Gemini

Google plans to bring Project Astra to the Gemini app and its other products, starting later this year. While there's no specific launch date yet, Google is committed to making these capabilities available, with CEO Sundar Pichai stating it's their "vision for the future of AI assistants."

Google confirmed the memo and outlined a clear set of actions to address the issue, including structural changes, updated product guidelines, improved launch processes, robust evaluations and red-teaming, and technical recommendations.

Given all the red-teaming, testing, and responsible AI practices that Google touts with each model release, how did such an obviously flawed feature end up in one of the world's most leading foundation models?

The feature is called "Help Me Write," and is powered by Gemini. The AI analyzes the content and context of the website you are on to provide tailored recommendations.

This provides Premium subscribers access to AI-powered productivity features like helping draft emails, itineraries, budgets, presentations and more.

By integrating generative AI directly into familiar Workspace apps, Google is not only hoping to make its AI tools more accessible but also ensure they are practical and relevant to everyday tasks.

Some critics have accused Google's culture of pushing a "pro-diversity bias," while others see this as a good faith but imperfect attempt to showcase variety. In any case, the company says it is prioritizing upgrades to improve accuracy.

This enables much more sophisticated reasoning and problem-solving abilities than previous models.

Gemini Advanced and the mobile apps are launching first in the US in English and will continue expanding globally and adding language support over time.

This consolidation of branding signifies Google's intent to unify its AI offerings under the Gemini moniker, streamlining its AI identity across products and services.

Early next year, Bard will receive another significant upgrade with Gemini Ultra, Google's most capable AI model intended for complex, multimodal tasks.

Gemini is the first general purpose foundation model to outperform GPT4-V on the majority of benchmarks.