Fireside Chat, Conference Presentation
The Collapse of the Human Empire: Reflexions from the AGI | RAISE Summit 2024 | Paris
- Laurent Alexandre predicts General Artificial Intelligence (AGI) will emerge by 2030, citing his 2016 book The Fall of the Human Empire as a prophetic analysis of current developments.
- Charles-Edouard Boué forecasts a specific AGI timeline of August 15, 2038, while challenging the notion of "artificial" intelligence as an oxymoron, viewing it as a proxy for human cognition.
- The discussion identifies Geoffrey Hinton's assessment that there is a 10% probability AI could wipe out humanity by 2044.
- The immediate market trajectory is defined as an unsustainably crowded bubble driven by excessive capital, me-too startups, and unsustainable unit economics (high token acquisition costs versus zero revenue).
- Alexandre predicts the AI bubble will burst within six to eighteen months, leading to a mass exodus of players where only survivors will dominate the post-bubble market.
- The "bubble burst" will be followed by the rise of "techno-politics," a new paradigm characterized by aggressive regulation and the convergence of entrepreneurs, politicians, and technologists.
- A critical technical limitation identified is the transition of AI from a 2D digital environment to interacting with the physical 3D world.
- Current models face constraints regarding entropy management, where high openness risks generating biases (e.g., racism), while low entropy enforces robotic, repetitive behavior.
- Boué argues that AGI will likely result from an accident rather than a designed process, as current systems are limited to modeling the past rather than predicting the future.
- Alexandre provides a specific medical case study: Google's new AI model achieved 60% diagnostic accuracy compared to a human doctor's 30%, yet the combination of AI and doctor yielded only 52%.
- This example illustrates a potential "negative complementarity" where human intervention decreases the quality of AI-driven outcomes, posing an ethical challenge for professionals.
- Both speakers agree that society faces a complex synchronization problem between the rapid pace of AI technology and the slower adaptation of education systems and political structures.
- Boué expresses long-term optimism, predicting that while a "machine aristocracy" may emerge around 2040, humans will ultimately find ways to reassert control, similar to the historical example of Genghis Khan's technological advantage.
- The speakers conclude that the "fall of the human empire" is more likely to result from internal societal failures—such as elite corruption, disorganization, and lack of collective intelligence—than from machine hostility.
- Future success for humans will depend on the ability to navigate a world where machines become superior at data integration, forcing a difficult recalibration of human value and work.