Darling, Kate
▶ Watch on YouTubeCore ideas
Darling opens by arguing that the common instinct to define and evaluate robots by comparison to humans is both technically inaccurate and practically limiting, since AI systems don't "think" like people and the most interesting opportunities come from building complementary skill sets rather than recreating human intelligence. Her central proposed reframe, developed at length in her book, is to think about robots the way humans have historically thought about domesticated animals — not as replacements for people but as different kinds of agents used for companionship, labor, sensing abilities humans lack (citing dolphins for underwater mine detection, ferrets for narrow-space wiring work, pigeons for early aerial photography and messaging), and, crucially, as entities whose social and emotional relationships with humans vary enormously by culture and individual disposition rather than by any single fixed rule — just as some cultures eat animals others keep as pets. She argues this analogy helps escape the false binary of "robots will replace jobs" versus "robots will just be tools," pointing to mining automation (where robotic trucks removed people from a dangerous task but also required new remote-operator jobs) as a template for how the labor picture is likely to actually evolve.
On design, Darling argues that humanoid form is generally unnecessary and often counterproductive for social robots — R2-D2 elicits more affection than the humanoid C-3PO — and that simple, well-designed cues (a name, a voice, subtle movement) are enough to trigger anthropomorphization, which she says happens whether or not designers intend it, making it irresponsible for engineers to avoid naming or gendering robots on the assumption that doing so sidesteps ethical responsibility for how people will relate to them. She and Lex discuss real-world design failures (the Stop & Shop "Marty" robot, whose oversized googly eyes and spill-detection function led shoppers to perceive it as a surveillance device rather than a helpful presence) and the broader pattern that large companies' risk-averse, PR- and legal-department-driven processes tend to strip out the creative, "edgy" personality choices that make social agents charming, a dynamic she compares to Microsoft's Clippy, which people resented specifically because it was seen as a social agent failing at social expectations rather than as an inert tool.
On AI and language models, Darling discusses the 2022 case of a Google engineer who believed the LaMDA chatbot was sentient, arguing the tech press's dismissive coverage of him as naive missed the more important point: human-robot interaction research shows even trained experts anthropomorphize systems they know are not sentient, so widespread emotional attribution to conversational AI should be treated as an inevitable, serious design and policy consideration rather than a fringe delusion to be corrected away. She raises significant concern about the privacy and manipulation risks of personalized, embodied conversational agents — proposing that such systems should be funded by direct payment (like an operating system) rather than advertising, that users should own and be able to fully delete their data, and that a trustworthy assistant should optimize for user well-being rather than being influenced by any promoted or sponsored content — while acknowledging that companies large enough to build this technology may lack the "guts" to commit to that model given the short-term business incentives toward data monetization. She also discusses subtle algorithmic bias (as opposed to overt bias, which she expects companies can more easily filter) as an especially difficult unsolved problem for future social and voice agents, and highlights research suggesting children can be confused about who a companion robot's marketing "actually" serves, often assuming a robot that knows them well is acting purely in their own interest rather than a company's.
On humanoid robots specifically (prompted by discussion of Tesla's Optimus and Boston Dynamics' Atlas), Darling is skeptical that the humanoid form factor is necessary outside a few edge cases, arguing that infrastructure built for humans could instead be redesigned for broader accessibility (comparing this to wheelchair-accessible design), and that she finds the pure engineering and cost-reduction ambitions behind such projects scientifically interesting but likely to run into unavoidable human-robot-interaction and social-perception challenges regardless of original intent. She reiterates a recurring theme — that meaningful social connection with a robot does not require arms, legs, or human-like appearance — while acknowledging the humanoid form's power to attract engineering talent and public fascination.
In the episode's final section, discussing MIT's institutional culture following the 2019 revelation that Media Lab director Joi Ito had accepted funding from Jeffrey Epstein, Darling states she believes Ito did not know about the criminal abuse and credits him, distinct from others in that same social circle, for publicly taking responsibility and resigning rather than obscuring his role — while stating plainly that accepting money despite what she describes as an environment full of "red flags" was a serious lapse. She is more critical of what she calls institutional "cowardice" more broadly: leaders and administrators who prioritize protecting the institution over transparency and accountability, drawing a parallel to displaced anger (illustrated via an anecdote about homeless-shelter bathroom vandalism) in which public criticism often lands hardest on whichever nearby figure has enough integrity to take responsibility, rather than on those most culpable. She separately discloses that she personally reported a supervisor for inappropriate physical contact and comments to MIT in 2019, and describes finding the institutional reporting process discouraging given the lack of corroborating evidence. Asked about Richard Stallman's remarks on an MIT mailing list connected to the Epstein discussion, she describes a separate incident in which, she says, a colleague told her Stallman had licked her arm after a handshake, and argues that being unusual or "neurodivergent" does not excuse behavior that makes colleagues feel unsafe, while distinguishing this from institutional rules that exist mainly to avoid controversy (citing Joi Ito's decision to install a gender-neutral bathroom against MIT bureaucratic objection as an example of a line worth crossing). Asked about Ghislaine Maxwell, she declines to comment on the specific facts of the case (saying she hasn't followed it closely) but states a general view that she does not believe in intrinsically "evil" people, only in accountability for harmful actions, and that leadership and institutions succeed when they prioritize transparency, humility, and listening over self-protection.
Excerpts
On why she considers the animal analogy more useful than the human-comparison framing common in robotics:
“Animals... we domesticated them not because they do what we do, but because what they do is different, and that's useful... there's just so many things we can draw on in that history from these entities that can sense, think, make autonomous decisions, and learn, that are applicable to how we should be thinking about robots and AI.”
On why she believes large companies systematically fail to build compelling, personality-driven social agents:
“I think what's really about it is... you get departments that look for harm in things, and so they will find harm in things that have no harm... I think the way to solve that in large organizations is to have Steve Jobs types of characters — a benevolent dictator who cuts through the lawyers and the PR but has a good heart.”
On what she says she learned about institutional behavior in the aftermath of the Epstein/MIT controversy:
“Sometimes cowards are worse than assholes... the assholes at least, you know what you're dealing with, they have integrity in a way, they're just living out their asshole values. And the cowards are the ones you have to watch out for... they'll throw others under the bus [to protect themselves or the institution].”