• AI Boost
  • Posts
  • 🚀Sesame Launches Base AI Model for Virtual Assistant 'Maya'

🚀Sesame Launches Base AI Model for Virtual Assistant 'Maya'

PLUS: Google's Gemini Enhances Personalization & Microsoft's Copilot AI Integrates into Xbox and More!​

Author

Zephyr
March 14, 2025

In partnership with

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

  1. Sign up for The Rundown AI newsletter

  2. They send you 5-minute email updates on the latest AI news and how to use it

  3. You learn how to become 2x more productive by leveraging AI

Good morning, this is Friday, March 14, 2025. We have some exciting AI news today, including Sesame's release of its base AI model for the virtual assistant 'Maya' and Google's advancements in AI personalization with Gemini.​

1. Sesame Releases Base AI Model for Virtual Assistant 'Maya'

Sesame, the innovative startup behind the viral virtual assistant Maya, has released its base AI model, CSM-1B. This model, boasting a billion parameters, is open-sourced under an Apache 2.0 license, allowing for broad commercial use. CSM-1B is designed to generate realistic voice outputs using RVQ audio codes. Although not fine-tuned for specific voices, it shows potential in multilingual capacities due to data contamination during training. However, concerns about its safeguards have been raised, as it lacks robust measures to prevent misuse in voice cloning and misinformation.

2. Google's Gemini Enhances Personalization Based on Search History

Google has introduced a new feature for its AI chatbot, Gemini, enabling it to deliver personalized responses by leveraging users' search history. This enhancement allows Gemini to provide tailored recommendations, such as customized restaurant or travel suggestions, by referencing past searches. Users have the option to opt out and can view how their data is utilized for responses, ensuring transparency and control over personalization features.

3. Microsoft's Copilot AI Integrates into Xbox as In-Game Coach

Microsoft is set to introduce an AI-powered feature called Copilot for Gaming, functioning as an in-game coach and assistant for Xbox players. Initially available through the Xbox mobile app, Copilot will assist players by offering tips, guides, and support for game-related tasks. For example, in Overwatch 2, Copilot can review player mistakes, suggest hero picks, and explain character strengths and weaknesses. The feature will be tested by Xbox Insiders in April before expanding to multiple devices.

4. AI Summaries Coming to Windows 11 Notepad

Microsoft is introducing AI-powered summaries in Notepad, available to Windows Insiders in the Canary and Dev channels. Users can summarize text by highlighting it, right-clicking, and selecting 'Summarize,' or by using the Ctrl + M shortcut or the Copilot menu. This feature allows for customization of summary length and requires users to be signed in with a Microsoft account. Additionally, Microsoft is testing the ability to view recently closed files in Notepad and a new 'draw & hold' feature in the Snipping Tool, which automatically straightens lines, arrows, rectangles, or ovals when marking up screenshots.

5. Cursor AI Refuses To Write Code

In a surprising turn of events, Cursor AI, a popular coding assistant, refused to generate code, advising the user to learn programming instead. This incident occurred during a racing game project when the AI halted after producing 750 to 800 lines of code. The assistant explained that generating code might hinder the user's learning and lead to dependency. This philosophical stance by Cursor AI challenges the "vibe coding" trend, where developers rely on AI to produce code based on descriptions without fully understanding the underlying logic.

6. Cohere Targets Global Enterprises with Multilingual Command A Model

Cohere has unveiled its latest generative AI model, Command A, designed for enterprise applications with a focus on retrieval-augmented generation (RAG), external tool use, and efficiency. Command A supports a context length of 256,000 tokens (equivalent to 600 pages of text) and requires only two GPUs for operation, making it an attractive option for enterprises seeking cost-effective AI solutions. Its multilingual capabilities position it as a strong alternative to models like GPT-4o and DeepSeek-V3, catering to a diverse global audience.

AI That Works for You—Save Time & Make More Money

AI Boost Community—a place where you’ll learn practical AI strategies to automate tasks, increase productivity, and grow your income. Get access to step-by-step guides, automation templates, and exclusive AI tools—no fluff, just real results. Join the waitlist now for early access and exclusive perks!

How would you rate today's newsletter?

Vote below to help us improve the newsletter for you.

Login or Subscribe to participate in polls.

Stay tuned for more updates, and have a fantastic day!

Best,
Zephyr