Few thinkers have shaped the global debate on artificial intelligence as deeply as Gary Marcus.
A cognitive scientist by training, Prof. Marcus bridges neuroscience, psychology, and machine learning — and has become the world’s most articulate advocate for real intelligence over hype.
He has advised governments and international organizations on AI policy, testified before the U.S. Senate, and consistently challenged Silicon Valley’s tendency to oversell generative AI.
Through his writing and research, he has emerged as the ethical compass and skeptic of the AI age — calling for systems that genuinely understand the world rather than simply predict the next word.
Article Zero:
In your recent Substack writing (“GPT-5: Overdue, overhyped and underwhelming. And that’s not the worst of it”), you emphasize that LLMs are overhyped, technically inadequate, and lack meaningful alignment with human values. We would like to ask you the following questions:
If ChatGPT is a public experiment rather than a real product, what does that say about the maturity of the AI industry itself, are we witnessing genuine progress, or just the world’s biggest beta test dressed up as innovation?
Gary Marcus:
All of the above.
ChatGPT is a real product, but in many ways not a fully mature product (and I am not sure it ever will be). ChatGPT will answer any question, but not all its answers are correct, and many that aren’t correct are presented as gospel truth when they aren’t.
In many ways ChatGPT is arguably the biggest beta test in history. Consumers shouldn’t really rely on it, and it comes with a lot of risks that haven’t been ironed out.
I have yet to see a clear, cogent argument the benefits outweigh the costs. Yet the beta test continues, at very large scale.
Article Zero:
Given the repeated cycle of overpromising and underdelivering in generative AI, what concrete steps would you suggest to rebuild public trust both with everyday users and policymakers?
Gary Marcus:
The public and policymakers need to realize that they are being played by big businesses with huge vested interests who aren’t always completely candid about the downsides of their tech.
Interview by Article Zero, 16 September 2025
brAuthor%20of%20Rebooting%20AI%20and%20Kluge){kind=link}