• AI Time to Impact
  • Posts
  • . . AI: Google Says Gemini Advanced Beats Human Experts (2.8.24)

. . AI: Google Says Gemini Advanced Beats Human Experts (2.8.24)

1X Androids, HyperWrite, Stable Audio

Friends, here are today's top stories in AI. Today's developments represent another leap toward creating more intuitive, autonomous, and intelligent systems that blur the lines between human and machine capabilities. Plus check out the weird robots in the 1X story.

Speaking of up-leveling capabilities, if you’re working on how organizations can utilize AI, newsletter community member Dr. Phil Hendrix has published a remarkable new Roadmap for Deploying and Capitalizing on Generative AI. It’s an impressive resource.

And now here’s today’s top stories…

-Marshall Kirkpatrick, Editor

First impacted: Software developers, mobile app users
Time to impact: Short

Google has launched Gemini Advanced and a mobile app, previously known as Bard. Google says that Gemini Advanced, powered by the Ultra 1.0 model, can perform complex tasks such as coding and logical reasoning. The release also stated that "in blind evaluations with our third-party raters, Gemini Advanced with Ultra 1.0 is now the most preferred chatbot compared to leading alternatives". Google is offering Gemini Advanced as part of its new Google One AI Premium Plan at $19.99/month [Bard becomes Gemini: Try Ultra 1.0 and a new mobile app today] Explore more of our coverage of: Google Gemini Advanced, AI Coding, Mobile AI Applications. Share this story by email

First impacted: Robotics experts, AI experts, Android operators
Time to impact: Medium

1X Technologies says it has developed a system that uses neural networks to fully control androids, with a single network managing all functions. The company is seeking AI researchers to improve the system and android operators to collect data and train models, enabling the androids to learn new skills within minutes. Check out a video of the human-like bots in action at the link below. [All Neural Networks. All Autonomous. All 1X speed] Explore more of our coverage of: Neural Networks, Android Control, AI Research. Share this story by email

First impacted: Chess players, AI researchers
Time to impact: Short

Google Deepmind has trained a transformer model on 10 million chess games, and reached a Lichess blitz Elo of 2895, according to its latest research paper. The interesting part is that this model operates without explicit search algorithms. This level of performance in chess without search algorithms suggests that the model has developed a deep understanding of chess strategies and principles, and a significantly more efficient decisioning capability. [Paper page - Grandmaster-Level Chess Without Search] Explore more of our coverage of: Google Deepmind, Chess AI, Transformer Models. Share this story by email

First impacted: AI developers, Business professionals, IT specialists
Time to impact: Short

HyperWrite is preparing to launch Agent Trainer and Agent Studio in the coming weeks. The tools allow users to instruct AI to automate recurring and repetitive tasks, according to the CEO, Matt Shumer. Users can record their workflow once for the AI to replicate. The link shows a video where an agent is trained to order the 'usual' lunch. [via @mattshumer_] Explore more of our coverage of: AI Automation Tools, Workflow Optimization, HyperWrite. Share this story by email

First impacted: Audio Engineers, Data Scientists
Time to impact: Short

Stability.AI has launched their latest audio model (Stable Audio), making it possible to quickly create high-quality, custom audio based on written instructions. The model leverages PyTorch Lightning for multi-GPU and node training and fine-tunes from prior trained checkpoints. [GitHub - Stability-AI/stable-audio-tools: Generative models for conditional audio generation] Explore more of our coverage of: PyTorch Lightning, Stable Audio, Amazon S3. Share this story by email

That’s it! More AI news tomorrow.