AI is implemented almost everywhere and in everything, but global companies, not tall as this girl, but quite high, past social networks and online shopping where the recommendation system actively takes part as a main favorite, are always interested in additional areas to bring innovation to the whole world, not one segment of it.
Facebook AI recently introduced TextStyleBrush, a system that analyzes a banner ad, newspaper or any other medium of text information in real time, allowing it to replace what is written on the medium with text that the user creates.
The model, which uses the underlying StyleGAN2 framework, handles both handwriting input and typography and is able to analyze various subtleties of style and account for transformations and deformations such as rotations and twists. The authors note, however, that improvements still need to be made, particularly in the realism of replacements on metal supports and in the handling of reflections.
A little while ago, Facebook also launched GrokNet, a unified computer vision model with which they intend to create the world’s largest social media shopping platform. The model is currently running on Facebook Marketplace. The company soon plans to expand GrokNet to new apps on Facebook and Instagram.
GrokNet detects the products in the picture and predicts their categories. Unlike previous models, Facebook’s product recommendation system is a universal model that scales to billions of photos vertically, including fashion, auto and home decor.
If we talk about computer vision, it is impossible not to mention Google, which are big fans in the practice of this branch, especially in the practice for various areas such as in this case, medicine.
The University of Waterloo has been noted as using artificial intelligence for its researches, in particular now – the detection of skin cancer. It was noted by Google, which subsequently presented their solution to detect skin diseases by photos in real time.
The user takes three pictures of an area of skin, hair or nails that they think has a dermatological problem and answers a few questions about their skin type and the problem itself (other symptoms, pain and/or how long etc.).
And for the development of AI itself, in particularly the participation of ordinary users who are not programming gurus, Apple is creating a non-code AI platform.
The platform allows machine learning researchers and non-technical geospatial specialists to experiment with domain-specific signals and datasets to solve different problems. It adapts complex spatiotemporal datasets to standard deep learning models, in this case convolutional neural networks (CNNs), and formulates disparate problems in a standard way, such as semantic segmentation.
Perhaps to maintain this platform, gurus like people who have devoted years to learning certain languages will act merely as mentors, because IBM has announced Project CodeNet, a large dataset that aims to help teach AI how to understand and write code by itself.
CodeNet features 500 million lines of code, 14 million examples, and spans 55 programming languages including Python, C++, Java, Go, COBOL, Pascal, and more. Projects such as OpenAI’s GPT-3 are showing how AIs are becoming quite adept at penning the languages of us humans, but writing their own native code could be innovative for finding another different methods to solve the problems with code.