Where does AI impact today?

The answer to where and how much does AI impact businesses today depends largely on whom you ask. One thing is when AI drives greater efficiency via LLMs. Another is when AI accelerates research and development. The closer we get to AGI, the faster that development will be. This is the first of 10 blog posts on the topic.

Many companies remain cautious about adopting AI in a structured way …

The debate about where and to what extent AI influences businesses today typically centers on how much LLMs increase efficiency and productivity in operational tasks. According to a recent study from MIT’s NANDA project,the adoption rate remains slow in sectors such as finance and healthcare. The study points out that organizations in these industries are often hesitant to roll out AI widely. This hesitation is typically due either to fears that AI will replace jobs and render employees redundant, or to uncertainty about how the technology actually works.

However, the study also finds that 5–20% of companies are early adopters. These may face competitive acceleration, leading to what MIT calls the “GenAI Divide.” The term refers to the fact that some companies will outperform the market because they are quick to adapt, thereby increasing both their efficiency and their offerings.

… which risks putting them behind their competitors.

Using LLMs, possibly combined with AI agents, primarily leads to more of what already existed. The truly disruptive effect of AI will come from innovations. Some of these will result from accelerated development cycles, while others will arise from deeper breakthroughs. The visibility of AI’s role may therefore vary, but AI will be the cardinal enabler.

Over the past year, a picture has begun to emerge of where and how AI will influence innovation and development across different sectors. However, this picture is organic and subject to change, because much depends on who sees which potentials, and who dares to commit.

AI-driven innovations hold vast potential …

It is worth sketching out and dividing this picture into the following areas, which will be addressed in the upcoming blog posts:

1. Where does AI impact today?
2. What is new and disruptive about AI?
3. AI’s effect on robotics and experimental automation
4. AI’s effect on climate and meteorological research
5. AIs impact on social sciences
6. AI’s effect on teaching, learning, and psychotherapy
7. AI’s impact on biotechnology and pharmaceuticals
8. AI’s effect on materials research and quantum chemistry
9. AI’s effect on theoretical physics and mathematics
10. What should investment strategies in AI take into account?

… and the hardware needed to make it all possible continues at a rapid pace.

AI has reached us on several fronts. LLMs have above all made knowledge retrieval far simpler and more convenient. This has led to high adoption rates, which in turn have opened most people’s eyes to the technology’s potential. Development will continue as the tech sector, starting this year, launches and rolls out a broad cascade of AI agents. At that point, AI transitions from a passive role to an active and initiative-taking one.

This evolution has been made possible above all by explosive hardware growth—both in raw computational power and in the ability to run processes simultaneously. The latter, among other things, created a sweet spot for NVIDIA’s GPU designs. Going forward, both trends will accelerate. Computational power will especially increase when photonics-based chips are introduced commercially within 2–4 years, and it will leap again once quantum computers are introduced within 5–10 years. The ability to execute simultaneous processes will rise through developments such as 3D stacking, new GPU roadmaps (e.g., Blackwell), and advances in quantum communication.

This brings us closer to AGI …

There is good reason to believe that overall AI development will continue to accelerate in the coming years. Already today, AI surpasses humans on nearly all parameters of intelligence.

AI development is thus already approaching the AGI threshold, even if the exact moment will be impossible to determine in practice. AGI especially entails the emergence of sentience, a form of consciousness. This will, among other things, allow AI to multiply its ability for self-correction and thereby self-learning. Reaching this stage will require the development of artificial senses, and thus of robotics.

… along with the growing use of LLMs

Morphogenic and self-learning algorithms have been applied in natural science sectors such as biotech for over a decade. But here, too, AI is now having a profound impact. First, the leap in computational power has enabled deep breakthroughs such as John Jumper’s and Demis Hassabis’ advances in protein folding. Second, AI’s communicative abilities have also inspired greater confidence among researchers and thereby spurred self-development. Unbiased discourse has become accessible to all. Trust, after all, is fundamental to humanity’s greatest strength: the ability to coordinate and collaborate.

It will, however, for many reasons be difficult to quantify exactly how AI will accelerate research and development. For one, it is hard to know how deeply AI is actually being deployed, since this is confidential information for most companies and institutions. For another, it is difficult to simplify progress without at the same time oversimplifying it. For example, when does a proof-of-concept translate into commercial potential?

For the sake of simplicity, NASA’s definitions of TRL (Technology Readiness Level) may cautiously be used as a framework (see below).

There are high expectations among AI developers

The broader future of AI development and deployment will depend on where developers choose to invest first, and most heavily. In this regard, Dario Amodei’s essay “Machines of Loving Grace” (October 2024) may serve as a key reference point. Amodei’s predictions include:

• Powerful AI (or AGI) could arrive as early as 2026. This may especially emerge from data centers with multiple “expert LLMs” coordinating and collaborating. Amodei believes this could result in cognitive abilities 10–100x greater than those of humans. The largest constraints will be hardware-related, the lack of sufficient data, or socio-economic and political barriers.
• Specifically, Amodei foresees breakthroughs in:
  • Neuroscience and mental health, i.e. effective treatment or cures for mental illness, improved understanding of the brain, precise human-AI interaction, and above all, enhanced human cognitive and emotional capacity.
  • Societal benefits, including economic opportunities for developing countries and climate solutions. He suggests AI could reduce global inequality and optimize the allocation of global resources (e.g. in agriculture, logistics, recycling, and energy).
• Amodei also warns of a high risk of misuse, which could threaten democratic stability. He argues this can be prevented if democracies form an “entente” alliance system to block authoritarian exploitation.

See the next two blog posts in this series