Earlier in March, Stanford’s Institute for Human-Centered Artificial Intelligence (HAI) release the 2021 AI Index report . The importance of ethical questions in relation to this technology is reflected in a dedicated chapter on AI ethics and a second one on AI Policy. HPC is only briefly mentioned and mostly indirectly. The report highlights, for example, a potential “compute divide” which gives an AI research advantage to large established academic as well as industry organizations because of easier access to powerful compute hardware such as HPC systems. Of the reports list of “Top 9 Takeaways” at least five have implications when viewed through an ethical lens (in the order of the report):
- “AI has a diversity challenge”: This is especially pronounced when considering PhD graduates in the US of which only around 3% each identify as African American or Hispanic, while 45% were white. Similarly, when considering gender only around 20% of PhDs are female.
- “China overtakes the US in AI journal citations”: While the number of publications and citations from the US still is higher when considering conferences and journals, it is notable that
- “Surveillance technologies are fast, cheap, and increasingly ubiquitous”: Capabilities across a wide spectrum of applications enabling large-scale surveillance matured considerable in 2020.
- “AI ethics lacks benchmarks and consensus”: Even though the discussion on AI ethics enjoys increasing attention, agreed on benchmarks are missing. In addition, researchers and civil society view AI ethics as more important than industrial organizations do.
- “AI has gained the attention of the U.S. Congress”: Legislators are considering AI with mentions tripled in the 116th Congress in comparison to the 115th.