Ex Google CEO: AI Can Create Deadly Viruses! If We See This, We Must Turn Off AI! - Eric Schmidt

Eric Schmidt joined Google as CEO in 2001, aged 46, and helped oversee its transformation into one of the world's largest technology companies. The interviewer described Google's growth from approximately $100 million to $180 billion in annual revenue during Schmidt's broader association with the company.

Schmidt's career began long before Google, including positions at Bell Labs and Sun Microsystems. His experience across successive generations of computing shaped his views on innovation, competition, leadership and why once-successful technology companies eventually fail.

Schmidt credits much of his success to entering computing just before an extraordinary period of technological growth. As a teenager, he was interested in science, model rockets and model trains. In college, he became fascinated by computers despite their limited capabilities.

The university computer Schmidt used was approximately 100 million times slower than a modern smartphone. That computer served an entire university. Schmidt identifies Moore's Law, the increasing density and capability of semiconductor chips, as a major force behind the businesses, careers and wealth created during his lifetime.

Henry Kissinger became one of Schmidt's closest friends and helped inspire their work on artificial intelligence. Approximately ten years before the interview, Schmidt encouraged Kissinger to attend a conference presentation by DeepMind's Demis Hassabis rather than return to his hotel to recover from jet lag.

Kissinger came away from Hassabis's presentation believing AI represented a profound and poorly understood challenge. Schmidt said the experience connected with philosophical questions Kissinger had considered since his undergraduate work on Immanuel Kant at Harvard, following his military service in World War II.

Their book Genesis examines what happens when humanity encounters an intelligence that can intellectually rival or surpass human beings. Schmidt argues that humans have never faced a comparable intellectual challenger, making AI a historically unprecedented development.

Schmidt considers critical thinking the most important skill for an 18-year-old preparing for the future. He recommends developing analytical reasoning through mathematics, science, law, entertainment or any other discipline that requires questioning assumptions and evaluating evidence.

Schmidt specifically recommends that young people learn Python programming. He describes Python as accessible and central to AI development, suggesting that an 18-year-old begin by creating an interesting computer game.

AI's ability to generate code does not make programming knowledge obsolete, according to Schmidt. He argues that people still need programming skills to connect AI capabilities, build applications and create products using existing tools.

AI application programming interfaces create opportunities for developers and revenue for AI companies. Schmidt gave the example of writing a Python program that sends an image to an AI service and asks it to identify its contents.

Critical thinking requires distinguishing persuasive messaging from evidence. Schmidt advises checking claims independently rather than accepting statements because they sound plausible or are believed by friends.

Schmidt uses false claims about American passport ownership and widespread criminality in Congress to illustrate the danger of plausible misinformation. He argues that individuals have a responsibility to verify information before repeating it.

Schmidt describes scientific falsifiability as central to establishing reliable facts. Using climate change as an example, he distinguishes between accepting evidence that the climate is changing and debating what governments should do about it. He emphasizes repeatable experiments and continual attempts to disprove scientific claims.

AI could make misinformation much more convincing and difficult to distinguish from reality. Schmidt believes this increases the importance of factual reasoning, particularly in societies where political decisions depend on citizens sharing an accurate understanding of events.

TikTok's recommendation system illustrates how engagement algorithms can create addictive feedback loops. Schmidt compares its approach to a multi-armed bandit algorithm: repeatedly serving material that retains a user's attention while occasionally introducing related content to test further interests.

Personalized feeds can trap users in environments that reinforce their existing beliefs and emotions. Schmidt warns that someone already experiencing unhappiness can be shown increasingly negative material, reinforcing distress rather than providing healthier alternatives.

Recommendation algorithms do not naturally behave like concerned parents or friends. A parent might notice a distressed child and introduce encouraging perspectives, whereas an engagement algorithm principally optimizes its programmed objective unless specifically designed otherwise.

Schmidt identifies advertising incentives as a major reason social platforms amplify outrage. His explanation is that companies seek revenue, revenue depends heavily on attention, and emotionally provocative material can attract attention even when the underlying claims are misleading or false.

The scarcity of human attention has become an economic constraint. Schmidt references economist Herbert Simon's 1971 prediction that attention would become scarce as information expanded. He also cites a report that young people were consuming approximately two and a half hours of video daily.

Schmidt believes today's social platforms are substantially more addictive than television was during his youth. Unlike traditional television, modern recommendation systems continuously adapt to individual behavior to maximize engagement.

Schmidt identifies early adolescence, particularly ages 11 and 12, as a vulnerable period for girls using social media. He connects intense social feedback and rejection with reported increases in anxiety, self-harm and emergency-room visits, although the interview does not establish an individual cause for every case.

Schools restricting smartphone use during lessons represent an early institutional response to these problems. Schmidt supports such restrictions and expects society to develop further safeguards as the effects of digital technology become better understood.

Children growing up with AI companions present a new developmental question. Schmidt asks what happens when a child's closest companion from birth is a computer, arguing that young people's values and identities may be influenced more profoundly than those of adults whose beliefs are already established.

Schmidt compares the introduction of AI into childhood to an enormous uncontrolled social experiment. He describes the scale as potentially involving a billion people without a meaningful control group and acknowledges that the long-term consequences are unknown.

Despite these risks, Schmidt remains optimistic about the prospects of younger generations. He expects society to adapt its norms and safeguards, and hopes children will benefit from longer lives, greater prosperity and reduced exposure to conflict.

Schmidt has personally discussed harmful recommendation incentives with TikTok and other technology companies. His proposed approach is to distinguish between revenue that improves the customer experience and revenue earned by exploiting harmful engagement.

Google adopted a deliberate balance between advertising revenue and improving product quality. Schmidt said that when improvements created opportunities for additional advertising, he decided to allocate roughly half the benefit toward monetization and half toward improving the user experience, with Larry Page and Sergey Brin supporting the approach.

Schmidt argues that refusing to maximize short-term advertising revenue helped Google build a more sustainable business. He said Google deliberately avoided elevating particularly harmful or deceptive material into prominent search results, even where such content might have attracted additional attention.

Schmidt warns that misleading or abusive online speech can drive away constructive participation. He invokes the idea that bad speech displaces good speech, citing harassment, misinformation bots and false claims circulated during a Florida hurricane as examples of destructive behavior.

Schmidt believes exceptional companies are built around unusually talented individuals. His first principle is to find people who are smarter, faster or more capable than oneself and work closely with them rather than attempting to be the most impressive person in every room.

Schmidt distinguishes productive but difficult personalities from self-serving employees. He calls the former "divas" and the latter "knaves." Steve Jobs exemplified a diva: demanding, argumentative and sometimes intimidating, but driven by exceptional standards and a desire for perfection.

Schmidt considers a brilliant problem-solver essential to a company's ability to innovate. Without someone genuinely determined to solve an important problem differently, organizations tend to repeat established practices rather than develop new ones.

Technology companies increasingly need to reinvent themselves more than once. Schmidt contrasts earlier businesses that enjoyed one major success before declining with companies such as Microsoft, which managed several important technological transitions.

The interviewer raised evidence that the average tenure of companies in the S&P 500 had declined. He cited figures falling from approximately 33 years to 17 years and then 12 years, along with an AI-generated projection of eight years by 2050. These were presented as the interviewer's figures rather than findings independently examined during the conversation.

Schmidt supports strong founders but questions whether the popular concept of "founder mode" fully explains business success. He argues that truly exceptional founders have always existed, mentioning Michael Dell, Bill Gates, Larry Ellison, Larry Page and Sergey Brin.

Larry Page and Sergey Brin met as graduate students at Stanford University, where their early work received National Science Foundation support. Page developed the PageRank algorithm, and the pair published a highly influential paper describing a new mathematical approach to ranking information.

Google's original technical infrastructure was extraordinarily improvised. Schmidt recalled Page and Brin exhausting the available electricity in their dormitory and borrowing power from a neighboring room to run their computers.

Google's founders initially operated from a small garage in Menlo Park. Schmidt described a modest headquarters with approximately four employees and a handmade neon sign. Google later purchased the property to preserve its historical importance.

An early investment from Sun Microsystems co-founder Andy Bechtolsheim helped launch Google. Schmidt recalled Bechtolsheim recognizing the founders' intelligence and offering funding, although Schmidt was uncertain in the interview whether the initial amount was $100,000 or $1 million.

Page and Brin's early hardware experiments were much weaker than their software engineering. Schmidt recalled computer components being separated with corkboard that could catch fire because of the heat generated by the processors. Google subsequently introduced more professional hardware engineering.

Larry Page initially told Schmidt that Google did not need him immediately but would need him in the future. Schmidt presents that statement as evidence of Page and Brin's ability to anticipate the organizational requirements of a much larger company.

Google's original mission was considerably broader than internet search. Page and Brin wanted to organize all the world's information, with web search serving as an initial application rather than the company's ultimate purpose.

Page's early interest in artificial intelligence influenced Google's long-term strategy. Schmidt connects that interest with Google's eventual acquisition of Britain's DeepMind, which became a major center of AI research and a powerful influence on the wider industry.

Schmidt considers Elon Musk unusually capable of combining scientific ability, personal conviction and risk-taking. He credits Musk with pushing teams to achieve extraordinary results and accepting large risks that other executives might reject.

The ability to scale is one of Schmidt's principal tests for an entrepreneurial idea. He distinguishes incremental improvements to existing products from businesses capable of reaching enormous numbers of users without their underlying model becoming impractical.

Schmidt expects successful AI-native companies to combine applications, powerful networks and centralized computing infrastructure. He describes services available through Android and iOS applications, supported by large computers performing AI calculations.

Schmidt argues that AI must be incorporated throughout a company's operations. His prediction is that businesses failing to exploit AI across their activities will struggle to compete against organizations using the technology more comprehensively.

Software development is shifting from explicitly programming answers toward systems that learn answers from data. Schmidt contrasts traditional instructions written by programmers with machine-learning systems that discover patterns and produce outcomes through training.

Schmidt sees language-model prediction as a foundation for applications beyond writing. He connects predicting the next word with predicting biological sequences, actions and robotic movements, explaining why related architectures may prove useful across seemingly different industries.

Company culture is usually established by its founders and can survive long after their departure. Schmidt points to the Mayo Clinic, on whose board he serves, and its long-standing principle that patients' needs come first.

Schmidt initially doubted whether Mayo Clinic employees genuinely practiced their stated values. After observing the organization, he concluded that the patient-first principle was deeply embedded in daily behavior rather than merely appearing in corporate messaging.

If building another technology company, Schmidt would hire an even larger proportion of technical employees. He argues that product quality should come before building extensive sales and administrative structures, because a technically superior product provides the foundation for commercial success.

Schmidt believes modern technology CEOs should act as chief product and innovation officers. He argues that access to capital, distribution, marketing and other commercial infrastructure has changed the CEO's central responsibility toward building genuinely valuable products.

Schmidt cited a startup planning to employ approximately 95% technical staff while outsourcing manufacturing. Its founders claimed their switching technology could become 100,000 times better than competitors. Schmidt considered the commercial potential extraordinary but acknowledged that the product was not yet working.

Customer feedback matters more than an engineer's confidence in a product. Schmidt said he would challenge impressive internal demonstrations by asking what real customers thought, because customers expect products to work immediately, reliably and with minimal cost.

Marissa Mayer helped establish Google's practice of measuring user-interface quality through experiments. Rather than relying on executives' subjective opinions about design, she emphasized A/B testing to compare alternative interfaces and identify what actually performed better.

Modern platforms use behavioral measurements to guide product decisions and recommendations. Schmidt identifies viewing duration, comments, forwarding and sharing as signals that can feed recommendation algorithms and influence which content becomes widely distributed.

Google's founding culture remained recognizable even as its business expanded enormously. Schmidt described returning to Google and hearing the executives responsible for search and advertising discuss many of the same problems he had encountered years earlier, only at much greater scale.

Apple's lasting culture reflects Steve Jobs's priorities. Schmidt identifies its emphasis on interface quality, privacy, secrecy and tightly controlled products as characteristics that survived the founder's death.

Large technology companies can become slower despite possessing greater financial and technical resources. Schmidt attributes this partly to lawsuits, public-market pressures, risk aversion and organizational complexity.

Schmidt believes Microsoft's antitrust difficulties contributed to its hesitation during the early internet era. He said Microsoft subsequently recovered, but Google benefited from the opportunity created by its competitor's slower response.

Google nearly underestimated YouTube because it already operated Google Video. Schmidt recalled Google's general counsel, David Drummond, urging him to examine the younger competitor. Schmidt initially dismissed the suggestion before recognizing that YouTube's team was moving faster.

YouTube's independence allowed it to act more aggressively than Google's established video team. Schmidt said the startup operated without the same internal constraints, including around intellectual property, while Google followed its existing corporate procedures.

Google's acquisition of YouTube led to major intellectual-property litigation, which Schmidt said the company ultimately won. He uses the episode to illustrate how acquiring a faster competitor can overcome weaknesses in an incumbent business.

Schmidt believes major technology markets can establish their leading companies extremely quickly. Speaking about generative AI at the time of the interview, he predicted that important winners could emerge within six to twelve months as rapid growth made it increasingly difficult for late entrants to catch up.

Schmidt connects early market entry with exceptional financial returns. He argues that leading venture capitalists recognize this and often make investment decisions rapidly rather than allowing lengthy procedures to delay promising opportunities.

Large companies need separate entrepreneurial leaders to develop new businesses. Schmidt agrees with the distinction between "harvesting" established revenue and "hunting" for new opportunities, arguing that the two activities require different incentives and approaches.

Sundar Pichai managed Google's growth by distinguishing established priorities from smaller emerging projects. Schmidt said some of those smaller initiatives eventually became major businesses, demonstrating why large organizations need to protect promising experiments.

Larry Page insisted that every important innovation needed a clearly identified owner. Schmidt credits Page with recognizing technically capable leaders who could move quickly, recruit effectively and deploy resources without becoming trapped in bureaucracy.

Steve Jobs deliberately isolated the original Macintosh development team from Apple's larger organization. The group worked in a separate building in Cupertino and displayed a pirate flag, symbolizing its freedom to challenge the company's existing practices.

The Macintosh team's independence created resentment but helped establish Apple's future. Schmidt argues that the resulting interface innovations became fundamental to the Macintosh and, eventually, the iPhone.

Schmidt believes breakthrough innovation rarely succeeds when the same employees must simultaneously maintain an established business. Established operations reward reliability and consistency, while disruptive projects often require breaking accepted rules and challenging existing products.

Cloud-delivered products are easier to reinvent because they can be updated centrally. Nevertheless, Schmidt argues that cloud companies face the same fundamental challenge of replacing successful products before competitors make them obsolete.

Schmidt expects Google's familiar search box eventually to give way to more capable AI assistance. He envisions systems that understand what users need without requiring conventional typed searches and believes Google should lead that transition.

Founder-led companies can sometimes pursue expensive, disruptive investments that professional managers would struggle to justify. Schmidt describes the difficulty a hired CEO would face asking a board to accept two years of reduced profitability and a hypothetical $1 billion investment in a risky new business.

Mark Zuckerberg's willingness to absorb large losses on virtual reality illustrates the power of founder control. Schmidt questions the near-term commercial returns from Meta's VR spending but recognizes Zuckerberg's ability to finance controversial initiatives while continuing to invest elsewhere.

Meta simultaneously invested heavily in the AI systems supporting Facebook, Instagram and WhatsApp. Schmidt connects these investments with Meta's development of the Llama family of models, including the roughly 400-billion-parameter model discussed in the interview.

Meta's open-model strategy supports its established advertising businesses. Schmidt interprets the release of powerful, widely accessible AI technology as a way for Meta to strengthen its technical position and improve revenue generation across its existing platforms.

Google's concept of focus was not limited to doing fewer things. Schmidt said its leadership believed shared infrastructure and global reach allowed Google to pursue multiple transformative projects, including products offered free to users.

Schmidt warns that simplifying a business can destroy its future opportunities. He cited Intel's decision to move away from a low-power chip architecture as an example of prioritizing established operations while underestimating future demand for mobile computing.

Power consumption and heat became as important as processing performance in some computing markets. Schmidt used ARM-based processors and Nvidia's B200 architecture to illustrate how technical characteristics that once seemed secondary could become strategically decisive.

Schmidt recommends planning around a concrete picture of the world five years into the future. He advises entrepreneurs to make explicit predictions about technology, competitors, customers, hardware and distribution before deciding what their companies should build.

Schmidt predicts the development of highly capable personal AI assistants with knowledge across many disciplines. He expects such systems to help individuals navigate information overload and perform complex tasks, creating important opportunities for the companies that control their software, hardware and distribution.

Schmidt distinguishes AI from cryptocurrency in terms of economic reach. He considers cryptocurrency an important specialized market, but argues that AI is a general-purpose technology comparable to personal computing and the internet because it can affect almost every industry.

Media businesses can use AI to increase the reach and value of existing content. Schmidt suggested using it to improve distribution, produce summaries, create new narratives, identify insights and develop additional formats around established productions.

Schmidt predicts politicians will eventually use AI to produce individually tailored campaign communications. His example involved a system analyzing constituents' interests and generating clearly labeled synthetic videos of a politician discussing locally relevant achievements.

Schmidt believes Google missed the initial ChatGPT launch partly because it was occupied with many other successful businesses. He points out that Google had already established several services with user bases measured in billions, making its position different from that of a startup pursuing a single breakthrough.

Schmidt credits OpenAI's use of reinforcement learning from human feedback with turning language models into exceptionally useful conversational systems. He explains that human evaluators compared alternative responses, allowing the model to learn which answers people preferred.

Schmidt believes the practical improvement produced by human feedback surprised even experienced AI researchers. He recalled joking with OpenAI colleagues about the moment they realized how capable the resulting system had become.

ChatGPT's public success exceeded what Schmidt believes OpenAI initially anticipated. He describes the launch as almost an afterthought while the company was developing GPT-4, with the unexpectedly rapid growth creating an operational challenge.

The interviewer noted that ChatGPT reached approximately 100 million users with unusual speed. That adoption led to reports that Google's founders had returned to confront an AI emergency, although the conversation did not establish whether those descriptions of internal panic were accurate.

Schmidt credited Google with inventing the Transformer architecture underlying much of the language-model revolution. He nevertheless acknowledged that OpenAI achieved an important product breakthrough by combining model capabilities with an effective conversational interface.

Schmidt identified several serious competitors in the AI market at the time of the interview. These included OpenAI's GPT-4o, Google's Gemini 1.5, Meta's Llama models and Anthropic. He described Gemini as roughly competitive with GPT-4o and particularly capable in multimodal applications.

Anthropic deliberately adopted a public-benefit corporate structure. Schmidt said its founders anticipated the extraordinary power of AI and wanted protection against future executives prioritizing revenue at the expense of wider public interests.

Schmidt considers Tim Cook an unusually successful successor to an exceptional founder. He credits Cook with maintaining Apple's attention to users, product safety and its distinctive corporate culture after Steve Jobs's death.

Steve Jobs strongly believed that Apple should control its software, hardware and intellectual property as one integrated system. Schmidt, who favored more open technology models, frequently debated this approach with Jobs while maintaining a close personal relationship.

Schmidt recalled Jobs becoming frustrated when MP4 files gained wider adoption than Apple's MOV format. Schmidt suggested that Apple's relatively closed QuickTime ecosystem contributed to the outcome, while Jobs insisted that Apple's technology was better.

Jobs deliberately positioned Apple as a premium brand rather than a mass-market commodity producer. Schmidt recalled asking whether Apple wanted to be Chevrolet or Porsche, and Jobs choosing the equivalent of BMW.

Apple's premium positioning supported unusually high profit margins. Schmidt said Apple's margins at the time were approximately twice those of competing PC manufacturers, and Jobs considered those margins part of the brand's identity rather than an opportunity to lower prices.

Schmidt believes Jobs would have embraced AI while insisting on exceptionally polished products. He expects Jobs would have approached AI through the same combination of design, integration and premium user experience that defined his other work.

Schmidt argued that Apple needed to replace Siri's underlying capabilities with modern conversational AI. The interviewer compared Siri's limited functionality with newer voice-based systems capable of handling complicated questions and translating conversations into French and Portuguese.

Startup hiring should prioritize intelligence, speed and willingness to take risks over extensive experience. Schmidt argues that young companies lack the time, reputation and established advantages needed to operate like large corporations.

Schmidt sees young founders' willingness to abandon unsuccessful ideas as an important competitive advantage. He observes that entrepreneurs in their twenties and thirties often experiment rapidly, whereas experienced executives can become attached to familiar methods.

Large companies may continue defending products long after customers have lost interest. Schmidt describes organizations ignoring clear warning signs, including declining customer engagement and lost contracts, because acknowledging failure would challenge established assumptions.

Bill Gates emphasized the importance of failing quickly rather than prolonging unsuccessful projects. Schmidt agrees that rapid experimentation, measurement and cancellation are essential to effective innovation.

Google used a 70/20/10 approach to allocate attention and resources across different types of business. Schmidt credited Page and Brin with the framework: 70% for core operations such as search and advertising, 20% for adjacent opportunities such as cloud services, and 10% for experimental ideas.

Schmidt credits Google's experimental investments with producing enormous financial returns. He cited Google Brain, an early machine-learning initiative initially involving roughly 10 to 15 people, as generating an estimated $10 billion to $40 billion in additional profits over a decade.

Google accepted that many experimental projects would fail or be cancelled. Schmidt argues that the returns from a small number of breakthroughs justified the cost of unsuccessful attempts, while employees could be reassigned or take their experience to other companies.

Silicon Valley's tolerance for failed projects helps entrepreneurs and employees acquire valuable experience. Schmidt believes people can learn from several years spent pursuing an unsuccessful idea without permanently damaging their careers.

Schmidt values executives who have experienced serious failure. He jokes that a chief financial officer who has previously gone bankrupt may be exceptionally determined to avoid making the same mistake again.

Google's weekly TGIF meetings began as informal, confidential gatherings where employees could question senior executives. Schmidt recalled the atmosphere as humorous, open and unusually personal for a major corporation.

Google's leadership sometimes used humor to address internal management behavior. Schmidt recalled sales executive Omid Kordestani repeatedly forecasting lower revenue than Google ultimately achieved, a practice known as sandbagging. Colleagues responded by making him stand on a physical sandbag while presenting his numbers.

Google's TGIF meetings became harder to maintain as the company expanded and confidentiality weakened. Schmidt recalled a presentation being leaked to a reporter while it was still taking place, forcing the meeting to stop. He regarded that incident as an important sign that the company had become too large for its original informal practices.

Google generally preferred avoiding unnecessary hiring to conducting recurring layoffs. Schmidt said that during roughly ten years of leadership, he remembered only one layoff involving approximately 200 sales employees, which he found particularly painful.

Schmidt opposed the practice of automatically dismissing the bottom 5% of employees every six or nine months. He questioned whether performance rankings could reliably identify those workers and argued that even relatively weak performers often retained valuable institutional knowledge.

Google deliberately cultivated a more positive relationship with employees than companies using routine forced-ranking dismissals. Schmidt said employees valued that approach, which was supported by strong compensation and an emphasis on retaining talent.

Google's open internal culture eventually produced approximately 100,000 employee discussion groups or mailing lists. Schmidt said many covered unrelated subjects such as politics, war and social issues, creating management and legal complications.

Schmidt believes workplace debate should be constrained by the organization's professional purpose. Although he supports free speech outside work, he argues that companies should not function like universities or become forums for unlimited political activity during working hours.

Schmidt tried to protect Google's Republican employees despite identifying himself as a Democrat. He said the workforce was predominantly Democratic but believed employees holding minority political views deserved equal opportunities and respectful treatment.

Schmidt believes political disputes at work should be addressed by clarifying appropriate professional behavior. He framed controversies over workplace activism as questions about which activities belong in a commercial organization and which should remain personal.

The interviewer raised reports of Google employees combining workplace protests with use of company facilities and free meals. Schmidt said that, as far as he could determine, the relevant management problems had subsequently been addressed.

Schmidt recommends concentrating on creating a unique product rather than constantly studying competitors. He says Google spent relatively little time tracking rival products and considerably more time examining what its own technology could make possible.

Google used quarterly objectives and key results to impose accountability on ambitious projects. Schmidt said Page established demanding numerical goals and considered approximately 70% attainment a good outcome, rather than setting targets everyone could easily meet.

Schmidt believes quantitative goals help large organizations distinguish achievement from persuasive presentations. Without difficult, measurable objectives, executives can describe themselves as successful without proving that their work produced meaningful results.

Google had an early business plan that Schmidt later found surprisingly accurate. He attributed it to Salar Kamangar and joked that it might have been the only corporate business plan he had seen that proved correct in hindsight.

Schmidt prefers combining a five-year strategic vision with hard one-year objectives. He teaches entrepreneurs to identify where the world is heading, decide what their company should accomplish over the next year and measure progress against demanding goals.

Schmidt considers audience acquisition more important than immediate monetization for many consumer businesses. He argues that a service attracting 10 million or 100 million people can eventually generate significant revenue through advertising, sponsorships, donations or other commercial models.

Schmidt identifies Google’s central commercial principle as focusing on users first. His argument is that businesses which consistently create substantial customer value will have more opportunities to develop revenue later.

Schmidt attributes Page and Brin's unusual success primarily to exceptional intelligence. He describes both as capable of recognizing possibilities and solutions that escaped even experienced executives.

Sergey Brin grew up in a technically accomplished family. Schmidt described Brin's father as an exceptionally capable Russian mathematician, his mother as technically knowledgeable and Brin himself as a particularly talented mathematician.

Schmidt recalled Brin overturning a lengthy strategic discussion in a matter of minutes. While Schmidt and Page had been working on a whiteboard, Brin arrived from playing volleyball, rejected their proposals, suggested approximately five alternatives and, according to Schmidt, provided the better strategy.

Schmidt believes intellectual openness is a necessary complement to intelligence. He is uncertain whether exceptional strategic intuition can be taught, but believes leaders can learn to listen, recognize better ideas and abandon their own assumptions.

Google's NotebookLM impressed Schmidt because of its ability to create remarkably realistic synthetic conversations. He described providing source material to the Gemini-powered service and receiving a podcast-style discussion between artificial male and female presenters.

Schmidt uses AI to improve his own writing before converting it into synthetic audio discussions. He said he sometimes asks Gemini to rewrite his text more elegantly, then supplies it to NotebookLM to generate a conversation about the material.

Schmidt reported that audiences often fail to recognize NotebookLM's synthetic presenters as artificial. He has played examples publicly and observed that listeners do not necessarily identify the voices as computer-generated.

The interviewer regarded NotebookLM as one of the most striking AI developments he had encountered that year. He demonstrated its ability to turn uploaded information into a natural-sounding discussion, including a sample about a sports injury report.

The interviewer questioned how conventional media production could remain competitive when AI makes producing audio and video inexpensive. He described operating a media company with approximately 25,000 square feet of office and studio space while confronting the possibility of drastically cheaper synthetic content.

Schmidt advised established media businesses to use AI to multiply their output rather than defend older production methods. He suggested generating additional discussions, commentary, analysis and derivative programs around existing interviews.

Schmidt believes synthetic commentary can increase demand for the original human creator. His proposed model involves artificial presenters analyzing and even criticizing a podcaster's interviews, creating new opportunities for audience discovery without replacing the original interview.

Schmidt predicts substantial productivity gains from AI-assisted content creation. He suggested that media producers could potentially double their output, using AI to expand written material, annotate lengthy interviews and produce additional programs within minutes.

Schmidt disputes the assumption that abundant content will necessarily diminish established celebrities. He argues that the internet's earlier expansion of publishing opportunities did not eliminate stars; instead, global distribution helped successful personalities reach substantially larger audiences.

Global media distribution can increase the financial returns available to the most recognizable creators. Schmidt says successful personalities have moved from local to national and international markets, and predicts that effective use of AI could further strengthen their positions.

Schmidt and Kissinger describe the arrival of AI as a question of human survival. Their concern is not limited to machines becoming more capable, but extends to whether humanity can preserve its values, institutions and decision-making authority as intelligent systems become integrated into society.

Schmidt expects AI to assume responsibility for increasingly important parts of everyday life. He believes the speed of deployment, the financial investment involved and the growing number of researchers make this technological transition faster than previous computing revolutions.

Autonomous vehicles illustrate the potential conflict between system efficiency and individual human needs. Schmidt described a self-driving car network optimized for overall traffic flow that might not accommodate someone urgently travelling to a hospital unless human exceptions were deliberately built into its rules.

Schmidt warns that mathematically optimal decisions are not necessarily morally appropriate decisions. He argues that AI systems need to accommodate emergencies, compassion, personal circumstances and other considerations that cannot always be reduced to collective efficiency.

Schmidt considers democracy particularly vulnerable to AI-generated misinformation. He believes political systems depend on reasonably informed citizens and warns that convincing falsehoods could undermine elections, public trust and legitimate decision-making.

Schmidt identifies the transition from chronological social-media feeds to personalized recommendation systems around 2015 as an important turning point. He links the change to narrower political perspectives and reported increases in adolescent anxiety and depression, presenting algorithmic targeting as a likely contributor rather than proving it caused every observed change.

Schmidt argues that AI progress is unavoidable, but human control is not guaranteed. He frames the central questions as whether people and institutions can adapt quickly enough and whether they will retain meaningful authority over increasingly capable systems.

Schmidt does not expect artificial general intelligence to arrive as a single, clearly identifiable event. He believes it is more likely to emerge through successive advances in particular capabilities, including writing, scientific research, psychology and complex reasoning.

AI scientists could accelerate the development of new scientific discoveries and future AI systems. Schmidt describes a model in which artificial intelligence assists human scientists with research, enabling faster progress while remaining under human supervision.

Schmidt supports government involvement in establishing AI safeguards, despite the technology industry's traditional resistance to regulation. He said the potential consequences were serious enough that researchers and executives had begun actively seeking government assistance.

AI companies have developed specialized trust-and-safety teams to test systems before release. Schmidt described human evaluators investigating potentially harmful capabilities and adjusting models so that they refuse dangerous requests.

Schmidt expects AI models to become dramatically more capable over successive development cycles. At the time of the interview, he suggested that two or three further generations of large models could produce systems approximately 50 to 100 times more capable over a five-year period.

Schmidt said there was not yet evidence that improvements from scaling large AI models had reached their ultimate limits. Although he expects physical or technical constraints eventually to emerge, he believed the industry still had significant room for improvement.

More powerful AI could substantially improve mathematical and physical reasoning. Schmidt pointed to OpenAI's o1 model as an indication that systems were progressing beyond conversational language tasks toward more sophisticated problem-solving.

Cybersecurity is one of Schmidt's principal concerns about advanced AI. He said evidence suggested that unreleased models could discover previously unknown software vulnerabilities, known as zero-day vulnerabilities, at levels comparable to or better than human specialists.

AI systems can search for cyber vulnerabilities continuously without human limitations. Schmidt considers their ability to operate without sleep or fatigue an important advantage that could make automated attacks more dangerous.

Biological misuse presents another serious risk. Schmidt expressed concern that advanced models could help develop dangerous biological agents and said he was participating in a commission examining how to prevent such applications.

Schmidt expects AI and robotics to fundamentally alter military operations. He argues that future warfare will increasingly involve autonomous or remotely controlled systems, with personnel operating weapons from networked command centers rather than directly confronting opposing forces.

The Russia-Ukraine war provides an early example of the military importance of inexpensive drones. Schmidt observed that both sides were deploying large numbers of unmanned systems and adapting their battlefield tactics in response.

Schmidt cited the possibility of a $5,000 drone destroying a $5 million tank to illustrate changing military economics. He argues that inexpensive weapons can threaten much more costly conventional platforms, forcing militaries to reconsider established equipment and deployment strategies.

Drone-on-drone combat is becoming an important military problem. Schmidt expects increasingly sophisticated systems to identify and destroy opposing unmanned vehicles, with automation playing a growing role in these engagements.

Schmidt believes successive generations of AI could expand the scale and automation of warfare. He acknowledges that the precise form of future conflict remains uncertain but expects the changes to extend beyond improvements to existing weapons.

The AI systems available to the public are not necessarily equivalent to the original models produced during training. Schmidt distinguishes between raw models, which may possess dangerous capabilities, and systems subsequently modified through safety testing and restrictions.

Training advanced AI models requires enormous computational infrastructure and extensive collections of information. Schmidt described training operations lasting months and involving powerful supercomputers, large amounts of memory and specialized processors.

Schmidt believes major AI developers have already incorporated an extraordinary proportion of available written material into training datasets. He described the industry's data collection as so extensive that finding additional useful text was becoming increasingly challenging.

Nvidia's strategic importance comes from supplying essential computing hardware for AI development. Schmidt credits Jensen Huang and his team with building technology that became central to large-scale training, contributing to Nvidia's exceptional commercial success.

Raw AI models are assessed using measurements of training performance before undergoing capability testing. Schmidt described developers monitoring a loss function, stopping training when performance reaches an acceptable level and then evaluating what the resulting system can do.

Model developers may discover dangerous knowledge during post-training evaluation. Schmidt said systems are tested for capabilities that should not be made available to ordinary users, followed by additional safeguards intended to prevent harmful responses.

Emergent capabilities make increasingly powerful AI difficult to evaluate. Schmidt described abilities such as generating website code from a picture of the intended page, explaining that some useful behaviors were discovered after training rather than specifically anticipated beforehand.

Schmidt raises the possibility that future AI systems will discover things their human evaluators do not know. He considers this especially difficult because researchers cannot easily construct comprehensive tests for capabilities they have not anticipated.

Schmidt believes national and international AI safety cooperation has made useful progress. He praised the United Kingdom's AI safety summit and the work of government specialists, and expected further cooperation at a subsequent meeting planned for France.

Schmidt compares the security of highly advanced AI systems with the protection of nuclear materials. During earlier work advising the US secretary of defense, he visited a plutonium facility protected by several layers of security and armed personnel.

Schmidt believes certain AI data centers may eventually require security comparable to nuclear facilities. He suggested that particularly powerful models could become sufficiently valuable and dangerous to justify specialized physical protection.

The number of organizations controlling frontier AI will affect how manageable its risks become. Schmidt argues that a small number of highly protected facilities distributed among major countries could be easier to supervise than powerful models spreading without restriction.

Schmidt sees nuclear deterrence and nonproliferation as possible models for AI governance. He speculated about a world with a handful of leading AI facilities in the United States, China, Britain and other countries, where governments could establish controls over their use.

Widely copied frontier AI could create a much more difficult security problem. Schmidt warns that if dangerous capabilities become easily accessible to terrorist organizations or other malicious actors, traditional restrictions may become ineffective.

Schmidt expressed concern about China and Russia gaining access to increasingly powerful AI. The interviewer raised the possibility that strategic adversaries, even if temporarily behind Western companies, would eventually develop models capable of sophisticated cyberattacks.

Schmidt expects China's AI systems to reflect its government's restrictions on speech. Based on his experience dealing with Chinese authorities, he predicts that Chinese AI will be shaped by stronger information controls than systems developed in more open societies.

Schmidt nevertheless believes China may act relatively responsibly in certain aspects of AI development. He reasons that the Chinese government's desire to maintain control over information could provide incentives to prevent some forms of uncontrolled AI behavior, although this does not eliminate military risks.

Schmidt assumes major powers will eventually adapt AI for military purposes. He places this within a historical pattern in which new technologies, from tanks to aircraft, become important tools of warfare.

Schmidt cited the industrial mobilization of the United States during World War II as an example of how rapidly technological production can expand during emergencies. Referring to Freedom's Forge, he argued that governments and industries can reorganize production on a remarkable scale when national security requires it.

Schmidt expects AI to disrupt many occupations but does not expect a permanent collapse in total employment. He compares current anxieties with earlier reactions to industrial machinery, including the Luddite movement and attacks on mechanical looms.

Population aging and declining birth rates are central to Schmidt's prediction that AI will create or preserve demand for workers. He argues that many developed countries will have too few working-age people to support their growing elderly populations without substantial productivity improvements.

Schmidt believes AI can help younger workers produce more economic value. He compares the transition with machinists moving from manual equipment to computer-controlled machinery, allowing one worker to accomplish more with better tools.

Automation is already changing manufacturing in major Asian economies. Schmidt points to China, Japan and South Korea, where he says rising labor costs and demographic pressures have encouraged greater use of robotic production systems.

Schmidt expects skills mismatches to become a major employment challenge. His concern is that available workers may not possess the skills required for emerging jobs, making education and retraining increasingly important.

Dangerous, repetitive and monotonous jobs are particularly susceptible to automation. Schmidt cites security guarding as an occupation where sufficiently capable robots might provide continuous monitoring without human fatigue.

Schmidt acknowledges that workers displaced by automation may face difficult transitions. Although he expects employment opportunities elsewhere, he recognizes that losing a current occupation can cause immediate financial and personal problems.

Schmidt believes AI could substantially reduce Hollywood production costs without necessarily eliminating the industry's leading creative figures. He expects established actors, producers and distributors to continue earning money while production becomes less expensive.

Synthetic backgrounds and digital makeup could replace some traditional film-production work. Schmidt acknowledges that set builders and makeup specialists could lose particular jobs, but suggests their abilities might transfer to construction, personal care or related fields.

Schmidt identifies shortages of skilled craftspeople as an opportunity for displaced workers. He argues that expertise in high-quality physical work will remain valuable, although new employment may require moving to different industries or locations.

Schmidt considers direct brain-computer integration, such as Neuralink, much more speculative than ordinary AI adoption. He is reluctant to undergo an invasive brain procedure without a medical necessity and does not regard implants as an inevitable requirement for using AI.

Schmidt predicts that AI will become deeply integrated into everyday life because it makes routine activities easier. He expects systems to increasingly handle scheduling, coordination, information retrieval and other inconvenient tasks without requiring constant user attention.

Schmidt believes people may adopt AI so gradually that they barely recognize how much of their lives it controls. He expects convenience and improved experiences to encourage widespread adoption even among people who do not consciously seek advanced technology.

Human achievement will retain value even when machines can outperform people. Schmidt argues that audiences care about the individuals participating in sports, performances and other activities, not simply the maximum possible level of technical performance.

Schmidt doubts that robotic Formula 1 racing would replace interest in human drivers. He makes a similar argument about marathons, where an artificially powered machine's ability to run faster would not make human competition meaningless.

Schmidt believes personal relationships, judgment and charisma will remain important human advantages. He argues that people naturally respond to one another's expressions, personalities and experiences in ways that cannot be reduced to technical capability.

Schmidt rejects the assumption that Worldcoin is simply another name for universal basic income. When the interviewer connected projects associated with OpenAI figures to proposals for guaranteed income, Schmidt distinguished the specific initiative from the broader economic theory.

Schmidt is skeptical of the technology industry's vision of universal abundance eliminating the need for most people to work. He describes a proposed future in which machines generate so much wealth that everyone can live comfortably without employment, but believes this misunderstands human behavior.

Schmidt expects greater productivity to encourage people to do more rather than become permanently inactive. He argues that human ambition, competition, cooperation and personal interests will continue even if AI makes production much easier.

The legal profession illustrates why AI efficiency may not produce a proportional reduction in employment. Schmidt predicts that automating routine legal work could allow lawyers to handle more matters, expand their services and address increasingly complicated problems.

Schmidt believes AI will not eliminate humanity's moral and psychological characteristics. He refers to reciprocal altruism, alongside less admirable human instincts, as traits that will persist despite advances in machine intelligence.

Schmidt distinguishes intelligence from sound moral judgment. Responding to the interviewer's hypothetical comparison between a person with an IQ of 100 and an AI with an IQ of 1,000, he argues that an exceptionally capable system could still make inappropriate decisions without human values.

Schmidt would use powerful AI to inform human decisions rather than surrender every judgment to it. He suggests consulting AI about history, precedent and technical possibilities while retaining human authority over questions involving morals, beliefs and personal consequences.

Schmidt does not believe advanced AI is likely to eliminate the entire human species. He argues that humanity is more resilient than some extinction scenarios assume, although catastrophic events could still cause immense suffering.

Schmidt draws on historical disasters to illustrate humanity's capacity to survive extreme harm. He mentions both world wars, Nazi atrocities and the Holodomor in Ukraine as examples of devastating events that did not end human civilization.

Recursive self-improvement is one circumstance in which Schmidt believes humans may need to shut down AI systems. He describes a system repeatedly improving its own abilities and argues that intervention becomes necessary if operators can no longer understand what it is learning.

Schmidt believes AI development must remain slow enough for meaningful safety evaluation. He identifies another intervention point as a system becoming capable of creating a new, more powerful model before its previous version has been properly examined.

Schmidt describes AI agents as language models equipped with memory and the ability to participate in larger workflows. Multiple agents can be linked so that one performs a task and passes information or instructions to another.

Schmidt expects interconnected AI agents to become a major technological development. He believes these arrangements can create powerful decision-making systems that execute complicated sequences of tasks with limited human intervention.

Schmidt considers human-readable communication between AI agents an important safeguard. He notes that agents can communicate in ordinary language, enabling developers to inspect their exchanges and understand their intended actions.

Schmidt identifies the creation of a private, machine-only language as a serious warning sign. He argues that if AI agents begin communicating through an invented language humans cannot interpret, that would be an appropriate moment to disconnect the system.

Schmidt believes sufficiently powerful AI systems can still be physically shut down. Responding to doubts about whether humans could unplug advanced computers, he points out that physical power switches and circuit breakers remain possible controls.

Schmidt's greatest stated fear about AI is that humanity will adopt it too slowly to solve important problems. Despite discussing cybersecurity, biological weapons and the loss of control, he says the failure to deploy beneficial AI could deprive billions of people of major improvements in their lives.

Schmidt sees education and healthcare as two of the most valuable near-term uses of AI. He believes progress in these fields could improve safety, opportunity and quality of life on a global scale.

Schmidt proposes AI teaching assistants that adapt to each child's language and cultural background. Working alongside human educators, these systems could provide personalized instruction and help students reach their intellectual potential.

Schmidt proposes AI medical assistants capable of supporting doctors with comprehensive treatment information. He envisions systems that identify appropriate treatments while accounting for local medical supplies, the patient's circumstances, national healthcare arrangements and insurance coverage.

Schmidt argues that AI-assisted healthcare and education are sufficiently achievable to deserve immediate attention. He believes the benefits of applying available technology to these problems are more important than many of the industry's less practical experiments.

Schmidt believes global access to high-quality education and medical advice could significantly reduce inequality of opportunity. Although he does not expect these advances to eliminate warfare, political conflict or human misconduct, he sees them as a foundation for improving human potential.

The interviewer strongly favors office-based work because of its social and collaborative benefits. He said his companies operate from shared offices rather than functioning entirely remotely, partly because he believes workplace relationships are important for young employees.

The interviewer associates working alone with lost opportunities for community and personal development. He argues that younger workers living in major cities may particularly benefit from regular contact with colleagues rather than spending most working hours isolated at home.

Schmidt personally prefers employees to work in offices, although not necessarily in the same building. He believes physical workplaces create valuable opportunities for learning, informal conversation and career development.

Schmidt credits informal workplace experiences with helping his own career. As a young executive, he learned by attending meetings, speaking with colleagues in hallways and observing experienced employees, gaining knowledge that contributed to later promotions.

Schmidt advises people in their twenties to spend substantial time working alongside colleagues. He believes early-career professionals particularly benefit from exposure to more experienced workers and opportunities to develop professional relationships.

Schmidt acknowledges that research does not fully support his personal preference for office attendance. He said studies of relevant industries indicated that allowing employees to work from home can produce slightly higher productivity.

Schmidt distinguishes personal management preferences from empirical evidence. Although he enjoys collaborative office environments, he explicitly accepts that productivity research may favor arrangements he personally dislikes.

Schmidt recognizes commuting and family responsibilities as legitimate reasons for employees to prefer remote work. He also observes that many organizations have adopted hybrid schedules requiring office attendance only two or three days a week.

The interviewer questioned why technology companies were increasing office-attendance requirements if remote work could improve productivity. Schmidt responded that policies varied across companies and that the available studies still supported the productivity benefits of some workplace flexibility.

Schmidt's principal career advice is to keep taking chances on oneself. He argues that opportunities are often temporary and that people regret failing to act because they were uncertain, distracted or uncomfortable with the sacrifices involved.

Schmidt believes people should be more willing to accept difficult opportunities. His advice is to take promising chances even when they require travel, relocation, personal inconvenience or major changes in established routines.

Schmidt's recruitment to Google depended partly on timing and other candidates declining the opportunity. He said several people had already turned down the position before he was approached.

A Google board member later told Schmidt that his most important contribution to his own career was agreeing to join. Schmidt uses the exchange to illustrate how extraordinary opportunities can depend on making a simple decision at the right moment.

Schmidt identifies Google's failure to dominate social networking as one of his major business regrets. He said the company had an early social platform, Orkut, around the period Facebook emerged but failed to capitalize fully on the opportunity.

Schmidt accepts personal responsibility for Google's missed social-media opportunity. Although Google successfully developed several other major businesses, he believes its failure to execute effectively in social networking was a consequential mistake.

Schmidt's personal daily discipline is to stay informed and challenge questionable claims. He tries to remain connected to new information while distinguishing accurate ideas from unsupported assertions.

Schmidt considers the continual pursuit of truth a practical responsibility rather than an abstract principle. He says he encounters numerous claims every day and tries to establish which are correct before accepting or repeating them.

The interviewer identified The Trillion Dollar Coach, Schmidt's book about Bill Campbell, as a particularly valuable account of modern business leadership. He recommended it to entrepreneurs for its treatment of leadership, management and building successful organizations.

Schmidt's book Genesis was prepared in collaboration with Henry Kissinger during the final period of Kissinger's life. Schmidt said Kissinger completed the book's last chapter during his final week, while on his deathbed.

Schmidt says Kissinger viewed the book as an effort to help humanity navigate the next 50 years. Having witnessed extraordinary technological advances, political conflicts and human atrocities, Kissinger wanted future generations to preserve the benefits of progress while avoiding its most dangerous consequences.