AI Insight
The AI stories you don't want to miss this week!
Howdy pardner. Jeff Jarvis and I wrapped another episode of the AI Inside podcast a few hours ago and we discussed a bunch of really interesting news and developments in the world of AI. Check it out!
Google Developing AI Agent 'Jarvis'
Google is reportedly working on a powerful AI agent called Jarvis, which will control Chrome browsers by interpreting frequent screenshots and carrying out tasks like clicking buttons or typing into text fields. This could make everyday tasks like research, shopping, and booking flights easier if automated correctly. Jarvis will be powered by Google's next-generation Gemini model, and we might see a preview as soon as December 6, when Gemini 2 is expected to launch. The full release of Jarvis is slated for 2025, but it's still unclear if this will be a premium feature for Gemini Advanced subscribers or available to everyone. It’s easy to speculate that Jarvis could someday evolve to control all of ChromeOS.
This news is reminiscent of last week’s story about Anthropic unveiling its own AI agent, Claude with Computer Use, which takes things further by controlling the entire computer, not just the browser.
Google’s Quarterly Results: AI-Generated Code
Google had a strong quarter, with Google Services revenue up 13% year-over-year and Google Cloud (which includes its AI infrastructure) up 35%. During the earnings call, Sundar Pichai dropped an interesting tidbit:
"More than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers."
This shows just how deeply Google is integrating AI into its own operations.
But it raises some questions: Is this AI-generated code writing entire chunks of new functionality? Or is it just adding comments or small improvements? Could Pichai be overstating things on an earnings call? No matter how you slice it, a quarter of all code is A LOT OF CODE at a company like Google. Talk about eating your own dog food.
GitHub Expands Copilot with Multi-Model Support
GitHub is expanding its popular Copilot tool to support multiple models, not just OpenAI’s. Soon, developers will be able to chat with models from competitors like Google and Anthropic in Copilot Assistant, though OpenAI will remain the default option. This shows Microsoft isn't putting all its eggs in the OpenAI basket.
GitHub also introduced Spark, a new tool that allows users to build micro-web apps using natural language. It’s designed for both non-coders looking to create simple apps and experienced developers who want to tweak the code. With more than one million paid subscribers and a $2 billion revenue run rate, GitHub Copilot is proving that AI-powered coding tools are big business.
Meta Building Its Own AI Search Engine
Meta is reportedly developing its own AI-powered search engine in an effort to reduce reliance on Google and Bing for real-time information in Meta AI. This project has been in the works since early 2024, and Meta AI now boasts 185 million active weekly users across its apps.
In addition, Meta recently announced a partnership with Reuters to provide current events and news answers—its first AI news deal. Ben Thompson from Stratechery believes Meta is well-positioned in the generative AI race, with short-term gains likely coming from AI-powered advertising and longer-term opportunities in content integration and XR projects like Project Orion.
Meta’s NotebookLlama: A New Take on NotebookLM
Meta has launched NotebookLlama, their own version of Google's NotebookLM. The tool is free and open-source but currently only accepts PDF inputs. It can turn those PDFs into podcast episodes with two voices discussing the topic in a conversational way.
While still rough around the edges, future updates are expected to add features like web links, audio files, and YouTube integration. Stay tuned for more developments!
Apple Intelligence Rolls Out (Softly)
Apple's much-anticipated Apple Intelligence has started rolling out to iPhones and Macs with M1/A17 Pro chips—but not everyone gets access right away. For now, users can enjoy writing tools, an improved Siri, and some photo-cleaning features as part of iOS 18.1 and macOS Sequoia 15.1.
More advanced features like ChatGPT integration, Image Playground, Genmoji, and Visual Search are coming in December. However, Apple won't release Apple Intelligence in the EU until April 2025. While this feels more like a soft launch than a big splashy debut, it’s clear that Apple is taking its time perfecting these tools.
Humane Shifts Strategy After AI Pin Flop
Humane—the company behind the much-maligned AI Pin—is shifting away from hardware after its disastrous launch earlier this year. The company is now focusing on licensing its CosmOS operating system for use in other devices like car dashboards. CosmOS is designed for agent-driven interactions rather than traditional app-based systems.
Humane founders Bethany Bongiorno and Imran Chaudhri say this was always part of their long-term plan—the AI Pin was meant to help them understand what’s needed for multimodal contextual computing. In an interesting move, they’ve also slashed the price of the AI Pin from $699 to $499 after reports surfaced earlier this year of potential buyout talks with HP.
Universal Music Group Partners with Klay Vision for Ethical AI Music
The music industry has been fighting against generative AI models like Suno and Udio while also trying to find ways to capitalize on this new technology without infringing on intellectual property rights. Universal Music Group (UMG) has now partnered with Klay Vision—an "ethical AI music company"—to create a foundational model for AI-generated music that works collaboratively with artists.
This partnership marks UMG's attempt to stay ahead of the curve while navigating ongoing litigation with Anthropic AI over copyright issues related to generative music models like Suno and Udio. Klay Vision's new Music Model (KlayMM) could set new standards for ethical collaboration between tech companies and musicians.
Here are a few other interesting stories
OpenAI CFO: Paid ChatGPT Subscriptions Make Up Most of Our Business - CNET
Hospitals use a transcription tool powered by an error-prone OpenAI model - The Verge
LeCun blasts Musk as the biggest threat to democracy today - Futurism
Linus Torvalds slams AI as ‘90% marketing and 10% reality’ - Tech Radar
Stability AI Is No Longer Controlled by Former CEO Emad Mostaque - Bloomberg
Arcade, a new AI product creation platform, designed this necklace - TechCrunch