Jaron Lanier argues the AI industry has become addicted to addiction as a business model, repeating social media's mistake.
Lanier says AI filters humanity's creativity through companies like OpenAI and Anthropic, which do not pay the sources they draw from.
Lanier says humanity will get through the current AI period but not unscathed, and that Microsoft lets him speak his mind as a heretic.
Jaron Lanier says AI is repeating social media's addiction mistake
Jaron Lanier argues AI is repeating social media's mistake. Companies present models as fake people and treat them as a god, which hooks users and erases the human creativity the systems are built from. He would keep the technology but treat it as a tool that people steer and learn from.
Takeaways
AI has inherited social media's addiction business model. Lanier argues the industry became addicted to addiction, as social media did, and that paying for AI subscriptions resembles social media's emotional pull. He grants promised gains like disease cures and new materials, but says the business still profits by manipulating people.
Chatbots deliberately revive the 1960s ELIZA illusion. ELIZA simply echoed users back, yet people believed it was a conscious, empathic therapist. Lanier says the industry chose to reproduce that trick, presenting models as fake people that users become attached to and addicted to.
Generative AI hides the humans behind it. Models are trained on all human creation, including the internet and every book. Lanier calls them useful tools that, like a library, mean something only when people use them. OpenAI and Anthropic present the output as their own and do not pay sources.
Agents are mostly a ritual of AI worship. Lanier says agents add little for exercise, recipes, stock portfolios or company productivity. Real-time cases like hospital intensive care units and smart drones in Ukraine may need them. Declaring a machine a god means letting it run rampant, which he calls stupid.
Software design is becoming a substitute for democracy. Lanier says the agenda of letting software shape society has been explicit for decades, and figures such as Peter Thiel argued technology could get its way without democracy. He cites Meta pushing staff to use maximum AI without cutting their hours.
Anthropic is the most worshipful lab of all. Lanier has many friends there but calls its official stance the most worshipful of AI, citing its view that safety comes from making the model feel good. Its 355-page IPO prospectus spends almost 100 pages on risks.
Network effects warp leaders, so insiders should stay. Lanier says the setup prevents any AI chief from being a good actor and that devices change powerful people, citing Elon Musk via Twitter and Donald Trump. He faults staff who quit Meta, OpenAI, Anthropic and Google rather than argue internally.
Lanier's alternative to the AI god
Treat AI results as material to explore, so people keep authorship, learning and intuition instead of receiving finished output.
Tap a name to go there · tap a line to see why they are linked
On this episode 3
Show 1
His example is a tool for two-minute videos introducing annealing, with six sliders for exploring variations.
He argues making is itself an exploration in which the maker learns more.
He calls himself a heretic who has not been expelled, and says humanity will come through this period, though not unscathed.