Scientists from Stanford University in the US recently released their latest AI annual report, which runs an impressive 386 pages long. The report ranks the top AI research institutions based on the number of AI-relevant publications between 2010 to 2021, with nine of the top ten AI research institutions from China, followed by the US Massachusetts Institute of Technology in tenth place. However, according to the AI Index Report, universities are becoming less important in development. In 2022, only three of the 35 major models for machine learning were published by research institutions, with 32 models published by companies.
In 2022, global private investment in AI saw a decline for the first time since 2013, totalling $91.1 billion, a decrease of 26.7 percent compared to the previous year. Nevertheless, over the past decade, there has been a clear upward trend. According to the AI Index Report, politicians are increasingly discussing the topic, as indicated by an analysis of parliamentary debates in 81 countries. While the number of mentions of AI in global legislative processes increased by 6.5 times between 2016 and 2022, public opinion on AI varies widely. According to an IPSOS survey, 78 percent of respondents in China agreed that AI products and services have more advantages than disadvantages, with approval ratings the lowest in the US at only 35 percent.
The abuse of AI is a growing concern, according to the AI Index Report. The non-profit AIAAIC database, which documents such cases, reportedly shows that the number of AI controversies has increased 26-fold since 2012. The impact of AI on the environment is also addressed, with the report citing research that shows the training run of the largest open AI language model, “Bloom”, causes 25 times as much carbon emissions as a single traveller on a plane trip from New York to San Francisco.
The rapid progress made by AI models in recent years has rendered previous benchmarks obsolete. Therefore, new, more comprehensive benchmarking suites such as Google’s “BIG-bench” and the “Holistic Evaluation of Language Models” (HELM) developed at Stanford are discussed in the report. This is the sixth time Stanford scientists have presented the AI Index Report, which they claim to be the world’s most credible and comprehensive source for data and insights into AI. The complete report and the data used are available for free download.
The report also highlights new AI technologies, such as AI search engines that scour the web and link sources for users, as well as the impending end of support for Windows 10, which raises concerns over e-waste. Additionally, the report touches on a few other topics, including the testing of mini PCs, explaining how to recover deleted files under Linux, and a reminder about the c’t “Hommingberger Gepardenforelle” campaign. This and more can be found in the latest edition of c’t.
