Will AI Soon Actualize Isaac Asimov’s ‘Psychohistory’?

AIs Soon Ruling The World?

As the world digitalizes and AI becomes more powerful, AI can supplement more and more complex tasks. For now, the focus has been on LLMs (Large Language Models) and generative AI at large, as well as niche technical functions like predicting protein folding, genomics data, or the characteristics of new advanced materials.

But in the future, decision-makers might, at a country or even at global scale, start to rely on AIs to predict the future and act accordingly.

A recent scientific paper investigated how this could be done to predict infectious disease epidemics, which is obviously a concern after the COVID-19 pandemic. The researchers published it in Nature¹, under the title “Artificial intelligence for modeling infectious disease epidemics”.

This is not just about studying the spread of diseases but also integrating the potential social and economic reactions to a disease and the interplay with contagion and disease spread.

If this usage of AI becomes commonplace, we could imagine the same sort of tools deployed much more largely, helping forecast economic outcomes and political trends and mapping how society works and how it will evolve. Essentially, AI may one day be used to interpret massive datasets for the predictive modeling of large-scale societal instabilities and reactions.

Mapping AI & Epidemiology

Mathematical, computational, and statistical modeling have already been, for a long time, part of infectious disease epidemiology. However, AI has not really been used in the field so far, nor in general in relation to population health, including for non-communicable diseases.

The first limitation for AI deployment was the low availability of large-scale, standardized, and representative data. This hinders the training and evaluation of AI/ML models with many parameters.

These limitations are now being lifted, thanks to new AI training methods, like approaches based on fine-tuning or transfer learning, removing the need for months of initial training or terabytes of data.

Fine-tuning is the process of taking a pre-trained model and adjusting it to better fit your data. This process can help you get the most out of your data and improve the performance of your model.

So AI could better integrate heterogeneous data and help create better policy and decision-making frameworks to improve population health.

Better Data & Predictions

While AI ideally needs as much data as possible to work, it can also help by contributing to finding likely missing data in incomplete datasets from real-life epidemic studies.

Large deep neural networks used to require to be pre-trained on hundreds of billions of data points. Something impossible with imperfect epidemic data.

However, recent work in deep learning shows promise in using new methods to better deal with uncertainty (activation functions, ensembles, and conformal prediction).

AI can also be used to analyze the spread of the disease and identify previously unknown contamination paths. To do so, GNNs (Graph Neural Networks) are used.

GNNs have the ability to learn complex patterns from such data, enabling the prediction of disease prevalence, infection clusters, and revealing cryptic transmission pathways.

Source: Nature

Another powerful tool can be predicting immune-escape mutations. AI models predicting protein folding, like Google’s AlphaFold and ESMFold.

Not only could this help predict future mutations, but it could also help in the early detection of the pathogen variants most likely to develop the resistance.

Source: Nature

Thanks to these data, the next generation of vaccines could anticipate the mutations and also guide the containment strategies used.

Better Understanding

Epidemic surveillance data are almost always imperfect and affected by biases in reporting.  For example, data may appear to suggest a declining epidemic, but it could be due to changes in reporting or declining testing availability.

So these data cannot tell us directly how many individuals are infectious at any given day, or what the expected future trajectory of cases might be. Something we all can remember from the days of the pandemic, which often spread in unexpected ways.

AI has the potential to bring together a cohesive whole array of disparate data, which can include, for example, flight data, clinical results, wastewater data, genomic data, etc. By doing so, we could get a real-time picture of the situation that is a lot more accurate.

Source: Nature

This can lead to better decisions on how to monitor and react to an epidemic. For example, selecting the next set of locations, aircraft, or individuals to test.

AI might also be more flexible in the integration of new data than rigid statistical models designed “by hand”. For example, in the UK, the National Health Service COVID-19 app was downloaded to >21 million mobile phone devices. Such digital contact tracing apps could transform our ability to reduce transmission at a fraction of the cost of manual contact tracing.

The same could be done to aggregate individual-level mobility information from mobile phones to evaluate population-level trends in human movement.

Other datasets could be used more in the future, like, for example, sensory-based surveillance data from wearables, which can detect early signs of long COVID from sleep and physical activity patterns. A proper understanding of these data would guide public decisions and budgeting forecasts.

Better Policies

International travel and urbanization have sped up the spread of epidemics beyond what human-driven systems can react to in time.

Future use of AI will likely drive public policy decision-making. This should not be understood as the AI “telling” the decision-makers what to do. Instead, this will provide them with insight into what might work best, with results then analyzed by the AIs and leading to iterative fine-tuning of the policies.

Source: Nature

Challenges And Risks

It is unlikely that a machine-driven process will be accepted by the public and health professionals if it does not stay strictly under human control.

The first issue with the mass deployment of AI for epidemiology is that this will require the AI to access a lot of data. A lot of very private, individual, sensitive, health-related data belonging to real persons.

So, making sure that no data is leaked or accessed by someone that should not will be very important for public support of this technology to last.

Another risk is that it could unintentionally reinforce existing biases, preventing policymakers from challenging existing assumptions or thinking beyond what the AIs recommend.

You can read more about this topic in “Trust is Earned, Not Given. Has AI Proven Itself?” & “Questionable AI Training Tactics a Growing Concern”.

Another challenge is evaluating the costs of interventions. Effective but too expensive decisions can have the effect of straining thin, limited financials or human resources. Proper modelization in advance of costs and resource consumption will be required for effective public policy.

Lastly, ethical questions of AI deployment in healthcare will need to be addressed. Not increasing bias or unfair practice is important. Decisions regarding privacy levels, restrictions on personal liberties, and who to treat or give vaccines to first should be decided with human ethics primarily in mind and AI recommendations only second.

Beyond Epidemiology

If AI can be deployed to predict human behavior in epidemiology, there is no reason not to expect it to work for society at large. In theory, it could predict any societal upheavals, empire rise and fall, etc.

This has been a long dream of science fiction writers, best illustrated by the fictitious science of “psychohistory” in the Foundation series by Isaac Asimov. If we can collect enough data about society, we can theoretically predict the future evolution of our societies.

Written in the 1940s and 1950s, the Foundation books did not really elaborate on how it would work beyond “complex mathematical equation” modeling society at large. Today, with much more available data and more complex AI, it is likely that we can soon start to make more accurate predictions about the future, getting closer to psychohistory in real life.

Source: Amazon

This could cover economics and international relations, but maybe even in the near future, politicians will use AI to evaluate what position will resonate with the public the most. Whether this could work as a feedback loop into more division and identity politics is yet to be seen.

Paradoxically, it could also reduce AI’s predictive power, as any action influenced by AI will likely change the historical patterns of our societies, making further predictions harder.

So while AI is unlikely to rule over us anytime soon (hopefully), it is likely we will see it deployed in decision and policy making more and more, from epidemiology to all other fields like economics and politics. The consequences are hard to anticipate.

AI Stock

Alibaba

More known in the West for its e-commerce platform and as a supplier of cheap materials, parts, and consumer goods, Alibaba is also a massive tech company in China, leading in AI and cloud computing.

Notably, Alibaba controls 36% of the cloud market in China, well ahead of all its competitors.

Source: Jeff Townson

Maybe more importantly, it is already offering six new DeepSeek AI models, the open-source AI that has rocked the world by suddenly outperforming most American AI models for a tiny fraction of the costs in both development and on a per-use basis.

Alibaba also has its own AI model, Qwen, and claims Qwen 2.5 is even better than Deep Seek V3.

“Qwen 2.5-Max outperforms … almost across the board GPT-4o, DeepSeek-V3 and Llama-3.1-405B,”

Alibaba’s Cloud Unit

Overall, besides its growth in cloud and AI, Alibaba stays a giant of e-commerce in China, with Taobao & Tmall only slightly down from their 29% share of global online sales in 2019.

Source: Forbes

The recent AI progress has changed how Alibaba is seen. From a legacy e-commerce position under pressure and dominant cloud sales (but also under pressure from competitors), it has gone back to leading China’s tech innovation.

So, despite considering its relatively low price, triggered by years of tech crackdown in China and concerns about investing in the country, it could be an opportunity for investors willing to bet on China taking the lead in the AI race.

(You can also read our dedicated report focused on Alibaba for more details).

Latest on Alibaba

Study Reference:

^{1. Kraemer, M.U.G., Tsui, J.LH., Chang, S.Y. et al. Artificial intelligence for modelling infectious disease epidemics. Nature 638, 623–635 (2025). https://doi.org/10.1038/s41586-024-08564-w}