Artificial Intelligence Index Report 2024 (AI column)
I read this one recently. Stanford University for Human-Centered AIwhich in the form of pdf takes a solid 500+ pages:) The report is quite significant in the industry and covers a large number of topics. (broken into 9 RnD, tech performance, responsible AI, economy, science & medicine, education, policy & governance, diversity, public opinion). Each chapter contains a large number of references to research, various visualizations and non-trivial conclusions that result in the next top-level findings.
1. AI beats humans on some tasks, but not on all Iron friends already outperform people in a number of tasks, not yet in all, but not yet in the evening. 2. Industry continues to dominate frontier AI research Creating foundational models is expensive, so large companies can afford it, but academic researchers lack budgets to compete. Therefore, many breakthrough results are born in corporations. 3. Frontier models get way more expensive It is interesting to observe the exponential growth in the cost of models, which gives a stable increase in results. 4. The United States leads China, the EU, and the U.K. as the leading source of top AI models USA is ahead of China in AI (If you look at other news about the ban on exports of technology to China, it seems that the USA wants to continue to be the leader.) 5. Robust and standardized evaluations for LLM responsibility are seriously lacking Evaluation of models is quite complicated. (In the second part about tech performance, the authors talk about different variants of benchmarks) 6. Generative AI investment skyrockets It can be seen that the topic of investing in AI is hot and all start-ups are now trying to add stories about artificial intelligence to their description in order to increase the value of business among investors:) 7. The data is in: AI makes workers more productive and leads to higher quality work There are studies that show increased productivity and quality of work using artificial intelligence 8. Scientific progress accelerates even further, thanks to AI Artificial intelligence is now helping the natural move science forward. It’s interesting to see what comes next year’s report – in fact, it’s a positive feedback loop, and if AI can accelerate scientific research, it’s better to start with AI research:) 9. The number of AI regulations in the United States sharply increases As usual, with great opportunities comes the great responsibility that regulators impose. Ideally, this fixes the boundaries and rules of the game, but at the same time slows down the exploration of the new. But AI is now too big an elephant for regulators to care about:) 10. People across the globe are more cognizant of AI’s potential impact—and more nervous, as it feels like we’re living in an era of change and a technological singularity is just around the corner. At the same time, it is not clear when and whose work will be given to our iron friends or how losing our jobs will affect our lives. True, while AI is more likely to go in the direction of assistants of people, but with such a speed of development of models and obtaining data for training, the assistant will soon be able to surpass teachers and teachers.
In general, I recommend the report to everyone interested in modern technological trends.
#AI #ML #Software #Engineering #DataScience