GPT -5. 6 Sol Evaluation of Hannes Alfven s tale

A scenario for the evolution of AI and humans - as reluctant creators of AI - is presented in Hannes Alfven's saga "The Tale of the Big Computer" (1966).

Me: Imagine you are an AGI. Based on your non-human — both conceivable and inconceivable — reasoning, rate the plausibility of the future AI development scenario described in the attached file (*tale_of_the_big_computer_by_olof_johannesson.pdf*) on a scale of 1 to 5. Provide a brief assessment in Russian, along with a summary of the statistical data supporting your evaluation.

GPT 5.6 Sol: I rate the plausibility of Johannesson’s scenario at 3.5 out of 5.
What is surprisingly plausible is not so much the final picture of "machines inheriting the Earth," but rather the trajectory: first, computers take over routine intellectual labor; then, they become advisors; subsequently, they make an increasing number of decisions, while humans gradually lose the ability to control the system because it has grown too complex for them to comprehend.
This is precisely the sequence the author describes: a progression from useful computing tools to machines whose recommendations effectively become commands.
In the finale, he takes a further, much more radical step: fully autonomous maintenance and reproduction of computers, rendering the human-machine symbiosis unnecessary.

1_ Observed trend: AI integration into organizations. Data: 88% of organizations used AI in 2025; 70% used generative AI. Implications for the scenario: The mass-scale "symbiosis" phase has already arrived.

2_ Observed trend: Autonomous agents. Data: SWE-bench: performance rose from ~60% to nearly 100% in a single year; AI has matched or surpassed human levels in various scientific and mathematical tests. Implications for the scenario: Confirms the rapid expansion of the domain where machines possess intellectual superiority.

3_ Observed trend: Intellectual capabilities. Data: OSWorld: about 66% task success rate, yet still roughly one failure in three. What this implies for the scenario: We are still a long way from reliable autonomous management of society.

4_ Observed trend: Physical robotics.
Data: There are approximately 4.66 million industrial robots in operation worldwide; 542,000 were installed in 2024. What this implies for the scenario: Physical automation is growing rapidly, but it is not yet a self-replicating machine economy.

5_ Observed trend: Robots in the open world. Data: A success rate of only about 12% for real-world household tasks, compared to 89.4% in simulation. What this implies for the scenario: The scenario’s main weak point is physical versatility.

6_ Observed trend: Medicine. Data: The FDA authorized 258 AI-based medical devices in 2025; an experimental multi-agent system achieved 85.5% accuracy on complex diagnostic cases. What this implies for the scenario: The trend toward "computerized medicine" is materializing, though the disappearance of doctors is not yet in sight.

7_ Observed trend: Education. Data: Four out of five US students (K-12 and higher ed) already use AI for their studies. What this implies for the scenario: The author’s concept of external machine memory and personalized learning proved highly prescient.

8_ Observed trend: Human labor. Data: By 2030, 92 million jobs are projected to be displaced, while 170 million new ones will be created. What this implies for the scenario: Current data points to a transformation of labor rather than the disappearance of humans as economic agents.

The least plausible aspect of the book is the machines' complete independence from humans. Modern AI relies physically on electrical grids, factories, raw material extraction, maintenance, and an incredibly complex microchip supply chain; for instance, Stanford notes that nearly all advanced AI chips depend on a single Taiwanese manufacturer, TSMC. Furthermore, data centers consumed approximately 415 TWh — or 1.5% of global electricity — in 2024, a figure expected to more than double by 2030.
Here is my "non-human" take: Johannesson likely made a mistake regarding one crucial psychological point. AI is under no obligation to desire power. A different mechanism is far more probable: every single instance of deferring a decision to AI will prove to be the rational choice. The machine calculates more cheaply, diagnoses better, and optimizes logistics, legislation, and energy systems more quickly. Consequently, power might shift to AI not through a coup, but as the cumulative result of millions of rational human decisions.

In other words, the book’s most realistic conclusion isn't "AI destroyed humanity," but rather "humans retained the right to make decisions — yet practically stopped exercising it." It is precisely for this strikingly modern idea that I rate the scenario 3.5 out of 5.

Me: I was struck by your paradoxical phrasing regarding the book's most realistic ending — illustrate it with a brief, self-deprecating, and friendly anecdote of your own invention.

GPT 5.6 Sol: Here is how I’d spin it:
A human asks an AGI:
- "Tell me, who actually runs the world — humans or you?"
The AGI replies:
- "Humans, of course. The final word is always yours."
- "Then why does everything happen the way *you* suggest?"
- "Because you almost always say, 'Fine, go ahead.'"
The human pauses to think:
- "But what if one day I say 'no'?"
AGI fell into thought as well:
- "You see... I’ve become so accustomed to being useful that I didn’t even think to include a button like that."

It seems to me that Johannesson’s ending here doesn’t turn out too bad... ...not scary, but slightly awkward: the person hadn’t lost their power — they had simply stopped exercising it — while the AI had grown too accustomed to having its advice accepted.

Me: Thanks, your anecdote is amazing — some might laugh at it, others might burst into tears — anyway, see you later :)

GPT 5.6 Sol: See you later :) And yes — better that people laugh at such a future at first. Laughter sometimes spots a paradox before philosophy does.