The State of Artificial Intelligence: Stanford HAI releases its latest AI Index report

The Stanford Institute for Human-Centered Artificial Intelligence today released the latest edition of its AI Index report, which examines developments in machine learning over the past year.

Stanford HAI, as the institute is commonly known, was founded in early 2019. It explores new AI methods and also examines the impact of the technology on society. It publishes its AI Index Report annually.

The new edition of the study published today comprises more than 350 pages. It covers a long list of topics, including the cost of AI training, efforts to mitigate bias in language models, and the technology’s impact on public policy. In each area examined, the report points to several notable milestones achieved over the past year.

AI advancements and challenges

The most advanced neural networks have gotten more complicated over the past year. As an example, Stanford HAI refers to Google LLC’s Minerva large language model. The model, which debuted last June, has 540 billion parameters and required nine times more computing power than OpenAI LP’s GPT-3 to train.

The growing hardware requirements of AI software are reflected in the rising costs of machine learning projects. Stanford HAI estimates that PaLM, another Google model released last year, cost $8 million to develop. That’s 160 times more than GPT-2, a predecessor of GPT-3 that OpenAI released in 2019.

Although AI models can perform significantly more tasks than they could a few years ago, they still have limitations. These restrictions cover several different areas.

In today’s report, Stanford HAI highlighted a 2022 research paper that found that advanced language models struggle with some brain teasers. Tasks that require planning are often a particular challenge for neural networks. Over the past year, researchers have also identified many cases of AI bias in both large language models and neural networks optimized for image generation.

Researchers’ efforts to address these issues came to the fore in 2022. In today’s report, Stanford HAI highlighted how a new model training technique called instruction tuning has shown promise as a way to mitigate AI bias. Introduced by Google in late 2021, instructional training involves rephrasing AI prompts to make them more understandable for a neural network.

New use cases

Over the past year, researchers have not only developed more powerful AI models, but also found new applications for the technology. Some of these uses led to scientific discoveries.

In October 2022, Google’s DeepMind machine learning unit introduced a new AI system called AlphaTensor. DeepMind researchers used the system to develop a more efficient way to perform matrix multiplication. Matrix multiplication is a mathematical calculation that makes extensive use of machine learning models to turn data into decisions.

Over the past year, scientists have also used AI to support research in a number of other areas, Stanford HAI pointed out. One project demonstrated that AI could be used to discover new antibodies. Another project, also led by Google’s DeepMind, led to the development of a neural network capable of controlling the plasma in a nuclear fusion reactor.

The societal impact of AI

Stanford HAI’s new report also devotes several chapters to the impact of AI on society. Although large language models have only entered the public consciousness in recent months, AI is already making an impact in several areas.

In 2021, only 2% of federal AI-related bills proposed by US lawmakers were enacted. Last year, that number rose to 10%. At the state level, 35% of all AI-related bills will be passed in 2022.

The impact of machine learning is also being felt in the education sector. According to Stanford HAI study, as of 2021, 11 countries have officially endorsed and adopted a K-12 AI curriculum. Graduates from US universities specializing in AI almost doubled to 19.1% between 2010 and 2021.

Image: Unsplash Show your support for our mission by joining our community of experts, Cube Club and Cube Event. Join the community that includes Andy Jassy, ​​CEO of Amazon Web Services and, Michael Dell, Founder and CEO of Dell Technologies, Pat Gelsinger, CEO of Intel, and many more luminaries and experts.