Les travailleurs vont-ils être remplacés par des IA ? (La Silicon Valley y croit déjà)

This summary of the *Silicon Carné* podcast, recorded in San Francisco, explores the rapid acceleration of artificial intelligence and its profound impact on security, leadership, labor markets, and the creative industries. ### The Hidden Mechanics of Identity Verification A significant portion of the discussion centers on a major security discovery by Johann Sido regarding "Persona," a third-party identity verification service used by OpenAI, LinkedIn, and Roblox. Sido identified a server named "OpenAI Watchlist" that exposed the underlying code for these verification processes. The findings reveal a level of surveillance far more invasive than users realize. When a user submits a selfie for identity verification on ChatGPT, the system performs 269 separate checks. This includes scanning the background of the photo to see if the environment is "known" and comparing the user's face against global databases of political figures, oligarchs, and suspected terrorists. Suspicious reports are automatically forwarded to the U.S. Treasury. Jérémy Michel notes that while this is legal, it represents a "mass surveillance machine" where private companies act as proxies for government interests. This creates a "Minority Report" scenario where "suspicious face" scores are generated without transparency, potentially allowing governments to target specific ethnicities or religious groups by simply adding a few lines of code. ### The Rise of the "AI CEO" and Augmented Leadership The conversation shifts to Sam Altman’s suggestion that OpenAI could eventually be led by ChatGPT. This introduces the concept of the "Augmented CEO." While a human CEO is still essential for public relations, high-level networking, and "shaking hands," AI is becoming superior at data aggregation. An AI leader can instantly analyze Slack logs, emails, and product data—tasks that would take a human dozens of reports and meetings to synthesize. This trend is already visible in "vibe coding" and small-scale entrepreneurship. Some founders now use AI agents to manage business operations via APIs, allowing a six-person team to achieve the output of a 100-person company. The consensus is that while the CEO might be the "last person to turn off the lights," their role is merging with the machine to overcome the human "bottleneck" of information processing. ### The Disruption of the Labor Market The podcast addresses the alarming statistics of early 2026: over 100,000 job cuts in the U.S. tech, healthcare, and transport sectors in a single month. Eric Gervé argues that 100% of jobs will eventually be changed by AI and robotics. A demo of "Vision Claude" (AI agents using Meta glasses to see and act in the real world) illustrates a future where agents might "prompt" humans. In this "dystopian yet incredible" world, an AI agent could identify a task in the physical world and hire a human via a marketplace to execute it. A critical concern is the "lost generation" of young workers. Junior roles—tasks like note-taking, spreadsheet management, and basic research—are being entirely automated. This leaves 22-year-olds with no entry point to gain the experience needed to compete with "augmented" seniors. Jérémy Michel adds a geopolitical layer, arguing that Western "tertiary" elites (lawyers and politicians) are unprepared for a future that requires a return to primary and secondary industries—extraction, energy, and robot manufacturing—where countries like China are currently leading. ### Hollywood’s Panic and the "Middle-to-Middle" Philosophy The creative world is in a state of "narcissistic injury" following the release of *Dance 2-0*, a Chinese AI capable of generating hyper-realistic videos of celebrities like Brad Pitt without their consent. While Hollywood unions are in "PLS" (distress), some see an opportunity. Matthew McConaughey’s approach is highlighted: since AI cannot be stopped, actors must trademark their "IP"—their voice, likeness, and signature catchphrases—to license them for AI use. The panel discusses the "Jevons Paradox," suggesting that as the cost of producing content drops, the volume of production will explode. In the video game industry, individual creators are already using AI to build complex RPGs that previously required massive studios. The final conclusion of the episode is the "Middle-to-Middle" theory. AI is now exceptionally good at the middle part of any process—the structure, the drafting, and the technical execution. However, humans remain essential for the "start" (the creative spark and original framing) and the "finish" (the soul, the imperfection, and the final human touch). The hosts urge listeners to stop resisting the automation of the "middle" and instead focus on becoming "creators" rather than mere "consumers" to survive the coming shift.