The field of AI is constantly evolving, with new innovations and trends emerging all the time. Despite this fast-paced environment, Stanford University's recent 386-page report on AI provides some valuable insights into this complex domain. The report was compiled by experts from academia and private industry, with involvement from the AI Index at the Institute for Human-Centered Artificial Intelligence.
This year's report delves into topics such as foundation models, including their geopolitics and training costs, the environmental impact of AI systems, K-12 AI education, and public opinion trends in AI. It also examines policy in over 100 new countries.
One trend highlighted in the report is the shift from academia-led to industry-led development of AI over the past decade. However, the report also notes that while the energy footprint of AI training and use is becoming significant, its delivery is still not meeting expectations.
Another interesting finding is that the number of "AI incidents and controversies" has increased by a factor of 26 since 2012. Despite this, the rate of AI-related job postings and investment growth has not been as high as anticipated, with a temporary stall in recent years.
The report also highlights differences in public opinion on AI. While over 70% of Chinese, Saudi, and Indian respondents felt AI had more benefits than limitations, only 35% of Americans shared this view.
One key concern in the field of AI is technical ethics, particularly around issues of bias and toxicity. The report notes that while it is difficult to reduce these factors to metrics, unfiltered models are more prone to problematic outcomes. Tuning instructions and passing output through a second mediator model can help mitigate these issues, but there is no simple solution.
Despite AI's potential for fact-checking, the report suggests that AI is still not effective at evaluating factuality. However, there is a growing interest in ethical considerations related to AI, with increasing attention given to issues of fairness, privacy, and interpretability.
The Stanford report highlights the need for ongoing attention to ethical considerations and the importance of continued investment in AI development.