Anthropic, Google, Microsoft, and OpenAI are forming the Frontier Model Forum, an industry-led body focused on the safe and responsible development of powerful AI models. The forum will start with a research focus and work to develop standardized best practices for evaluating frontier model risks and to issue a public library of solutions. The companies are among seven that recently committed to minimizing AI risks and sharing safety best practices with the White House. The Forum is open to companies developing powerful models that exceed the capabilities currently in the most advanced models.
New research in edge AI has made it possible for large language models to run on devices such as iPhones, Androids, PCs, and browsers. This could widen the use of generative AI and reduce the need for renting servers from cloud providers like Azure, AWS, and Google.
Amazon plans to use generative AI to revitalize its Alexa voice assistant and give it a more conversational nature. The company sees Alexa’s presence in the home and its existing ties to other devices as a key advantage over rivals. Amazon’s goal is to create an assistant that can answer questions and take action, but it must ensure that generative AI systems do not compromise security and privacy. Amazon is expected to announce new devices at an event in September, with more complete integrations to follow next year.
Due to low accuracy rates, OpenAI has shut down a tool designed to differentiate between human and AI-generated writing. The company plans to develop new mechanisms to identify AI-generated audio and visual content. Misinformation via AI-generated text is a growing concern, and governments have yet to figure out how to regulate it.
Bing’s AI chatbot is now available on Google Chrome and Safari, with some limitations, such as a 2,000-word prompt limit and conversation reset after five turns. Microsoft has also introduced dark mode for Bing Chat, accessed through the hamburger menu. Previously, the chatbot was only available on Edge, making it inconvenient for users on other browsers.
Microsoft’s AI-powered coding companion, GitHub Copilot for Business, has 27,000 customers using the service, including Airbnb, Dell, and Scandinavian Airlines. Microsoft CEO Satya Nadella touted the value of adopting generative AI solutions like GitHub Copilot, which infuses AI across Microsoft’s product inventory. However, the impact of AI on Microsoft’s earnings this quarter was negligible.
CoPilot for Business’s adoption numbers are solid – it’s not been widely available, and while revenue numbers aren’t a big deal yet, this is encouraging news. Plus, the ability to use LLMs on edge devices will push the relevance further out.
No one should be surprised that Amazon is looking to leverage its Alexa capabilities with AI, nor continued pushes to make the products more available by Microsoft.
There are complicated issues here – as OpenAI quickly learned, a tool alone won’t solve the coming changes due to AI. This is good news for service organizations, continuing to highlight the relevance of the business questions to be answered about its use. Good news. The players involved try to get ahead of regulation with their standards. That alone won’t do it… yet encouraging all the same.