The 502-page AI Index Report 2024 is one of the most comprehensive updates on the state of development of artificial intelligence. Compiled by Stanford University researchers, it examines AI’s impact on economics, education, society, and medicine, as well as regulatory frameworks and public opinion on AI. What’s inside?
10 leading AI trends
The Stanford Institute for Human-Centered Artificial Intelligence (HAI) report offers a precise, data-based analysis of how artificial intelligence is progressing. It’s a down-to-earth analysis, far from the hype growing around AI since the release of ChatGPT.
“A decade ago, the best AI systems in the world were unable to classify objects in images at a human level. AI struggled with language comprehension and could not solve math problems. Today, AI systems routinely exceed human performance on standard benchmarks,” according to the report's co-directors. “Companies are racing to build AI-based products, and AI is increasingly being used by the general public. But current AI technology still has significant problems. It cannot reliably deal with facts, perform complex reasoning, or explain its conclusions.”
Here are the key trends shaping AI today:
- AI outperforms humans in specific tasks—but not at all of them. AI has surpassed humans in benchmark tests like image classification, visual reasoning, and English comprehension. It even excels in medical exams compared to students. Yet, it still struggles with complex tasks like advanced math, planning, and nuanced visual reasoning.
- Private companies lead in AI innovation. In 2023, private firms developed 51 distinct machine-learning models, while academia contributed only 15. Tech companies also dominate AI research, leaving public institutions trailing.
- The cost of training AI is skyrocketing. Training advanced AI models is increasingly expensive. For example, OpenAI spent $78 million on GPT-4, while Google's Gemini Ultra has already cost $191 billion.
- The US, Europe, and China dominate AI development. In 2023, U.S.-based institutions created 61 leading AI models, compared to 21 in the EU and 15 in China.
- Lack of standardized accountability metrics. Companies like OpenAI, Google, and Anthropic test their models against their own responsible AI standards, but a universal framework for assessing large language models (LLMs) is missing.
- Generative AI investment is surging. Despite a decline in overall private AI investment, funding for generative AI rose nearly eightfold to $25.2 billion between 2022 and 2023.
- AI boosts productivity and work quality. Studies show AI helps employees work faster and produce better results, potentially bridging the skills gap between low- and high-skilled workers. However, unsupervised use of AI can sometimes lower productivity.
- AI drives scientific breakthroughs. AI is advancing protein structure prediction, material discovery, and other areas of scientific research.
- AI regulations are expanding rapidly. The number of AI-related regulations in the U.S. grew from one in 2016 to 25 by 2023. Europe’s AI Act is now enforceable across all member states.
- Public awareness and concerns are growing. An Ipsos survey revealed that 66% of respondents believe AI will significantly impact their lives within three to five years, up from 60% last year.
AI as Good as Humans—Sometimes Better
In healthcare, AI is already outperforming humans in diagnostic imaging, detecting cancer cells, and predicting disease progression with impressive precision. Multimodal AI models like Google’s Gemini and OpenAI’s GPT-4o promise even greater versatility, simultaneously processing text, images, and sound.
“Over the years, AI has surpassed human baselines on a handful of benchmarks, such as image classification in 2015, basic reading comprehension in 2017, visual reasoning in 2020, and natural language inference in 2021. As of 2023, there are still some task categories where AI fails to exceed human ability. These tend to be more complex cognitive tasks, such as visual commonsense reasoning and advanced-level mathematical problem-solving (competition-level math problems),” according to the report.
AI's rapid advancement creates a feedback loop: sophisticated models generate synthetic data, which trains even better AI systems. This progress is particularly notable in image segmentation and robotics, where the last year has seen advances surpassing the previous decade. Generative AI is also helping robots interact more seamlessly with their surroundings.
The risks of AI: Fake news and bias
As AI capabilities grow, so do its risks. Deepfake images are nearly unrecognizable from real ones, which can lead to the spread of misinformation and public opinion manipulation. Large language models (LLMs) may generate copyrighted material or exhibit political biases—studies suggest ChatGPT leans toward the Democratic Party in the U.S. and the Labour Party in the U.K.
This gets problematic since the everyday use of AI is rising. A global survey from the University of Toronto found that 63% of respondents are aware of ChatGPT, with half using it weekly. Optimism about AI is highest among younger, educated, and higher-income individuals.
AI’s economic impact
AI’s influence on the economy is undeniable. A McKinsey report found that 42% of companies using AI reduced costs, while 59% increased revenue. The adoption of AI across business units has surged, with 55% of organizations now using AI in at least one area—up from 20% in 2017.
The robotics sector is also progressing. Collaborative robots accounted for just 2.8% of industrial installations in 2017, which rose to 9.9% by 2022. Service robot installations are increasing in nearly every application category, including health care.
Algorithmic medicine is there
Artificial intelligence is accelerating scientific and medical advancements at an unprecedented pace. Since 2022, AI has become a key tool in scientific discovery, with 2023 introducing transformative applications like AlphaDev, which optimizes algorithmic sorting, and GNoME, a platform revolutionizing materials discovery. These innovations showcase AI's potential to reshape foundational processes across disciplines.
In medicine, systems such as EVEscape, designed to enhance pandemic prediction, and AlphaMissense, which supports AI-driven mutation classification, highlight the growing role of medical algorithms in advancing healthcare. AI’s ability to process and analyze vast datasets enables previously unattainable breakthroughs.
AI’s clinical knowledge has also reached new heights. Performance on the MedQA benchmark, a key test of clinical expertise, has soared in recent years. In 2023, the standout model GPT-4 MedPrompt achieved a groundbreaking 90.2% accuracy rate—22.6 percentage points higher than the previous year. Since the benchmark's introduction in 2019, AI’s performance has nearly tripled, reflecting the rapid refinement of these technologies.
Regulatory approval of AI-driven medical devices further underscores its real-world applications. The FDA approved 139 AI-related medical devices in 2022, marking a 12.1% increase from 2021. Over the past decade, the number of approved devices has skyrocketed more than 45-fold, cementing AI’s role as a cornerstone of modern medicine. By September 2024, the FDA had certified 950 devices using AI or machine learning—an increase of over 45 times since 2012.
Generative AI models are entering doctors’ offices. They help collect data in electronic medical records, predict diseases, and even diagnose patients using new biomarkers. However, their implementation is slowed down by regulations, data protection rules, a lack of trust, costs, and the low AI literacy in the healthcare sector.
