Meitu released 7 AIGC products
At the Meitu Image Festival held today, Meitu (01357. HK) released 7 AIGC tools, including the AI vision model "MiracleVision", the digital human product "DreamAvatar", and the AI-inspired design tool "WHEE". According to the observation of the "Science and Technology Innovation Board Daily", these 7 products are mainly biased towards the B-end market, and will be launched in July after another.
According to the market, Meitu (01357. HK) rose late in the day. At press time, the company's shares were up nearly 27%.
LG releases Captioning AI, the first generative AI commercialization service
At CVPR 2023, the world's largest computer vision conference held yesterday in Vancouver, Canada, the LG Artificial Intelligence Research Institute, a subsidiary of South Korean technology giant LG, unveiled the company's first generative AI commercialization service Captioning AI, providing a more efficient solution for customers who need to manage a large number of pictures.
LG introduced that Captioning AI is based on zero-shot learning technology, which allows the AI to use its previously accumulated experience and knowledge to interpret the object or scene it sees for the first time, just like a human. On average, the service generates 5 complete descriptive sentences and 10 keywords for an image in 10 seconds. If the number of images is expanded to 10,000, these will be completed in two days, allowing a customized image search and management system to be established in a short period of time.
LG said that Captioning AI is a deep collaboration between LG Artificial Intelligence Research and Shutterstock, one of the world's largest image libraries, which has a wealth of expertise in image labeling. In addition, AI ethical validations such as filtering and filtering of learning data were carried out to ensure copyright transparency.
Sezal Amin, Chief Technology Officer at Shutterstock, said: "We are currently further developing this service through an Early Access program for 10 customers worldwide. This service is intended to help people focus on more important content and creative tasks rather than repetitive mechanical tasks. ”
Global demand for AI talent is surging, with salaries approaching $400,000 a year in some roles
According to USA Today, in the face of a huge talent gap, more and more American technology companies are trying to compete for limited talent with generous salaries and benefits, so as to take the lead in the competition on the AI track, and the annual salary of some positions has even approached 400,000 US dollars (currently about 2.852 million yuan).
OpenAI, the parent company of ChatGPT, is looking for experienced AI trainers with annual salaries ranging from $200,000 to $370,000 (currently about 1.426-2.638 million yuan), and even generous benefits such as equity dividends, 29 weeks of paid parental leave, family insurance and unlimited time off. AI security and research firm Anthropic is also looking for the right people to "discover, test, and document AI best practices," create tutorials and work directly with customers. The expected salary range for this position is $280,000 to $375,000 (currently about 1.996-2.674 million yuan), which also includes equity dividends.
In addition, Adobe is looking for a senior creative "evangelist" who specializes in AI-related work, the core of which is to guide and support the use and understanding of AI within the company, hence the name "evangelist". The salary for this job varies depending on the location, ranging from $101,000 to $202,000 per year (currently about 72-1.44 million yuan).
Princeton University launches Infinigen AI model
Alexander Raistrick, a researcher at Princeton University in the United States, has released an open-source AI model called "Infinigen", which can generate realistic 3D natural world scenes.
It is reported that the Infinigen model is built on the open-source modeling software Blender, which can generate a series of "plants, animals" scenes in nature through random mathematical rules, and can customize the terrain by entering prompt words, and add natural phenomena such as "clouds, rain, snow, thunderstorms, and sky fire". But the current efficiency of the Infinigen model leaves much to be desired, and researchers have tested that it takes 3.5 hours to generate a 1080P image (a double image of two scenes).
Meta lost a third of researchers who published AI research last year
Analytical data shows that Meta lost at least a third of researchers who published AI research last year, and many expressed burnout or lack of confidence in Meta, believing that Meta could not keep up with competitors. A Meta spokesperson said the company has been continuing to recruit and bring in new AI talent.
For more metaverse information, please pay attention to the public account "Metaverse Chronicles"
Reply to "Reports" to get the latest and most authoritative industry reports