Suggestions
Ross Harper
Founder & CEO at Limbic
Ross Harper is a computational neuroscientist and the CEO of Limbic, known for using probabilistic models to understand the brain. He possesses a diverse skill set, being an experimentalist, theoretician, entrepreneur, communicator, and self-proclaimed loverer of suffixes.
His educational background includes a Doctor of Philosophy (PhD) in Computational Neuroscience from UCL, a Master of Research (MRes) in Mathematical Modelling in Biology from UCL, and a Master of Arts (MA) in Natural Sciences (Neuroscience) from the University of Cambridge. He also studied at Colfe's School.
Ross has been involved in various organizations, having previously served as the Founder & CEO of Limbic, an EF9 Cohort Member at Entrepreneur First, Bogue Research Fellowship at Emory University, Deputy Editor/Director at Guru Magazine Ltd., Founder of Wriggle Ltd, and Founder of BuyMyFace.com.
Highlights
Today, I met a cat called Larry 🐈
Honoured to have been invited to 10 Downing Street today to discuss the future of #AI in the UK. Sometimes I have to stop and appreciate the journey. Limbic has achieved something truly groundbreaking as the only mental health chatbot in the world to secure Class 2 medical device certification. Our deep clinical evidence base is setting new standards, and I’m proud to represent our progress at the highest levels.
The conversation was under Chatham House rules, so below are my personal views:
🇬🇧 The UK is the perfect place for AI in healthcare. Now is the moment to start a company. The UK government must continue encouraging entrepreneurship through strategic funding and developing the startup ecosystem. 🏥 The #NHS is special. Contrary to popular belief, the NHS is one of the most innovative healthcare systems in the world, driven by passionate individuals committed to patient outcomes—don’t underestimate its potential. 🤖 The UK’s AI talent is unmatched. This is our secret weapon in becoming a global AI leader (I can personally vouch for the quality at UCL and Cambridge, but I’m told Oxford isn’t bad). 🔑 Adoption is key. The government should aim to incentivise AI adoption across sectors. Safety must be prioritised, but we cannot forget that progress requires courage. Large-scale AI adoption will drive efficiencies, raise GDP, and stimulate substantial economic growth.
I’m excited about the future we’re building—one where the UK leads the world in AI innovation. Grateful for all the incredible people I get to work with on this important mission.
Sebastiaan de Vries Entrepreneur First Matt Clifford Everyturn Mental Health Bruce Armstrong Khosla Ventures Startup Coalition NHS
Ever made a photocopy of a photocopy?
Yarin Gal and his team at the University of Oxford have shared thought-provoking research on the recursive impact of training generative AI models on web data, which increasingly includes content generated by previous versions of those same models. The result? “Model Collapse.” The modelled probability distribution diverges from the true data distribution.
This is a crucial consideration for the future of generative AI.
- Highlights the necessity of preserving original datasets that accurately reflect the system of interest.
- Stresses the importance of developing models that can reliably detect and exclude AI-generated content from training datasets.
- Suggests a first-mover advantage, as later models may struggle with training due to the contamination of web data with AI-generated content.
It’s great to see our top scientists tackling these important issues early on. Yarin’s work has personally influenced me before - his paper, “Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning", inspired Limbic's paper "A Bayesian Deep Learning Framework for End-To-End Prediction of Emotion From Heartbeat". It's an exciting time to be alive.
Nature paper: https://lnkd.in/epz79rKY
[Links to the other papers in the comments 👇]
