Everyone says that AI is revolutionary technology. But what does that really mean? The answer comes in the AI Index 2025 report published by Stanford University. The conclusions are both exciting and alarming: AI has surpassed doctors in diagnostic reasoning, computer scientists in programming, and graphic designers in the speed of creating videos and images. What’s more, the cost of AI models has fallen 280-fold. Still, complex reasoning remains an area where humans retain the advantage.
Stanford University’s annual AI Index Report makes it clear that the AI revolution is no longer theoretical. Over the past year, AI has broken through barriers once thought unattainable: it has learned to program and generate high-quality video, it understands human language with increasing fluency, and it reasons at a near-doctor level in medicine. It can combine information from multiple sources and instantly identify correlations, for example, between lifestyle and disease.
AI will be the core technology for medicine
For the first time, Stanford researchers conclude that AI in medicine is no longer experimental or merely complementary, but is becoming an essential, foundational technology.
OpenAI’s o1 model achieved 96.0% accuracy on the MedQA (Medical Question Answering) benchmark test in 2024, a record that is 5.8 percentage points higher than the previous best result. Since 2022, accuracy on this benchmark has risen by more than 28 percentage points. In clinical studies, GPT4 achieved 92% diagnostic accuracy, outperforming physicians (74%). Importantly, combining doctors with AI did not always improve results, indicating that integrating AI into clinical practice remains a challenge.
AI systems already outperform radiologists in detecting cancer and predicting patient mortality risk. Crucially, the best outcomes were achieved through collaboration between humans and AI—pointing to a future in which medical decision-making should always remain a joint effort.
By 2023, the number of AI-enabled medical devices approved by the U.S. Food and Drug Administration (FDA) had reached 223, compared with only 6 in 2015. Most approvals were in radiology, followed by cardiology, neurology, and anesthesiology.
In 2024, a wave of specialized medical foundation models emerged. Google released Med-Gemini, a tool tailored to the healthcare industry. Other models included EchoCLIP (echocardiography), ChexAgent (radiology), and VisionFM (ophthalmology). Their strength lies in training on domain-specific data, making them more accurate in their respective areas.
AI in medicine also faces a shortage of real-world data, much of which is locked in local databases. Here, synthetic data generated by AI is already helping in drug discovery and in identifying social determinants of health, areas where traditional medicine often struggled due to data scarcity.
A landmark recognition came in 2024, when Demis Hassabis and John Jumper received the Nobel Prize in Chemistry for their work on protein folding with AlphaFold, while Geoffrey Hinton and John Hopfield were awarded the Nobel Prize in Physics for foundational contributions to neural networks.
As AI’s role in medicine grows, so does concern for its ethical development and deployment. The number of publications on AI ethics in medicine nearly quadrupled, from 288 in 2020 to 1,031 in 2024. Key debates center on equal access, privacy, and accountability in clinical practice.
AI breaks down barriers and drives robotics
Beyond medicine, AI is advancing across other disciplines. In general knowledge tests across multiple languages (MMMU), AI performance increased by 18.8 percentage points, from 34.3% in 2023 to 53.1% in 2024. In a scientific reasoning test at the master’s level (GPQA), results improved by nearly 49 percentage points.
The most dramatic progress came in programming. On the SWE-bench test, which evaluates real-world software engineering tasks, AI effectiveness leapt from 4.4% in 2023 to 71.7% in 2024, thanks to its growing capacity for solving complex problems.
The acceleration of AI is driving progress in other technologies as well. Autonomous vehicles are no longer experimental: Waymo now provides over 150,000 driverless rides per week in the U.S., while Baidu’s Apollo Go robotaxis serve passengers across numerous Chinese cities.
Private investment in AI is also surging. In the U.S., investment nearly doubled from $67.2 billion in 2023 to $109.1 billion in 2024. At the same time, AI adoption in business climbed from 55% in 2023 to 78% in 2024. Companies report measurable gains in customer service, marketing, logistics, and software engineering. Studies confirm that AI boosts productivity and helps bridge the skills gap between experienced and less experienced employees.
The battle of the AI giants: US vs. China
The U.S. remains the global leader in producing frontier AI models, publishing 40 notable models in 2024, compared to 15 from China and three from Europe. Chinese models, however, have quickly closed the performance gap, achieving near parity with U.S. systems on key benchmarks.
China also dominates in patents: as of 2023, it accounted for 69.7% of global AI patent grants. This signals not only scientific competition but also the geopolitical stakes of AI-driven economic power.
Another striking trend is the concentration of AI innovation in private hands. Nearly 90% of top-performing models in 2024 came from private companies, up from 60% in 2023.
Generative AI is also spreading rapidly among the general public. ChatGPT is used by roughly 300 million people each month. Global attitudes are shifting: in a 2024 survey, 55% of respondents said AI is more beneficial than harmful, a two percentage point increase since 2022. Optimism is highest in China (83%), Indonesia (80%), and Thailand (77%).
AI is a technology that is changing the world
Another notable trend is the falling cost of advanced AI. Running a model comparable to GPT-3.5 costs $20 per million tokens in late 2022, but by October 2024, it had dropped to just $0.07; a 280-fold reduction. This democratizes access for researchers and startups.
At the same time, smaller models are becoming more efficient. Microsoft’s Phi-3-mini (3.8 billion parameters) now matches the performance of models more than 100 times its size.
The authors of the AI Index 2025 Report conclude that we are at a historic turning point in the development of AI. The unprecedented pace of progress demonstrates how rapidly AI is advancing into domains once reserved for humans.
Just a few years ago, IT specialists were in extremely high demand; today, AI systems can already solve many of their tasks. And yet, the report stresses that AI is not only a technology—it is also a force reshaping medicine, geopolitics, and society. The challenge for governments is to ensure that the benefits of this revolution are shared broadly, while mitigating risks such as inequality and job displacement.
