Article originally posted on InfoQ. Visit InfoQ

This week’s roundup for April 3rd, 2023 includes the most recent news and information from the fields of data science, machine learning, and artificial intelligence. This week, OpenAI, Microsoft, Meta, and Bloomberg are just a few of the leading competitors in the market whose new breakthroughs and innovations are the focus of our attention.

Here are the most recent developments in natural language processing (NLP) models, new developer tools for AI, ethical AI practices, and other subjects.

OpenAI

OpenAI recently announced GPT-4, the next generation of their GPT family of large language models (LLM). GPT-4 can accept both text and image inputs and outperforms state-of-the-art systems on several natural language processing (NLP) benchmarks. The model also scored in the 90th percentile on a simulated bar exam.

Also, OpenAI announced their approach to AI safety, by improving AI systems’ ability to learn from human feedback and to assist humans at evaluating AI. The goal is to build a sufficiently aligned AI system that can help solve all other alignment problems.

Microsoft

Microsoft has open sourced Semantic Kernel (SK), a lightweight SDK enabling the integration of large language models (LLMs) with conventional programs which can leverage prompt templating, vectorized memory, intelligent planning, and other capabilities.

Also, Microsoft Researchers Introduce TaskMatrix.AI. A New AI Ecosystem that Connects Foundation Models with Millions of APIs for Task Completion. The concept involves integrating foundation models with millions of existing models and system APIs, resulting in a super-AI that can perform various digital and physical tasks. While AI models and systems are currently designed to address specific domains effectively, the diversity in their implementations and working mechanisms can make it challenging for foundation models to access them. This new ecosystem aims to overcome these obstacles by providing a unified framework for connecting these AI models and systems.

Meta

Meta just released their Segment Anything Model. This new model allows AI to extract objects within images or videos with a single click. It utilizes advanced computer vision technology, which enables computers to analyze and understand visual information- such as images or videos, similar to how humans perceive and interpret what they see.

Bloomberg

Researchers from Bloomberg and John Hopkins University train BloombergGPT, a language model with 50 billion parameters that serve a variety of financial sector operations. They adopt a hybrid approach rather than creating a tiny or general-purpose LLM solely based on domain-specific data. Standard LLM standards, open financial benchmarks, and proprietary benchmarks to Bloomberg are used to evaluate the model and ensure it functions as anticipated.