1. Meta released I-JEPA
I-JEPA is a self-guided computer vision technology that gains an understanding of its surroundings by forecasting them. This AI mechanism emulates the learning and reasoning processes found in animals and humans.
Source: Facebook Blog
2. DragGan released
DragGAN is an open-source AI image editing tool that allows users to easily adjust and deform images by dragging points on the image, with applications ranging from adjusting the layout, poses, shapes, and even expressions in photos, and the team plans to release a version that works with 3D models.
Source: GitHub
3. Open Orca LLM
Orca is a new 13B open-source language learning model from Microsoft Research, built on llama13b, that surpasses ChatGPT on the BBH benchmark, matches text-davinci-003 on the AGIEval reasoning benchmark, and has been trained on large-scale, diverse, and complex datasets.
Source: arxiv Cornell University
4. Apple utilizes AI
Apple has announced a modification to the iPhone's autocorrect feature to stop it from erroneously converting a common expletive to "ducking," while continuing to learn other text inputs from users.
Source Reuters
5. OpenAI Updates
New ChatGPt 3.5 16K model. Native JSON responses in GPT-4
6. Adobe Firefly for video announced
Adobe Firefly, a generative AI suite, will be expanded to include video, audio, animation, and motion graphics design, offering enhanced capabilities such as color alteration, royalty-free sound generation, customized text effects, and automated storyboarding, with the aim of accelerating creative processes and cutting post-production time
Source: Adobe Blog
7. Youtube is exmperimenting with Auto Dub
YouTube has developed a new tool that can automatically dub videos into different languages, helping to make content more accessible to international audiences. YouTube utilizes the Aloud dubbing tool. Currently, this new functionality is only available to the most active (larger) YouTubers.
Source: Aloud
8. Microsoft launched Azure OpenAI Service
Azure OpenAI Service enables the development of mission-critical solutions for analyzing images, understanding speech, and making data-driven predictions
Source: Microsoft Azure
9. Meta showed Voicebox
Facebook's Meta AI has developed Voicebox, a pioneering generative AI model for speech that can adapt to tasks it wasn't specifically trained for, producing high-quality audio clips across six languages and various styles, capable of noise removal, content editing, style conversion, and diverse sample generation
Source: Meta Blog
10. Stability AI launches SDXL 0.9
Stability AI has launched SDXL 0.9, a significant advancement in their Stable Diffusion text-to-image suite of models, which produces enhanced image and composition detail, and enables a range of creative applications for generative AI imagery in films, television, music, instructional videos, design, and industrial use, while being operable on a modern consumer GPU
*Source: Stability AI Blog