The latest update, spanning from June 12th, 2023, highlights the recent advancements and announcements in the domains of data science, machine learning, and artificial intelligence. This week's spotlight is on notable entities such as Google, OpenAI, UC Berkeley, and AWS.
Generative AI Support on Vertex AI Is Now Generally Available
Google has introduced the availability of Generative AI support on Vertex AI, enabling customers to utilize the latest platform features in creating and operating personalized generative AI applications. This update empowers developers to utilize various tools and resources, including the PaLM 2 powered text model, the Embeddings API for text, and foundational models in Model Garden. Additionally, the Generative AI Studio offers user-friendly tools for fine-tuning and deploying models. With the backing of enterprise-level data governance, security, and safety measures, Vertex AI simplifies the process for customers to access foundational models, customize them with their own data, and swiftly develop generative AI applications.
DeepMind Introduces AlphaDev
Google Deepmind has unveiled AlphaDev, an artificial intelligence system that leverages reinforcement learning to uncover improved computer science algorithms, surpassing the ones developed by scientists and engineers over many years. AlphaDev has discovered a more efficient sorting algorithm, which is used for organizing data. These algorithms play a fundamental role in various aspects of our daily lives, from ranking online search results and social media posts to data processing on computers and smartphones. The utilization of AI to generate superior algorithms is poised to revolutionize computer programming and have a significant impact on all facets of our ever-growing digital society.
OpenAI Announces Function Calling
OpenAI has introduced updates to the API, including a capability called function calling, which allows developers to describe functions to GPT-4 and GPT-3.5 and have the models create code to execute those functions. Function calling facilitates the development of chatbots capable of leveraging external tools, transforming natural language into database queries, and extracting structured data from text. These models have undergone fine-tuning to not only identify instances where a function should be invoked but also provide JSON responses that align with the function signature.
UC Berkeley Develops Gorilla, an LLM Connected with APIs
Researchers from the UC Berkeley released Gorilla, a finely-tuned model based on LLaMA, that surpasses GPT-4 in terms of performance when generating API calls. Its integration with a document retriever allows Gorilla to effectively adapt to changes in documents during testing, enabling flexible API updates and version changes. Moreover, Gorilla significantly addresses the issue of hallucination often encountered when directly prompting LLMs. To assess the model's capabilities, APIBench, a comprehensive dataset comprising HuggingFace, TorchHub, and TensorHub APIs, has been introduced. The successful combination of the retrieval system with Gorilla showcases the potential for LLMs to utilize tools more accurately, stay updated with frequently changing documentation, and enhance the reliability and practicality of their outputs.
AWS Introduces Falcon, a Foundational LLM Built and Trained with Sagemaker
AWS and the UAE Technology Innovation Institute have launched Falcon LLM, a foundational large language model with 40 billion parameters. Falcon matches the performance of other high-performing LLMs, and is the top-ranked open-source model in the public Hugging Face Open LLM leaderboard. It’s available as open-source in two different sizes, Falcon-40B and Falcon-7B, and was built from scratch using data preprocessing and model training jobs built on Amazon SageMaker. Open-sourcing Falcon 40B enables users to construct and customize AI tools that cater to unique users' needs, facilitating seamless integration and ensuring the long-term preservation of data assets. The model weights are available to download, inspect and deploy anywhere.
