Google presents SoundStorm – a new model for efficient audio generation. It can generate highly realistic dialogues via transcript annotations and short voice prompts. See demo in GitHub Link – It’s impressive.
Microsoft releases a new language for controlling large language models: ‘Guidance’. Guidance enables you to control modern language models more effectively and efficiently than traditional prompting or chaining .
Zapier launched two new AI beta features for their no-code automation platform:
Stability AI released StableStudio – the open-source variant of DreamStudio, their text-to-image app, with plans to create bounties for new features.
Project Ring: is a generative AI-based wearable device that has a camera and microphone. It can chat with you about what it sees and is powered by OpenAI Whisper (voice-to-text), Replicate (image-to-text), ChatGPT (text-to-text), and ElevenLabs (text-to-voice). The entire code – Raspberry Pi Python script, cloud application, HTML webpage, and Android app – was written by GPT-4! 1
- Meta shares plans for their next generation of AI infrastructure: a custom silicon chip (MTIA – Meta Training and Inference Accelerator) for running AI models , a new AI-optimized data center design and the second phase of their 16,000 GPU supercomputer for AI research .
Apple Point and Speak is a feature in the iPhone’s built-in Magnifier app that enables users who rely on the app to navigate the world around them to use “point and speak” by pointing their finger at something in front of the camera, and the app will read it out loud1. This feature is available in iOS 171.
Cloudflare introduced Constellation: a new feature to run fast, low-latency inference tasks using pre-trained machine learning models natively with Cloudflare Workers scripts.
Glide, the no-code tool for building custom apps, now includes integration with OpenAI in Glide apps.
Google’s Colab will soon have AI coding features like code completions, natural language to code generation and a code-assisting chatbot. Colab will use Codey, a family of code models built on PaLM 2, which was announced at I/O last week.
- Hippocratic AI has built a safety-focused large language model for healthcare to assist with tasks such as explaining billing, providing dietary and medication advice, answering surgery-related queries, patient onboarding etc. .
OpenAI is rolling out ChatGPT plugins and web browsing feature to all paid users. Check and enable via ‘Beta features’ in the ChatGPT ‘Settings’. Available if you have a paid $20/ month version