Sora in Hollywood and AI for Big Football

It’s high time to tell you what news from the world of artificial intelligence turned out to be  #InfocusAI in the second half of March. You will learn about OpenAI´s plans to promote the Sora neural network, another idea to improve computer vision, a football recommendation system from Google, and plans of technology companies to  loosen NVIDIA’s position on the market. And also, about which country can be considered a “talent factory” for AI research. 

Issue 38, March 7-29, 2024

OpenAI is going to promote its Sora video generation neural network among Hollywood filmmakers and creative studios for the production of various video content. According to Bloomberg, the company is already entering into negotiations with influential representatives of the film industry and mass media in Los Angeles for these purposes. For now, the developers do not open Sora to the public users. The access was granted only to a small group of professionals involved in creating video content. The first results of their work you can view on the OpenAI blog, if you haven’t yet. It seems that very soon there will be completely different professions in the film industry… 

Scientists from MIT, Microsoft, Google, and Adobe have figured out how to further improve image and video processing with computer vision algorithms. It is often the case that when extracting features (this is one of the main operations in data processing), spatial resolution is compromised. The FeatUp technology developed by scientists allows smoothing out this problem. It is based on the idea of using the consistency of multiple low-resolution views to supervise the construction of high-resolution signals. The authors were inspired by 3D reconstruction methods. To learn more, please, go to the MIT News website and this scientific article. 

Technology companies, including Qualcomm, Intel, and Google, have formed a coalition to compete with NVIDIA in the field of AI. They plan to loosen the position of the CUDA platform which is used by more than 4 million developers of AI and other applications around the world and which allegedly keeps them “tied” to NVIDIA chips. As an alternative, the companies intend to create independent open-source tools which power various AI chips. To learn more, please read this exclusive article from Reuters.

A MacroPolo study published recently showed that most leading scientists in the field of artificial intelligence are concentrated in the United States (42%), which is obvious, followed by China (28%). However, approximately 47% of influential AI researchers were granted undergraduate degrees in China, and highly qualified scientists who started their professional training in the United States were more than half as many, i.e., 18%. The study reports that the shares of the Chinese and US top scientists in the US institutions are approximately the same, i.e., 38% and 37%, respectively. Another interesting observation is that the number of AI researchers who stay in their country after graduation is increasing. For more figures, click here. 

This time we will finish the digest with sports news. Last week, the Google DeepMind team shared their new development, TacticAI. It is a recommendation AI system based on a combination of predicative and generative models for football team coaches. It is able to provide them with valuable data about the game, especially in terms of corner kicks, and help in choosing behaviour tactics for each player. The solution was under development for several years together with the Liverpool football club which tested it afterwards. It is reported that Merseyside appraisers preferred TacticAI´s tactical recommendations in 90% cases.