The Week in AI: Microsoft's New Copilot, Google's LLM Comparator, StarCoder2 Debut, More -- Pure AI

The Week in AI: Microsoft's New Copilot, Google's LLM Comparator, StarCoder2 Debut, More

By Pure AI Editors
03/04/2024

This edition of our weekly roundup of AI product and services news includes a preview of Microsoft Copilot for Finance; an early-stage GenAI music generation tool from Adobe; an update of its Commercial SaaS Agreement from Cohere; Tenstorrent's partnership with LSTC; and NVIDIA, Hugging Face, and ServiceNow's StarCoder2 the release.

Microsoft announced the public preview of Copilot for Finance, which the company described as "a new Copilot experience for Microsoft 365 that unlocks AI-assisted competencies for financial professionals, right from within productivity applications they use every day." Copilot for Finance connects to a company's financial systems, including Dynamics 365 and SAP, to provide role-specific workflow automation, guided actions, and recommendations in Microsoft Outlook, Excel, Teams, and other Microsoft 365 applications. It both suggests actions in the flow of work, and enables users to ask questions by typing a prompt in natural language. Sign up for the preview here.

Google researchers introduced a tool designed to provide side-by-side comparisons of LLM outputs. The LLM Comparator A team of researchers at Google Research has introduced the LLM Comparator tool, which facilitates the side-by-side comparison of LLM outputs, enabling an in-depth analysis of their performance. The LLM Comparator allows users to interactively explore the differences between model responses, clearly representing where and why one model may outperform another.

Adobe unveiled an early-stage GenAI music generation and editing tool called Project Music GenAI Control at last week's Hot Pod Summit in Brooklyn, NY. The experimental tool allows creators to generation music from text prompts, and then employ fine-grained control to edit the audio it generated. Using the tool, generative AI becomes your "co-creator," said Nicholas Bryan, Senior Research Scientist at Adobe Research and one of the creators of the technology, in a blog post. "It helps people craft music for their projects, whether they’re broadcasters, or podcasters, or anyone else who needs audio that’s just the right mood, tone, and length." The project is being developed by Adobe Labs.

Cohere announced that it is updating its Commercial SaaS Agreement with "Copyright Assurance" to provide full indemnification for any third-party claims that the outputs generated by its large language models infringe on a the party’s intellectual property rights. "This means we will safeguard our customers data and assume responsibility for any legal settlements or judgments that come from these claims," the company said in a statement. This expansion covers customers in compliance with Cohere’s guidelines and safeguards, including its Responsible Use Guidelines and Restrictions on Use, while using Cohere’s services, the company said.

NVIDIA, Hugging Face, and ServiceNow announced StarCoder2, a family of open-access large language models for code generation. Built by the BigCode community and managed by ServiceNow and Hugging Face, StarCoder2 was trained on 619 programming languages, and it can be further trained and embedded in enterprise applications to perform specialized tasks, such as application source code generation, workflow generation, and text summarization. Developers can use its capabilities to accelerate innovation and improve productivity. The BigCode project is an open-scientific collaboration focused on the responsible development of Large Language Models for Code (Code LLMs). StarCoder2 was introduced in a paper published on the Cornell University's arXiv lab site.

Japan’s "Leading-edge Semiconductor Technology Center," the LSTC, in conjunction with US-based hardware startup Tenstorrent and Tokyo-based Rapidus Corporation, announced a multi-tiered partnership last week to license and leverage Tenstorrent’s RISC-V technologies and chiplet IP for LTSC’s planned 2nm AI Accelerator for the edge. Headquartered in Toronto with offices in Silicon Valley, Tenstorrent develops AI processors for machine learning workloads. In addition to the IP licensing, Tenstorrent will also collaborate with LSTC to help co-design the chip.