, I was very lucky to be at neuro- at UCL in, uh, in Russell Square, which is a kind of the, the hub of neuroscience, um, at least in the UK.
Mishka NemesGUEST
Um, and there I learned a lot about decision-making, about metacognition, about, uh, kind of theory of mind, all things that we care now about in AI, but back then were kind of fascinating at the, at the, um, at the human level and cognition level.
Mishka NemesGUEST
Um, and then I guess my segue into, um, AI was, um, when I had the privilege to start a job with the Alan Turing Institute kind of in the early days.
Mishka NemesGUEST
Um, and I pretty much set up the, the skills work that the, um, the institute was doing.
Alex HudsonHOST
And if you link- look at the AI announcements from MI5, so 2017, they were sort of partner...
Alex HudsonHOST
Well, The Turing Institute was specifically working on AI, and then MI5 was very quick to partner.
Alex HudsonHOST
Uh, in 2021, GCHQ was operating the AI initiative.
Alex HudsonHOST
The AISI, the AI Security Institute, has been operating since last year.
15 more episodes mention The Alan Turing Institute.
Create an account to see the whole feed, search across every transcript, and follow the entities you care about.