Jon Carvill, VP of Marketing & Communications at MBZUAI, explains the shift from fluent AI to systems that can actually reason. The conversation highlights K2-Think V2’s open-source design, its advantages for science and industry, real-world use cases like drug discovery and climate modeling, and why transparency and sovereign AI matter for trustworthy decision-making.
21:35
forum Ask episode
web_stories AI Snips
view_agenda Chapters
auto_awesome Transcript
info_circle Episode notes
question_answer ANECDOTE
University Launched Frontier Models And Apps
MBZUAI combines academic research with practical product releases like K2 Think and a world model.
Jon Carvill recounts launching frontier models, mobile apps, and web apps from a university lab with open releases.
insights INSIGHT
Full Transparency Through True Open Source
Openness Means Releasing Model, Weights, Data, And Docs for full transparency.
Jon Carvill says MBZUAI publishes models, weights, datasets and technical docs so there is no black box and researchers can inspect and iterate.
volunteer_activism ADVICE
Apply Reasoning Models To Complex Engineering Problems
Use reasoning models for domain-specific heavy math or engineering tasks rather than just chat improvements.
Jon Carvill gives aerospace and oil and gas as examples where fine-tuned reasoning models can synthesize data and suggest optimal solutions.
Get the Snipd Podcast app to discover more snips from this episode
🧠 From Fluency To Thinking: Why Reasoning Models Are The Next Leap In AI
For years, AI progress has been measured by how convincingly machines can talk. The next phase is defined by how well they can think. This conversation explores why that shift matters and what it means for science, industry, and high-stakes decision-making. #MBZUAIPartnership#Sponsored
In this episode, I sit down with Jon Carvill, Vice President of Marketing and Communications at MBZUAI, to discuss the rise of reasoning models and why they represent a fundamental change in how we should judge AI capabilities.
🎯 WHAT WE COVER:
Why fluency alone doesn't guarantee understanding in high-stakes applications
The critical difference between predicting words and reasoning through problems
How MBZUAI's K2-Think was built for reasoning from the ground up, not retrofitted
What makes reasoning models essential for science, engineering, and medicine
Real-world applications: drug discovery, climate science, financial modelling, and manufacturing
Why openness matters: training data, methodologies, and full transparency
The strategic importance of sovereign AI development for trust and governance
How K2-Think V2 achieves strong performance with clear, explainable reasoning
The connection between reasoning models and world models for physical reality
What this means for the future: moving from conversation to reliable decision-making
Where to try K2-Think and access the full open-source release
🔬 Jon Carvill leads marketing and communications at the Mohamed bin Zayed University of Artificial Intelligence, home to the Institute for Foundational Models (IFM). Their work on K2-Think represents a critical step toward AI systems that can be inspected, trusted, and applied to complex real-world problems.
This conversation reveals why the next generation of AI will be judged by its ability to reason rather than talk, why transparency and sovereignty have become practical requirements for adoption, and how open reasoning models could shape the future of science and industry.
If you are curious to try it, K2 Think is accessible through the public web app at k2think.ai, while researchers can explore the full open releases, including data and training recipes, via IFM’s Hugging Face repositories.
Bernard Marr's Future of Business & Technology Podcast 