Giant Robots Smashing Into Other Giant Robots

557: AI from Automation to Transformation with Andrew Rashbass

·38 min·3 clips

Andrew Rashbass studied AI 35 years ago, then became a COBOL programmer before leading major media companies and now founding an AI startup.

The episode begins with host Sammy Birnbaum introducing Andrew Rashbass, highlighting his roles as former CEO of the Economist Group, Reuters, and Euromoney, and his current work with the Van der Schaar Lab at Cambridge University and as founder of Sculpture AI. Rashbass recounts his academic background in Greek, Latin, and philosophy, and how he studied AI 35 years ago through a government program aimed at making 'useless' graduates useful. He then worked as a COBOL programmer, rising to IT director roles before joining the Economist Group in the late 1990s. At the Economist, he led the development of its online strategy, recalling early internet experiences like witnessing real-time reports from Tiananmen Square in 1989 via academic networks. Rashbass describes his career progression to CEO of the Economist Group and later roles at Reuters and Euromoney, before returning to AI with Sculpture AI and the Van der Schaar Lab. He explains the lab's focus on 'reality-centric AI,' which addresses challenges in high-stakes domains like healthcare where data is limited and models must work in real-world conditions. The host raises concerns about AI hallucinations in forecasting and data querying, but Rashbass distinguishes between hallucinations in factual reporting versus probabilistic forecasts. He discusses techniques like interpretability and prompt engineering to improve reliability, but argues that current uses of LLMs for tasks like translating English to SQL are too limited. Rashbass envisions more ambitious applications, such as AI systems that answer causal questions like why sales dropped or which customers might buy new products. The conversation shifts to AI's broader societal impact, with Rashbass comparing it to the internet but predicting it will be an order of magnitude bigger. He outlines a progression from automation to transformation, where new entrants disrupt incumbents by doing things previously impossible. When asked if AI will make society better, Rashbass acknowledges dystopian risks but leans optimistic, citing examples like AI enhancing consultants' capabilities and acting as an 'exoskeleton' for human potential. He references historical fears about technologies like writing and calculators, suggesting AI could similarly expand access and creativity. The episode concludes with Rashbass sharing contact details via Sculpture AI and the host providing show notes.

As heard by us

A practical AI conversation that stays with the hard questions about trust and interpretation.

Andrew Rashbass, who has led the Economist Group, Reuters and Euromoney Institutional Investor PLC and now works with the Van de Schaar Lab at Cambridge University, anchors a conversation that stays alert to what AI can and cannot be trusted to do.

Read the full review in PlayNext →

Why you'd press play

You get a clear AI talk about why hallucinations matter differently for facts, forecasts, and data queries.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Giant Robots Smashing Into Other Giant Robots