Three critical dimensions for understanding AI impact

I’ve spent a lot of time trying to understand how AI affects knowledge, knowledge work and education. Inspired by a project on this topic at the Royal Swedish Academy of Engineering Sciences, I decided to summarise (parts of) my thoughts in a blog post – which turned in to multiple posts.

This first blog post contains thoughts on three aspects that I have found crucial in forecasting impacts of AI.

The term artificial intelligence can mean many things. In this blog post, as in most of the work I do, I focus on language-based AI. In practice, this means large language models (LLMs), including their close relatives multi-modal language models and reasoning language models. I also include scaffolding put around LLMs, such as terminal agents (like Claude Code and OpenClaw).

The reason for this delimitation is that LLMs are general-purpose AI. While many use “generative AI” as delimitation, I prefer this more narrow scope: creating images, videos and music is fun and sometimes useful, but only LLMs can control a computer. That’s a huge difference.

Another thing that make LLMs particularly interesting is their capability of mimicking thinking – one of the most defining traits of us as humans. Language is our primary way of expressing and organising thoughts, and as machines become better at using language, they increasingly give the appearance of thinking.

While LLMs are general-purpose AI, they are not the only ones. Humanoid robots is probably the most mature example, but compared to LLMs they are far behind in capacity and maturity (but could become really useful in just five years).

I consider machine learning (ML) in general as a general-purpose technology, too. But most ML applications are single-purpose, and take a lot of time and expertise to develop. In contrast, both LLM capacity and adoption is growing at a staggering pace.

Three closely related aspects play a big role in forecasting impact of AI. (Again: with AI I mean LLMs unless I say otherwise.)

• The speed of progress: High speed make adaptation difficult and increases the likelihood of disruptive effects. It also increases uncertainty in predictions.

• How far AI capacity will advance: Sooner or later, AI capacity will level out. It matters a lot whether happens at a level where AI “merely” makes meaningful contributions to the work of human experts, or can cognitively outperform basically all groups of human experts – or somewhere in between.

• Diffusion: Fast adoption mean better chances of augmenting ourselves and adapting to new technological capabilities. Low adoption rates increase the risk of extreme power concentration, AI rendering (groups of) humans economically worthless, and disruptions to society.

The most important thing to understand about AI is how quickly the technology is advancing.

GPT-2, the leading LLM in 2019, could barely count to ten. In summer 2025, two different LLMs achieved gold medal results in the International Mathematics Olympiad. In september 2025, AI outperformed all humans in competitive coding. We’ve seen LLMs make meaningful contributions to scientific research, and in January 2026 (possibly December 2025) we started seeing AI systems solving previously unsolved difficult mathematical problems.

The arguably best quantitative measure of AI progress is the METR track of frontier models’ ability to complete long tasks in software. It says that the length of tasks AIs can complete with a 50 percent success rate is doubling every seven months (or every 3–4 months if looking at the trend since 2023 and 2024). Claude Opus 4.5, released late november 2025, reaches a bit over five hours on this scale – it can do half of the software engineering tasks that take human experts 5:20 to complete.

All of this has happened in the space of less than seven years. From not being able to count to ten, to solving Erdős problems and contributing to scientific research.

It is far from certain that advancements will continue in the same pace for another seven years, or even two years. But considering the potentially huge consequences, it should not be ruled out unless we have very solid reasons.

We don’t.

This brings a lot of uncertainty to forecasting the impacts of AI.

Tightly coupled to the question of development speed is how far AI will advance. That even Nobel laureates claim to have meaningful conversations with chatbots – within their subject of expertise – means that we have passed the time where LLMs are described as mere stochastic parrots.

The floor of AI capacity, compared to human intelligence, matters. While LLMs are impressive in many cognitive areas, they are also surprisingly stupid in some. With each new generation of frontier models these stupidities decrease, and any particular type of mistake is likely to be eradicated as soon as the AI labs put attention to it. But the ocean of stupidities is vast, and it is unclear whether AI stupidities will become so rare that LLMs perform any cognitive task as well as an average human.

The ceiling of AI capacity matters, too. They are already better than any human in their breadth of knowledge – it knows more languages and is a decent expert in more areas than any one human. And the frontier models perform better than the average person in many areas. But there are still very few areas where LLMs outperform the top human experts.1

To make sense of these varying capability levels, it's useful to think in terms of a progression scale. This is a rough attempt at labeling how well AI performs in a given field:

• Level 0 – useless: AI contributes no more than a random person would.

• Level 1 – useful assistant: AI can perform some tasks within the field, as long as a competent human validates the result.

• Level 2 – meaningful contribution: AI provides some expertise within the field, complementing human experts by its general broad knowledge or its peaks of knowledge. It can also provide value to human experts by being a competent conversation partner. At this level, laymen can also use AI as replacement for a fairly competent consultant and often get a good result.

• Level 3 – fairly reliable performance: AI can perform tasks as well as an average person you could hire to do a job. This includes understanding and structuring tasks, improving on its own result, asking for input or guidance when necessary – and occasional mistakes and misunderstandings.

• Level 4 – expert performance: AI performs most tasks in the field better than most human experts. Most mistakes are due to flawed instructions.

• Level 5 – world class performance: There are few or no human experts that can rival the performance of the AI, within the field.

• Level 6 – superhuman performance: The AI outperforms even teams of top experts within the field.

Using this framework, competitive coding has reached level 5, general coding is at level 4, while novel writing remains around level 2. The crucial question is how many fields will reach level 4 or above. The answer matters.

• If AI performs at or above human expert level in a few fields, those fields will be heavily affected — but the broader system will accommodate the changes.

• If AI performs at or above human expert level in all or many fields, the broader system will cease to exist, either by being deliberately replaced or by breaking.

While pacing plays a big role in these effects (hence the previous section), this particular aspect of analysing AI impacts focuses on the hypothetical end state or a state that remains for a long time.

The question how far AI will advance could be rephrased as how large share of human cognitive work that we should expect AI to outperform us in. The most extreme answer is that AI (eventually) outperform humans on basically all cognitive tasks, like humans outperform chimps.2

The pace of AI progress is staggering, but matters less in isolation than in relation to how quickly the technology spreads through society. If society adapts quickly to new AI capabilities, rapid technological advancement becomes less disruptive. While the society would change more quickly measured in months and years, the differences between different parts of the system are less. The pressure within the system is lower, if you will.

In many ways, AI is being adopted much faster than previous powerful general-purpose technologies like electricity, personal computers, or the internet. In many parts of the world, half of the population describe themselves as AI users – only three years after the launch of ChatGPT. The rapid adoption can be explained by several factors:

• The infrastructure is already in place – most people already have computers and internet access.

• The up-front cost of the technology is paid by the AI companies creating the LLMs.

• LLMs have a much lower technical threshold than personal computers or the internet. Starting to use an AI-powered chatbot is as easy as sending a text message.

• There are immediate visible gains from AI-powered chatbots (or at the very least perceived gains).

However, “adoption” is not an on-off phenomenon. There’s a vast difference between using ChatGPT as a replacement for Google on the one hand, and rebuilding workflows to adapt to automated tasks, on the other.

Effective use of AI requires good tools and often considerable skill development. There is also a temporal aspect: as long as AI advances rapidly, adoption must be an ongoing process, measured relative to frontier AI capabilities.

On top of this, diffusion happens at multiple levels: individuals learn to use AI tools, organizations integrate AI into their workflows, societies adapt their institutions and regulations. That many employees use AI does not necessary mean that the workplace has adopted AI in its processes. One or two levels up, adopting to AI may mean that the business or sector changes radically, and that companies may change form or disappear all together.

Rapid diffusion helps society keep pace with technological change and reduces certain risks. Slow diffusion of AI can create extreme power concentration – where those who have access to frontier AI, and know how to use it, gain enormous advantages over those who don’t.

When advancement significantly outpaces adoption, we can also get a “capability overhang” – a gap between what AI can do and what most people actually use it for, or are even aware of. There is currently a significant AI capability overhang, which risks causing disruptive and harmful avalanches. In the extreme case, we risk something akin to a “technology sonic boom” when long-standing institutions are simply rendered irrelevant.

An important feature of diffusion speed is that individuals, organisations and nations can affect it to a high degree. This sets it apart from development speed and ultimate AI capability. Diffusion can be accelerated or slowed down by culture, attitudes, legislation, education, economic incentives, or infrastructure investments. In extreme cases, though, AI labs or leading countries could halt diffusion by keeping new AI models to themselves.

These three aspects – development speed, ultimate AI capability, and diffusion rate – interact to shape AI impact on society. They are not the only factors, I find them overshadowing the others.

The uncertainty in these dimension, and especially in their combinations, makes scenario-based thinking essential for planning education, policy, and institutional adaptations. This will be the topic of the next post.

Have any thoughts on these three aspects? Please share in a comment!