AI Weekly Rundown (May 13 to May 19)
OpenAI, Google, Meta, Tesla, Hugging Face, Amazon, and more making huge AI strides.
Hello, Engineering Leaders and AI Enthusiasts.
Another eventful week in the AI realm. Lots of big news from huge enterprises.
In today’s edition:
💬 ChatGPT rolls out 70+ plugins for users
🔍 InstructBLIP: Language model that handles both Vision and Language
🧠 Hugging Face introduced RWKV, combining the best of RNNs & transformers
🛍️ After Google, It’s Amazon Search to get an AI makeover
📚 Microsoft released Guidance language for controlling large language models
💊 Google Cloud launches AI tools for drug discovery and precision medicine
🤖 Humanoid bots make wave amid AI’s surge
🖥️ AI coding announced for Google Colab
🔌 Meta develops its own AI chip Meta Training and Inference Accelerator (MTIA)
Let’s go.
ChatGPT rolled out 70+ plugins for users
The latest release notes of ChatGPT highlight various updates and new features added to the model, such as the ability to turn off chat history, export data, and use AI plugins. In addition, there are updates on the model's performance, factuality, and mathematical capabilities.
ChatGPT Plus subscribers get early access to experimental features, such as the ability to choose between different versions of ChatGPT and new experimental models that can browse the internet and use third-party services.
InstructBLIP: Language model that handles both Vision and Language
InstructBLIP is a vision-language model that utilizes instruction tuning and instruction-aware visual feature extraction to improve performance on vision-language tasks.
It is based on pre-trained BLIP-2 models and achieves state-of-the-art performance on multiple benchmarks, including zero-shot evaluation. The model is designed to effectively integrate visual and linguistic information to solve complex language-domain tasks.
Hugging Face introduced RWKV, combining the best of RNNs & transformers
Hugging Face introduced a new architecture, RWKV, which takes the best from transformer-based models and RNNs (Recurrent Neural Networks) to provide faster inference and theoretical infinite context length.
It has been integrated into the Hugging Face transformers library and comes with the addition of the "chat" version of the model, called RWKV-raven. Moreover, the GPUs for training the RWKV models are donated by Stability AI.
After Google, It’s Amazon Search to get an AI makeover
From a recent job listing on Amazon, it looks like the company is working on giving Amazon Search a “once in a generation transformation” with AI-powered features.
Amazon Search may soon have an interactive conversational experience that will help users find answers to product questions, perform product comparisons, receive personalized product suggestions.
AI can detect signs of Alzheimer’s even before symptoms begin to show
Researchers at UT Southwestern Medical Center have found that AI-powered voice analysis can help diagnose Alzheimer’s and cognitive impairment in early stages. If confirmed by larger studies, these findings could primary care providers with an easy-to-perform screening tool for at-risk individuals.
The research used advanced machine learning and natural language processing (NLP) to identify even the subtlest changes in language and audio that individuals may not easily recognize.
Microsoft released Guidance language for controlling large language models
Microsoft has released a new guidance language for controlling large language models (LLMs) that allows developers to interleave generation, prompting, and logical control into a continuous flow, which can significantly improve performance and accuracy.
The tool features simple and intuitive syntax, rich output structure, support for role-based chat models, easy integration with HuggingFace models, and intelligent seed-based generation caching. It also offers playground-like streaming in Jupyter/VSCode notebooks and regex pattern guides to enforce formats.
Google Cloud launches AI tools for drug discovery and precision medicine
Google Cloud has launched two AI-powered tools to help biotech and pharmaceutical companies accelerate drug discovery and advance precision medicine. The Target and Lead Identification Suite aims to streamline the process of identifying a biological target and predicting protein structures, while the Multiomics Suite assists researchers in ingesting, storing, analyzing, and sharing large amounts of genomic data. Both tools aim to significantly reduce the time and cost associated with drug development.
Humanoid bots make wave amid AI’s surge
Sanctuary AI unveiled a new humanoid robot named "Phoenix," designed for general-purpose work. It’s a 5’7” and 155 lbs humanoid that can walk at 3mph. The robot has advanced sensors and control systems allow it to move and interact. Hours later, Tesla also rolled out images of their AI humanoids walking around and gathering information.
Now you can use AI without the internet with PrivateGPT
Private AI has introduced PrivateGPT, a product designed to help businesses utilize OpenAI's chatbot without risking customer or employee privacy. PrivateGPT is an AI-powered tool that redacts over 50 types of Personally Identifiable Information (PII) from user prompts prior to processing by ChatGPT, and then re-inserts the PII into the response for a secure and seamless user experience.
This allows companies to remain compliant with data protection regulations and prevents personal information from being exposed due to bugs or data breaches.
AI coding announced for Google Colab
Google Colab is set to introduce AI coding features, including code completions, natural language to code generation, and a code-assisting chatbot, powered by Codey, a family of code models built on PaLM 2. Codey has been fine-tuned on a large dataset of high-quality, permissively licensed code and customized specifically for Python and Colab use cases.
In addition to code generation, an integrated chatbot will be available to assist users with their questions directly in Colab.
OpenAI has released the ChatGPT app for iOS
It will sync your conversations, support voice input, and bring the latest model improvements to your fingertips. Also, ChatGPT Plus subscribers get exclusive access to GPT-4’s capabilities, early access to features, and faster response times, all on iOS.
They have started the rollout in the US and are expected to enter additional countries in the coming weeks. OpenAI said: “Android users, you're next!”
Meta develops its own AI chip Meta Training and Inference Accelerator (MTIA)
The custom chips will be optimized for AI workloads and allow for more efficient training and running of complex models. In addition, Meta is building a supercomputer with a capacity of over 100 petaflops, which will be used to train large-scale AI models for natural language processing and computer vision. The goal of these investments is to enable the development of more advanced products and services, such as virtual assistants and augmented reality applications.
That's all for now!
If you are new to ‘The AI Edge’ newsletter. Subscribe to receive the ‘Ultimate AI tools and ChatGPT Prompt guide’ specifically designed for Engineering Leaders and AI enthusiasts.
Thanks for reading, and see you Monday.