Public AI: Diffuse Impacts Over Decades

The impact of AI on daily life is described as weak and primarily indirect, far from the tangible benefits of industrial revolutions. The essence of the current dynamic seems to be concentrated in infrastructure and intellectual work, with long diffusion timelines and a risk of political backlash fueled by the history of Big Tech.

Social, Political Pressure, and Diffusion Timeline

The AI industry is evolving under rapid social expectations, with many eyes focused on its promises. If it were granted a 100-year timeline for diffusion, its effects would certainly become more visible, akin to past industrial revolutions. The current diffusion is presented as the first half of a decade in a process estimated to take 50 years. Two obstacles are highlighted: initially too indirect benefits and a political backlash related to the history of Big Tech in the West. Addressing either of these points would reduce pressure and allow more time to establish a positive case for changing the status quo. The situation is complicated by the frequent portrayal of AI as dangerous or negative, evoking catastrophe and mass unemployment. Prominent figures are beginning to address this issue, but more efforts are deemed necessary to gain public support.

Little Perceptible Concrete Gains in Daily Life

Family, food, transportation, and entertainment are still minimally affected by AI. Current touchpoints are marginal, not very beneficial, or confusing, as illustrated by the incident between OpenAI and HuggingFace that many have only heard about. The uses highlighted by optimists—image generation, enhanced online search—are considered limited, while negative associations dominate: addictive social media algorithms, ubiquitous chatbots for some, and debates around data centers. Personally, the author indicates having used AI only for research and occasional creative work.

Perceived Gap Between Digital Elites and the Rest of Society

Current artificial intelligence primarily serves as a tool for a privileged minority. Regarding intellectual work, which constitutes about half of the U.S. economy, it is seen as essential as electricity, and rapid advancements in agents are expected in the next 18 months. The fact that such a productive instrument only affects a portion of the population is described as highly disruptive and accentuates the impression of a thriving tech economy while the daily lives of the rest of society seem stagnant. AI is also viewed as a lever to grow tech companies and create small online businesses. The author does not expect a rise in tech employment despite massive success; the workforce could even decrease as the production of intellectual work increases significantly. This context, in already high-performing sectors, would tarnish the image of AI as a collective good, with the risk of a stalled trajectory reminiscent of American nuclear energy.

Infrastructure Today, Benefits Tomorrow

The current stage is described as a construction of infrastructure and processes that accumulate over decades. A major breakthrough today might seem minor in light of future advancements along this cumulative path, and it remains difficult to anticipate how each everyday technology will benefit from these improvements. Optimistic perspectives include scientific discoveries and therapies for rare diseases, but these benefits are deemed too indirect for the public to easily attribute them to AI actors. Meanwhile, autonomous driving is progressing on a trajectory presented as largely independent of innovations driven by LLMs.

Lessons from Industrial Revolutions and the Engels Pause Concept

Previous major industrial revolutions gave rise to tangible material goods: in the 18th century, it was about affordable clothing, utensils, books, and transformations in lifestyles; in the 19th century, household appliances like the sewing machine, preserved foods, indoor plumbing, photography, improved lighting sources, and the bicycle emerged. These innovations were material in nature and remain in use. The "Engels Pause" refers to the period from 1790 to 1840, during which British working-class wages stagnated despite rapid GDP per capita growth. If this analogy applies to AI, those who do not benefit from it would be justified in opposing it. Furthermore, comparing current AI to these revolutions ignores the lack of immediate tangible goods and a social inertia deemed stronger today.

Scenarios of Materialization Through Robotics and Assistants

Robotics and autonomous driving could merge with the narrative of AI if mass production of LLMs accelerates the arrival of robots in daily life, offering tangible benefits. This scenario coexists with a paradox: although LLMs are often distinguished from previous advancements, the same dynamics could later either save or severely undermine them. The current period is presented as a growth phase, requiring the resolution of issues predating ChatGPT to unlock a long-term trajectory. Diffusion will take much longer than the controversy surrounding it. It is suggested that adoption could rise from 0% to over 90% over a lifetime. An AI deeply integrated into businesses and in the form of personal assistants is beginning to become viable, but it will take longer to establish than applications like ChatGPT. Additionally, mathematical or research advancements, such as solving a Millennium problem, might not engage the general public much, while daily life could, according to estimates, resemble today’s in 50 years, even if AI received significant credit in the meantime.