Microsoft Phi-3 Mini announced
Microsoft introduces Phi-3 Mini, its smallest AI model to date, which can compete with models such as GPT-3.5 despite its small size, making it ideal for companies with smaller data sets and limited budgets.
Tools
I’ll keep you up-to-date with the latest tools and updates. There’s a lot to cover …
VideoGigaGAN improves video upscaling
VideoGigaGAN outperforms previous methods of video upscaling, creating videos with a high level of detail and consistency. The approach is based on the GigaGAN image upscaler and solves its video processing problems through special techniques that result in sharper and smoother videos. Source: Hacker News
“Expressive Avatars” are more lifelike than ever
London-based Synthesia introduces “Expressive Avatars,” a new generation of AI avatars that adapt their facial expressions, gestures, and tone of voice to the context of the spoken content. This makes it possible to create more realistic and emotional AI videos for marketing, training or patient communication.
Microsoft’s VASA-1 generates video from a photo and audio
Microsoft’s VASA-1 can make human portraits sing and talk. It only needs a still image and an audio file with speech to generate moving lips, matching facial expressions and head movements. Microsoft emphasizes that this is a research demonstration only, with no plans to bring it to market.
AdaKWS claims better speech recognition than OpenAI’s Whisper
The new AI model AdaKWS from speech recognition specialist aiOla claims to be able to convert speech correctly into text, even if it is technical jargon. The model achieves an accuracy of 94.6% – better than OpenAI’s Whisper.
ChatGPT update brings “memory” feature and temporary chats
ChatGPT has received several updates. One of the most interesting is the “memory” feature, which allows ChatGPT to remember information that users communicate to it. For example, you can store details about yourself or your company that the chatbot can access when needed. Other new features include the ability to have temporary chats that are …
Meta’s impressive ChatGPT alternative Llama 3
Meta presents Llama 3, the latest generation of its speech models, which is freely available for download. The models are said to surpass the performance of many competitors and can even compete with some of the best proprietary models. Llama 3 is said to excel at multiple-choice questions, programming tasks, and mathematical problems. In addition …
Tested: Generative AI by iStock
There are many AI image generators. But they often have one problem: copyright. First, it is not always clear where the training material comes from. Second, it is not certain to what extent you can get into legal trouble with the generated images. Such image generators are often out of the question for companies and …
Google researches give AI “infinite” attention
With “Infini-attention”, Google researchers have developed a technology that allows language models to process texts of theoretically infinite length without requiring additional memory and computing power. Source: VentureBeat
Limitless helps with meeting notes
Limitless is a new AI tool that records, transcribes and summarizes meetings and provides real-time notes. The accompanying “Limitless Pendant” is a portable device that automatically records audio and sends it to the cloud. That sounds useful, but it also sounds like a huge privacy problem … Source: The Verge