The experts in AI – why they are fearful and what this means for surveying

Who the speakers are, and why they matter

These are not commentators. They are foundational figures.

Geoffrey Hinton

A cognitive psychologist and computer scientist, often referred to as one of the “godfathers of AI”.

– Pioneer of deep learning and neural networks – Formerly at Google – Winner of the 2024 Nobel Prize in Physics for foundational work in machine learning – Left big tech so he could speak more openly about AI risk

Yoshua Bengio

One of the most influential AI scientists in the world.

– Professor at the Université de Montréal – Turing Award winner – Regularly ranked among the most cited scientists globally across all disciplines – Leading voice on AI safety and governance

Stuart Russell

A long-standing authority on AI alignment and risk.

– Professor of Computer Science at UC Berkeley – Co-author of Artificial Intelligence: A Modern Approach, one of the most widely used AI textbooks – Author of Human Compatible

Yann LeCun

An important counter-voice in the debate.

– Turing Award winner – Pioneer of convolutional neural networks – Former Chief AI Scientist at Meta – Strong critic of AI “doomerism” and exaggerated timelines

What stood out

The people who built this are uneasy
This was the most sobering part. Several of these individuals helped create the foundations of modern AI, and now speak with visible concern.

It reminded me of Oppenheimer, because we’re racing ahead with something powerful without fully understanding its long-term consequences. The fact that there is more than a 1% chance of human extinction, with some saying it is as high as 20% yet we are not stopping or slowing down.

Speed is outpacing understanding

The technology is moving faster than regulation, faster than professional guidance, and faster than most organisations can safely absorb.

Surveying already lives with this tension. AI widens the gap.

Confidence is not correctness

AI systems can sound fluent, persuasive, and certain while still being wrong.

For a profession built on evidence, judgement, and accountability, that distinction matters.

Language models do not understand the physical world

A repeated technical point, made in different ways.

Today’s systems are excellent with language. They are far less reliable when it comes to physical reality, materials, buildings, and real-world cause and effect.

That has direct relevance for areas such defects, pathology, retrofit, sustainability, construction and risk.

Incentives cause more harm than intelligence

The danger is often not “superintelligent AI”.

It’s incentives. Speed. Scale. Market pressure. Cost-cutting. Power and of course £££££££

Safety cannot be bolted on afterwards

Guardrails, prompts, and disclaimers are not enough.

Responsibility has to be designed into systems and workflows from the start. This mirrors professional practice. You don’t add diligence at the end.

Power is concentrating

AI capability is increasingly held by a small number of organisations with vast resources.

That raises questions about dependency, cost, liability, and control for professions that may come to rely on tools they do not own or fully understand. I also see many firms building their own AI solutions – silos, lack of consistency, expertise and standards.

What I think we should all be doing right now

This applies whether you run a firm, work for one, or are simply trying to stay informed.

– Keep learning, deliberately Read long-form pieces. Watch full interviews. Sit with disagreement. Be wary of those who profess to be experts. Check their credentials.

– Use the tools, carefully The fastest way to understand AI is to use it. See where it helps, where it misleads, and where it confidently gets things wrong.

– Question outputs, not just inputs Ask where information comes from, what assumptions are baked in, and what still needs professional verification.

– Stay grounded in responsibility AI doesn’t remove accountability. Reports, advice, and decisions still sit with people.

– Talk about it openly Within teams and peer groups. Silence creates fear. Conversation builds literacy.

These are just some of the things I’m learning.

I do have concerns, real ones. But I’m also a natural optimist, and a pragmatic one. I believe the best way through uncertainty is understanding, not avoidance. That’s why I listen to different viewpoints, especially when they disagree, and then form my own judgement.

What’s hard to dispute is the common ground across many of the world’s leading AI figures. People like Ilya Sutskever, co-founder of OpenAI, Demis Hassabis, CEO of Google DeepMind, and Dario Amodei, CEO of Anthropic, all raise similar concerns in different ways.

Different organisations. Different incentives. Different views.

Yet again and again, the same themes appear: fear, uncertainty, unknowns, and deep questions about speed, control, and direction.

At the same time, this is a global race. The US and China are not slowing down. No one really is. And yet many of the people closest to the technology are saying we should slow down.

Watch the interviews and decide for yourself

– Geoffrey Hinton interview – Yoshua Bengio interview – Stuart Russell interview – Yann LeCun interview

For professions like surveying, judgement still matters. And it’s going to matter even more.