Redirected from Academic Headlines
The Stanford Institute for Human-Centered Artificial Intelligence (Stanford HAI), co-led by Feifei Li, has just released its Artificial Intelligence Index Report 2024.
The more than 300-page report, the 7th AI Index study published by Stanford HAI, tracks global AI trends in 2023.
"This is our most comprehensive report to date, and it comes at a significant time when the impact of AI on society has never been more apparent." ”
Vanessa Parli, director of the research program at Stanford HAI, said, "I think the most exciting advantage of AI research is the combination of these large language models with robots or agents, which marks a significant step forward for robots to work more effectively in the real world." ”
The AI Index Report 2024 can be downloaded at:
Hatps://iIndek.stanford.edu/up-content/upload/2024/04/is_ai-index-report-2024.pdf
Unlike previous years, Stanford HAI has expanded its research this year to cover fundamental trends such as technological advancements in artificial intelligence, public perception of the technology, and more. The new report reveals the top 10 trends in the AI industry in 2023:
1. Artificial intelligence outperforms humans in some tasks, but not in all.
AI has outperformed humans in several benchmarks, including in image classification, visual reasoning, and English comprehension. However, it still lags behind humans in more complex tasks such as competition-level math, visual common sense reasoning, and planning.
2. The industry continues to lead cutting-edge research on artificial intelligence.
In 2023, industry has produced 51 notable machine learning models, while academia has contributed only 15. In 2023, industry-academia collaborations also produced 21 notable models, a record high. In addition, 108 newly released base models are from industry and 28 from academia.
3. Cutting-edge models have become more expensive.
According to AI Index estimates, the cost of training state-of-the-art AI models has reached unprecedented levels. For example, OpenAI's GPT-4 is estimated to have used $78 million worth of computing resources for training, while Google's Gemini Ultra has a whopping $191 million in computing costs.
In comparison, some of the most advanced models released a few years ago, the original transformer model (2017) and RoBERTa Large (2019), cost around $900 and $160,000 to train, respectively.
4. The U.S. becomes a major source of top-tier AI models.
In 2023, 61 notable AI models originated from institutions in the United States, surpassing 21 in the European Union and 15 in China.
The U.S. also remains a top destination for AI investment. In 2023, U.S. private investment in AI totaled $67.2 billion, nearly nine times that of China.
However, China remains the biggest competitor to the United States, with the largest number of robot installations in the world, and the majority of the world's AI patents (61 percent) come from China.
5. There is a serious lack of reliable and standardized assessments of LLM liability.
The latest research from the AI Index shows that there is a serious lack of standardization in responsible AI. Leading developers, including OpenAI, Google, and Anthropic, primarily test their models against different responsible AI benchmarks. This practice complicates the work of systematically comparing the risks and limitations of top-tier AI models.
6. Generative AI investment is surging.
Despite an overall decline in private investment in AI last year, investment in generative AI has surged, increasing nearly eightfold from 2022 (about $3 billion) to $25.2 billion. Major players in the generative AI space, including OpenAI, Anthropic, Hugging Face, and Inflection, have all received a sizable funding round.
7. Data shows that artificial intelligence makes workers more productive and has a higher quality of work.
In 2023, several studies evaluated the impact of AI on the workforce, showing that AI can allow migrant workers to complete tasks faster and improve the quality of their output. These studies also suggest that AI has the potential to close the skills gap between low- and high-skilled workers. There are also studies that warn that the use of AI without proper supervision can have a negative effect.
8. Scientific progress is further accelerated thanks to artificial intelligence.
In 2022, AI began to drive scientific discovery. However, in 2023, more important science-related AI applications are launched – AlphaDev, which makes algorithm sequencing more efficient, GNoME, which facilitates the material discovery process, GraphCast, which provides an extremely accurate 10-day weather forecast in less than a minute, AlphaMissence, which successfully classifies about 89% of the 71 million possible missense mutations.
Today, AI can now do crude calculations that are difficult for humans to do, but are critical to solving some of the most complex scientific problems. On the medical front, new research shows that doctors can use AI to better diagnose breast cancer, interpret X-rays and detect deadly cancers.
9. The number of AI regulations in the United States has increased dramatically.
In 2023, there were 2,175 mentions of AI in the global legislative process, almost double the previous year. The number of AI-related regulations in the United States has increased significantly over the past year. In 2023, there are 25 AI-related regulations, compared to 1 in 2016. In the last year alone, the total number of AI-related regulations increased by 56.3%. Some of these regulations include copyright guidelines for generative AI materials and cybersecurity risk management frameworks.
10. There is a greater awareness of the potential impact of AI, but at the same time more anxiety.
A survey from market research firm Ipsos shows that in the past year, the percentage of people who believe that AI will greatly impact their lives in the next 3-5 years has risen from 60% to 66%. In addition, 52% expressed anxiety about AI products and services, up 13 percentage points from 2022.
In the U.S., data from the Pew Research Center (Pew) shows that 52% of Americans say they are more worried than excited about AI, up from 38% in 2022.
Attached: A letter from Ray Perrault, Co-Director of AI Index
Ten years ago, the world's best AI systems were also unable to classify objects in images at a human level. AI struggles with language comprehension and is unable to solve math problems. Today, AI systems routinely outperform humans on standard benchmarks.
In 2023, AI advancements are accelerating. Advanced models such as GPT-4, Gemini, and Claude 3 demonstrate impressive multimodal capabilities: they can generate fluent text in dozens of languages, process audio, and even interpret memos. With the advancement of artificial intelligence, it is also increasingly entering our lives. Companies are racing to build AI-based products, and the general public is increasingly using AI. However, there are still significant problems with current AI technology. It cannot reliably process facts, make complex reasoning, or interpret its conclusions.
Artificial intelligence faces two interconnected futures. First, technology is constantly improving and becoming more widely used, which has a significant impact on productivity and employment. There are good and bad uses for AI. Second, the application of AI is constrained by technical limitations. Either way, the government is paying more and more attention. The government is actively involved in encouraging the development of AI, such as funding university research and development and incentivizing private investment. The government is also committed to managing potential adverse factors such as impact on employment, privacy concerns, misinformation and intellectual property rights.
On the technology front, this year's AI Index reports that the number of new large language models released globally in 2023 has doubled from the previous year. Two-thirds of the models are open-source, but the highest-performing models come from industry players with closed systems.
Gemini Ultra became the first LLM to reach human level on the Massive Multitasking Language Understanding (MMLU) benchmark, where the model's performance has improved by 15 percentage points since last year. In addition, GPT-4 achieved an impressive average win rate score of 0.97 on the Comprehensive Language Model Overall Assessment (HELM) benchmark.
While global private investment in AI has declined for the second year in a row, investment in generative AI has risen sharply. Fortune 500 companies are mentioning AI more than ever on earnings calls, and new research shows that AI has significantly improved worker productivity. In terms of policymaking, AI is mentioned in the legislative process more than ever before. U.S. regulators passed more AI-related regulations in 2023 than ever before. Still, many people have expressed concerns about AI-generated capabilities such as deepfakes. The public is more aware of AI, and studies have shown that their response is also anxious.
Ray Perrault
AI Index
Reference Links:
Hatps://iIndek.stanford.edu/up-content/upload/2024/04/is_ai-index-report-2024.pdf
Hatpas://is. Stanford. edu/research/i-index-report
https://hai.stanford.edu/news/inside-new-ai-index-expensive-new-models-targeted-investments-and-more
HTTPS://High.Stanford.Seven/News/AI-Index-State-AI-13-Charts
https://hai.stanford.edu/news/ai-index-five-trends-frontier-ai-research