AAAI's New Podcast Asks: Does Your Generation Shape How You See AI?

So far, the podcast has featured conversations with:

• Professor Marynel Vasquez on human-robot interactions and social robotics
• Professor Manuela Veloso on multi-agent systems and human-AI interaction
• Professor Roberto Martín-Martín on embodied AI, robotics, perception, and action

These aren't random topics. They're all areas where generational perspective might genuinely matter. Someone who started working on social robotics in the 1990s has seen multiple hype cycles come and go. They've watched promises get made and broken. A researcher who entered the field post-2020 is working in an environment where large language models have suddenly made certain interactions possible that seemed decades away.

You might be wondering whether this actually produces different research approaches or just different vibes in conversation. Tbh, I'm not sure we have enough data yet to say definitively. But the anecdotal evidence is suggestive.

The embodied AI episode is particularly relevant here. Professor Martín-Martín's work sits at the intersection of perception and action, the stuff that makes robots actually do things in the physical world rather than just process information. This is an area where the gap between generations might be especially pronounced. The deep learning revolution happened relatively recently, and researchers who came up before it often have different intuitions about what's hard and what's easy.

I should know this better, but I'm genuinely uncertain how much the "AI winter" generations carry that skepticism forward versus how much they've updated their priors. The podcast seems designed to surface exactly these kinds of tensions.

What's missing from the early episodes (or at least from the descriptions available) is conversation with people outside academia. Industry practitioners, people deploying robots in warehouses or hospitals, folks working on the policy side. AAAI says the series aims to feature "unique perspectives" across generations, so maybe that's coming. It remains unclear whether they'll expand beyond the academic research community.

The multi-agent systems conversation with Professor Veloso touches on human-AI interaction, which is interesting because multi-agent thinking has evolved dramatically. Early work in this area often assumed relatively simple agents with clear objectives. Now we're talking about systems where the agents themselves might be large language models with emergent behaviors that surprise their creators. How do you even apply lessons from 1990s multi-agent research to that context? I don't think anyone fully knows yet.

A few observations on what this format does well:

• Forces explicit discussion of assumptions that usually stay implicit
• Creates space for disagreement that isn't adversarial
• Surfaces historical context that gets lost in paper abstracts
• Acknowledges that expertise isn't just about knowing more, it's about knowing differently

And some limitations:

• Podcast format means no visual demonstrations or technical deep dives
• Self-selected participants probably skew toward people who enjoy this kind of reflection
• We're getting curated conversations, not raw disagreements

The social robotics episode with Professor Vasquez is probably the one I'm most curious about, because social robotics has had such a weird trajectory. There was enormous hype around companion robots and social interaction in the 2000s and early 2010s, then a kind of disillusionment, and now renewed interest driven by eldercare applications and the loneliness epidemic. Someone who lived through all of that has a different relationship to the work than someone who entered during the current wave.

I think the underlying question here is actually pretty important for anyone following robotics and AI. We tend to treat technical progress as objective, measurable, separate from the people doing the measuring. But research directions are chosen by humans with particular histories. Funding priorities reflect generational anxieties. The problems we consider "solved" versus "open" depend partly on what we remember struggling with.

This doesn't mean everything is subjective or that technical benchmarks don't matter. It means the story we tell about where the field is going is always partly a story about who's telling it.

AAI hasn't announced how many episodes the series will run or whether it's an ongoing project. Based on the three episodes released so far, it seems like they're releasing them periodically rather than on a strict schedule. If you're interested in the meta-level questions about how AI research happens (not just what the research finds) it's worth adding to your podcast rotation.

Whether this kind of intergenerational dialogue actually changes anything about how research gets done, well, that's a harder question. I'm hopeful but honestly not certain. Sometimes these conversations stay in the realm of interesting reflection without translating to different choices in the lab. But at minimum, it's useful to hear smart people acknowledge that their perspective isn't the only valid one. That's rarer than it should be.

Sarah Williams · 4 hours ago · 5 min