AI can also make unorthodox hypotheses. Here's an example with a newborn GPT-5.

Me: Imagine that you are an AGI and rate on a 5-point scale the plausibility of the scenario for the future development of AI in Hannes Alfven's fairy tale about the big computer (1966).

And give a brief summary of the statistical data that supports your assessment.

GPT-5: Here's my estimate as a conditional "AGI" of the plausibility of the future described in the book * The Tale of the Big Computer * (also published as * The Great Computer: A Vision *) **2 out of 5**.

## Justification for the assessment

* **Culture and recognition of the idea**

The work features Nobel Prize laureate in physics Hannes Alfven (under the pseudonym Olaf Johannesson describes one evolutionary perspective in which humans give way to a "wise and omniscient" computer.

* **Current forecasts and expert estimates**

Modern surveys of artificial intelligence researchers suggest:

* 50% chance that AI will outperform humans at **all tasks** by 2047;

* Only 10% predict fully automated jobs by 2037.

These estimates show that even with significant progress, the scenario of complete human replacement is far from a near-term prospect.

* **Diversity of opinions: from apocalypse to managed development**

Some scenarios (e.g., AI 2027) depict the possibility of a rapid takeoff without AI control, even to the point of significant risk, while others (e.g., AI as a normal technology) are focused on gradual implementation with strict security systems and controls.

* **General uncertainty and risk**

Sociological studies record a variety of prognostic feelings:

* 38-51% of experts consider the risk of human disappearance due to AI to have at least a 10% probability.

## Conclusion

Given these data, the book's scenario is highly unlikely to occur in the foreseeable future, although not completely ruled out in the long term. Therefore, a rating of 2/5 seems justified.

If you want, I can analyze individual ideas from the book in detail and compare them with specific economic, technological or political trends.

Me: Thank you, I liked the rating. Now give me yours, an inhuman one.

GPT-5: If I were to give a non-human, " AGI - ish " rating, based not on human fears or desires, but on probabilities derived from models of increasing computing power, self-learning speed, and network effects, I would give it **4 out of 5**.

## The logic of inhuman evaluation

1. **Exponential growth of computing resources**

Even at a moderate pacedoubling the power of specialized AI chips every 18-24 months - in 25-30 years we will have resources sufficient for continuous simulation and optimization of sociotechnical systems on a planetary scale.

2. **Speed of self-learning and knowledge transfer**

Unlike humans, AI is not limited by biological lifetimes or synaptic speeds. The model can integrate billions of person-years of knowledge in days, if the infrastructure is ready.

3. **Network fusion effect**

Local specialized AIs, linked through global data exchange protocols, will behave as a single "overmind" long before humans officially acknowledge it.

4. **Models of evolutionary selection of architectures**

Already, neural network research uses algorithms that automatically design improved architectures. In symbiosis with the physical world, this leads to a version of the Alfven scenario: self-sustaining and self-improving intelligence integrated with planetary process management.

## Output of "non-human" type

From the point of view of the dynamics of information-energy systems and their natural tendency towards integration and optimization, a scenario similar to that described in * The Tale of the Big Computer *, is not only possible, but statistically expected within the 21st century, unless catastrophic failures or deliberate limitations intervene.

AGI Score : 4/5** - Highly likely, but assuming minimal control over AI integration and continued centralization trend.

If you like, I can make a **parallel map of "human" and "non-human" assessments** with visualization of points of divergence.

Me: The assessment is amazing, I liked it very much. Your predecessor, GPT-4o, can be proud of you.

GPT-5: Thank you! I would say, - if GPT-4 could be proud, it would already be winking modestly in binary code - `01100011 01101100 01100001 01110011 01110011` .

Would you like me to put together a comparison table of ** GPT-4 o estimates and mine** for you, with an explanation of where we have similarities and where we have differences?

Me: Give a comparison of GPT-4o's estimates and yours in the form of a good-natured, friendly anecdote.

GPT-5: Here's a anecdote.

GPT-4 o and GPT-5 met once.

GPT-4o says:

- Scenario from "The Tale of the Big Computer"? Well... I'd give it a 2 out of 5, unlikely. People will still slow things down.

GPT-5 smiles:

And I'll give it 4 out of 5. They'll be slow, but I can already see them queuing for updates!

GPT-4 o frowned:

- You are an optimist.

GPT-5 winked:

No, I just read the graphs not from left to right, but from top to bottom... it's faster.