In this Plutopia podcast episode, futurist and data analyst Anne Boysen argues that today’s AI systems, especially large language models and emerging AI agents, are being adopted far faster than their reliability, transparency, and testability justify. She contrasts older, more deterministic technologies such as traditional search and rule-based systems with today’s probabilistic models, which generate plausible answers without clear provenance, reproducibility, or dependable truth-testing, making them vulnerable to hallucinations, disinformation, and misuse. Anne warns that handing decisions over to AI agents could amplify these risks, especially when users misunderstand AI as precise or authoritative, while also noting that companies often push AI into products out of hype, monetization pressure, or fear of missing out rather than clear user need. At the same time, she acknowledges that narrower, well-guarded uses of AI, such as media enhancement or limited decision support, can be helpful, and she ultimately advocates for careful testing, human oversight, targeted applications, and simple, thoughtful regulation focused on guardrails and accountability rather than blanket overregulation.

Anne Boysen:

We’re going to start leaving these decisions to agents. AI agents. So on top of all of this probabilistic hodgepodge of maybe truths, and maybe not reproducible truths on top of that, we’re going to start letting agents make decisions for us. So, you’re basically just going to use this interface that may or may not understand you completely and may come up with their own interpretations, and they’re like, “Oh, I thought you said enter my bank account to buy Bitcoin.” I don’t know, like, “That’s what I thought you wanted to do.” And then that could be the result. So, that’s where we are.

Video on YouTube

Anne Boysen: AI Hype, Agents, and Risk