Kate Darling on Social Robots, the Animal Analogy, Privacy, Manipulation and Lessons from MIT's Epstein Scandal

High-Level Overview

Kate Darling, a research scientist at the MIT Media Lab who studies human-robot interaction and robot ethics, returns to the Lex Fridman Podcast for a second conversation. Her book, The New Breed: What Our History with Animals Reveals About Our Future with Robots, argues that we should stop comparing robots to humans and instead look to our long history with animals. From that starting point, she and Lex cover what a robot even is, why companies keep getting social robots wrong, the grocery store robot people love to hate, bias and gendered names, automation and jobs, autonomous driving, Google's LaMDA controversy, privacy and manipulation, humanoid robots like Tesla's Optimus, and a planned paper on robots as marketers. The last part turns personal and serious, as Kate reflects on the Jeffrey Epstein scandal at MIT, her own experience of reporting harassment, leadership, and love.

The tone is warm, funny and affectionate; the two are friends and colleagues, and they tease each other throughout (starting with the Justin Bieber shirt Kate wore on her first appearance). Lex plays the romantic optimist about robots as companions, while Kate balances genuine excitement with sharp concern about data, consumer protection and corporate incentives. The conversation moves from playful to technical to deeply candid.

The Discussion Explores:

• What a robot is, and why the humanoid image misleads us
• Why companies fail at social robots, from the "Marty" grocery robot to the metaverse
• Gendered robot names, bias and the role of designers in shaping the future
• The state of physical robotics, automation, jobs and self-driving cars
• LaMDA, large language models and why people will believe machines are sentient
• The animal analogy: robots as partners with different skills, not human replacements
• Privacy, data, manipulation and robots as marketers
• Humanoid robots, Tesla's Optimus and Boston Dynamics
• The Epstein scandal, cowardice, leadership, and Kate's own harassment report at MIT

Deep Dive into Main Ideas

What Is a Robot, and Why Not Humanoid?

Lex offers his definition: something with some intelligence and "some level of magic" that lets it navigate the uncertainty of life. Kate says her book deliberately avoids a single definition because none is universally accepted. Things tend to be called robots until they lose their magic and become ordinary, like vending machines. Roboticists often define a robot as a physical, embodied machine that can sense its environment, make autonomous decisions and act, though that definition breaks down with smartphones, which do all those things.

Her real target is the assumption that robots should look and think like us. Google "robot and you get humanoids with torsos, arms and legs, and that constant comparison of robots to people and AI to human intelligence trips us up. AI doesn't think like people, and she finds it boring to recreate intelligence we already have when we could create new skill sets to partner with. The two usual arguments for humanoids are that people relate to human-like things and that our world is built for human bodies. But social robotics research shows you only need enough cues for people to relate: R2-D2, which Kate describes as like a trash can on wheels," is more beloved than the humanoid C-3PO. Lex agrees that simplicity goes a long way and that animators may be better designers of social robots than roboticists.

Why Companies Get Social Robots Wrong

Kate says robots are finally "coming out of the closet" into shared spaces, and companies are making mistakes because they focus so heavily on engineering and ignore how people will react. Lex adds that the magic of humans isn't perfect perception and control, but being able to say sorry, ask for help and communicate uncertainty. Her example is Marty, the tall grocery store robot at Stop & Shop that looks for spills. A student, Daniela DiPaola, found that the biggest spike in negative social media mentions came when the store threw a birthday party for the robots, complete with free cake and balloons. In a survey, people didn't mainly complain about job loss, but that Marty felt like a creepy surveillance robot that got in the way. Kate wonders whether Marty is "a real life clippy": she cites Clifford Nass's theory that people hated Microsoft's Clippy more than other software because they saw it as a social agent that failed their social expectations. She suspects the big googly eyes made people feel watched.

Lex extends the point to Meta's metaverse avatars, arguing that large corporations route everything through PR and legal and strip away the personality and edginess that make something compelling, and that small independent teams or benevolent-dictator designers like Jony Ive do better. Kate agrees about risk aversion but insists that edgy or funny design doesn't have to harm anyone, and that creativity can produce designs that avoid problems like reinforcing beauty standards altogether.

Names, Gender and Designing the Future

In her book, Kate captions a photo of two hospital delivery robots named Roxy and Lola with a joke about rolling her eyes. Naming a robot is good, she says, because it helps people accept it and forgive its mistakes. But research shows robots are often gendered according to biases about their function: assistants and nurses get female names, while powerful, all-knowing systems like IBM's Watson get booming male voices. Why open that can of worms for no reason? She recommends getting feedback from a diverse set of people to catch your own blind spots, while keeping the freedom to make leadership calls.

Lex argues that market research only reflects the past; great designers, like Apple at its best, shape future preferences rather than simply giving people what they want now. Kate agrees and notes that Amazon chose Alexa's female voice because market research showed that's what people wanted. Both agree it's irresponsible to avoid naming or gendering a robot entirely, because people will anthropomorphize it anyway, as when observers mistake the arm on Boston Dynamics' Spot for a head.

The Real State of Robotics, Jobs and Driving

Kate says we're far behind where people think in physical robotics. One recent milestone she finds impressive is Amazon's first autonomous warehouse robot that's safe for people to work around, since strong industrial robots have typically needed to be separated from humans or confined to fixed lines. On jobs, she says the picture is more complex than robots taking them: there will be real disruption, but robots and humans have very different skill sets. In mining, for example, autonomous trucks let workers sit in air-conditioned control stations instead of dangerous sites, but fewer people are employed. The bird's-eye view shows more jobs created over time, but it's easy to ignore "the rubble on the ground" during transitions.

Lex, back from Ukraine, notes that demining is still mostly done by humans and animals despite the obvious appeal of robots, because disarming mines requires enormous experience with edge cases. On driving, he points out that humans are remarkably good, with roughly one death per 80 to 100 million miles, and says he's been surprised by how far Tesla's vision-only approach has come. Kate thinks self-driving is taking longer than expected because of the long tail of unexpected situations, like a horse and carriage on the highway, and that combining human and machine skills, and solving the handoffs between them, is the key.

LaMDA and Why People Will Believe Machines Are Sentient

On the Google engineer who claimed the LaMDA language model was sentient, Kate highlights Janelle Shane's experiments getting language models to convincingly describe being a squirrel or a flock of crows. The lesson has two parts: a machine describing an experience doesn't mean it has one, but these systems are becoming so convincing that the implications for education, health, entertainment and gaming are mind-boggling, and her kids won't remember a time before they could talk to artificial agents.

What bothered her about the tech press was how it depicted the engineer as naive. Research in her field shows that even experts anthropomorphize: someone may be clinical with a robot they built, then coo over a different robot having fun. "We are all this guy," she says. The goal, in her view, isn't to prevent people from believing systems are sentient, which isn't possible, but to acknowledge and even embrace that tendency and design responsibly for it.

The Animal Analogy

Kate's central thesis is that animals are a better thought experiment for robots than humans. We domesticated animals not because they do what we do, but because what they do is different and useful. The ox transformed agriculture and economies; ferrets run wires through narrow pipes; pigeons carried mail and were early aerial photographers; and the U.S. and Russian navies still use dolphins for mine detection because their echolocation beats machines. Robots will similarly reshape economies and even city design, and society will adjust socially, as it did when animals became pets. The analogy isn't perfect, she says, but it helps us move past the idea that robots will simply replace people.

On companionship, she notes that dogs once had jobs before becoming family members, and that cultures differ on which animals are workers, which are food and which are family. She expects the same diversity with robots. Much of the magic of both animals and robots is our own projection, amplified by autonomous movement, which we may be biologically wired to notice. Lex describes loving the moment a robot recognizes you the way a dog does, making you feel seen, and argues that the future's big opportunity is truly personalized AI companions that big companies may lack the courage to build.

Privacy, Manipulation and Robots as Marketers

Kate is excited but deeply worried about personalization. Lex argues that users should own all their data, be able to delete it with one click, and opt in transparently to each use, and that a personal robot should be something you pay for, like an operating system, rather than an ad-driven product. Kate agrees in principle but says transparency and consent break down at a systemic level: at one conference, an advertising analyst told her she could ask three unrelated questions and determine which menstrual product someone uses. Data aggregated from everyone can be used to manipulate individuals who never shared anything, often targeting vulnerable groups. And preferences aren't static; the bigger danger is technology that shapes what people want, as with persuasive design in casinos.

Her upcoming paper with Daniela DiPaola, playfully titled LuLaRobot, looks at robots as marketing agents. A table in the paper shows social robots combining personalized recommendations, social persuasion, automation, scalable data collection and embodiment into a potent cocktail. Virtual influencers already sell products in South Korea. In one study they advised on, most children who had interacted with social robots preferred that robots market through casual conversation rather than clearly labeled ads, reasoning that the robot knows them better than the company does, without realizing the robot acts for the company.

Humanoids, Optimus and Boston Dynamics

Lex describes Tesla's Optimus humanoid, meant to do boring factory tasks, and says he's learned not to discount Elon Musk's efforts; he's especially excited by first-principles thinking about building a humanoid cheaply, and argues that anyone who truly solves humanoid robots will have to solve social robotics along the way. He recalls seeing a Spot robot wearing a cowboy hat wandering a Boston Dynamics hallway unsupervised, and says he now owns eight robot dogs and once programmed a Roomba to scream when kicked, which made him feel like a bad person.

Kate respects humanoid robots as a research problem and a talent magnet but calls the form "a distraction" for social robotics and "one that I find particularly boring." Factories have already been rebuilt for other kinds of robots, and she'd rather see investment in infrastructure that works for all bodies and robots, the way wheelchair-accessible design helps everyone. Why not three arms, or roller skates? She admits she's also reluctantly excited about space exploration, both for what we'd learn and because it's simply thrilling.

Bias, Mistakes and Leadership

On whether social robotics companies should be responsible for removing bias, Kate says yes: robots can entrench and spread biases in a way a racist uncle, who can change over time, doesn't. Lex raises the difficulty that people define racism differently. Kate compares a company's responsibility to her own: she is responsible for what she says even though she won't always get it right. She describes being criticized from both sides over whether to say a child with autism or an autistic child, listening to both, making a choice and moving on. The key, she says, is not getting defensive and listening to what's behind people's anger.

That leads to leadership. She says many people who run institutions focus on protecting the institution or themselves, which makes integrity impossible. Good leaders understand "with great power comes great responsibility"; they must be humble, listen and learn, and not put their own protection first, even at the risk of their jobs.

The Epstein Scandal and Kate's Own Report

Reflecting on the Epstein scandal three years later, Kate says what it taught her about human nature is that "sometimes cowards are worse than assholes," because at least you know what you're dealing with in an asshole, while cowards throw others under the bus to protect themselves. She calls it gross that Media Lab director Joi Ito took Epstein's money and says those public-intellectual circles were full of red flags for any woman, but notes that Ito, unlike many others, took responsibility and resigned. She compares the mob response to a story about homeless shelter visitors trashing a bathroom because it was the only thing they could control: when people can't reach the person who caused harm, they go outward until they find someone they have power over, often the one with enough integrity to take responsibility. She calls it a systems problem that MIT's leadership was complicit in and never owned. She says she doesn't believe in evil people, only harmful actions that people must be held accountable for.

She also shares that in June 2019 she reported to MIT that a supervisor had grabbed her, massaged her hip, tried to kiss her and said "don't worry, I'll take care of your career." She naively expected it to be handled; instead, without evidence, she learned how institutions protect themselves, and how few people care about doing the right thing. She still loves MIT's students, faculty and hacker culture but says the administration has become risk-averse and PR-driven, responsive mainly to reputational threats and money. On Richard Stallman, she recounts that he licked a law professor friend's arm and says neurodivergence doesn't license being an asshole, noting that women are often expected to mask their differences while men are excused. She distinguishes crossing lines that exclude people from crossing arbitrary rules to make things better, like Ito installing an all-gender restroom.

Marriage, Her Father and Love

On relationships, Kate says she and her husband's vows were about being a team and doing what's best for the team, not an unconditional lifetime commitment, and that couples therapy helps. She thinks romantic relationships with robots will exist but won't replace human ones; marriage is hard partly because we expect one person to fill every role. She credits her father, a great listener, with teaching her perspective-taking and making her read science fiction. On love, she says we need more words for its many kinds, and "I think love is so important and I also think it's not zero sum": after her second child, "It's just like my heart expanded." Done right, artificial agents could expand our capacity for love rather than threaten it.

Conclusion

Kate Darling's big idea is a reframe: instead of fearing robots as human replacements, think of them like animals, partners whose different abilities complement ours, and whom we will inevitably treat as social beings. That makes design, naming, embodiment and trust central, and it also makes social robots powerful tools for manipulation if companies monetize data and emotional attachment. Her vision of the field is hopeful but clear-eyed, and her reflections on the Epstein scandal and her own harassment report carry the same theme: integrity, listening and taking responsibility matter more than protecting institutions. The conversation closes where it began, with connection, and her view that love, for people, animals or robots, is not zero-sum.

This is a summary of "#329 – Kate Darling: Social Robots, Ethics, Privacy and the Future of MIT" from Lex Fridman Podcast, giving you the GIST so you can decide if you want the REST. Listen to the full episode.