[AI One-Bite News] AI Finds the Product in Your Mind

Google Unveils Shopping AI Features
Try Dressing Avatars in Clothes

Editor's NoteSince the emergence of ChatGPT, the world has been buzzing with artificial intelligence (AI). Technology is changing rapidly, and news is pouring in every day. Even industry experts who are at the top of their game say it's hard to keep up. It's overwhelming to consume it all, but you can't afford to miss out either. We bring you the key points from the weekly AI issues, carefully selected and easy to digest.

"Recommend colorful and metallic padded jackets"

AI Shopping Search Feature Introduced by Google
Photo by Google

Google has introduced a shopping feature powered by AI. It shows customized products based on the buyer's preferences or the characteristics of the item they want to purchase. For example, if you type "a gift for a husband who loves cooking," it recommends a list of suitable products. You can also search by product design, color, or texture. For instance, you can ask to find "a padded jacket that is colorful and shiny like metal." If the recommended products are not quite what you want, you can get help from AI. The AI generates product images based on your request, and by clicking on them, you can see real products similar to the generated images. If you find the product you want but are unsure if it will suit you, you can even try it on an avatar. By selecting an avatar similar in age, gender, and ethnicity, you can see how the chosen clothes look on it.

AI Weathercaster More Accurate Than the Meteorological Agency

Google DeepMind has developed a weather forecasting model using machine learning.
[Image source=DeepMind]

A weather AI that is faster and more accurate than the Korea Meteorological Administration's forecasting model has been developed. The star is GraphCast, created by DeepMind, an AI development company under Alphabet, Google's parent company. According to the international academic journal Science, this weather AI can predict the weather worldwide for 10 days in just one minute. It is at least 1,000 to 10,000 times faster than existing models. Its prediction accuracy is also higher. In a competition against the European Centre for Medium-Range Weather Forecasts' model 12,000 times, it achieved a 99% win rate. GraphCast was trained by learning about 40 years of weather data from 1979 to 2017 to uncover weather variability.

Choose John Legend's Voice to Create Pop Songs

We have entered an era where AI can instantly create songs and shoot music videos. Google YouTube introduced "Dream Tracks," a generative AI that creates songs. Nine famous artists, including John Legend and Charlie XCX, participated in this project. You can select the voice of any of the nine artists and order the music you want, and it will produce a soundtrack up to 30 seconds long. You can hum a melody, and it will add guitar accompaniment or even transform pop music into reggae. Meanwhile, Meta unveiled "Emu Video," which converts text into video. When you input text, it draws images and animates them according to instructions. You can also request edits via text, and the AI will apply the changes.

MS Unveils Its First Self-Developed AI Chip

[Image source=Yonhap News]

Microsoft (MS) has unveiled its first self-developed AI chips. They introduced the "Maia 100," a chip specialized for AI model training and inference, and the "Cobalt 100," a graphics processing unit (GPU). To develop these chips, MS collaborated with OpenAI, the creator of ChatGPT. They tested and refined the chips using OpenAI's models. MS plans to integrate these AI chips into its own cloud service, Azure. With this, MS now covers everything from AI model-based software (SW) to devices, the cloud that runs them, and the chips themselves.