Grok’s now-removed ‘Aurora’ AI Image Generator is too realistic

Hello, AI Enthusiasts!

Welcome to FavTutor’s AI Recap! We’ve gathered all the latest and important AI developments for the past 24 hours in one place, just for you.

In Today’s Newsletter: 😀 

• Grok’s now-removed ‘Aurora’ AI Image Generator is too realistic

• Google says its new AI weather forecast model is too good

• AI models stumble in complex game tests, scoring low across the board

• Google releases PaliGemma 2 open-source vision language model

Grok
😲 Grok’s now-removed ‘Aurora’ AI Image Generator is too realistic

Last weekend, some users of Elon Musk's X saw Grok running a new image generator called Aurora for a few hours. TechCrunch reported Aurora's beta version appears to have few creative limitations, and in brief tests created images of public and copyrighted figures.

Insights for you:

• X’s in-house chatbot Grok has debuted a photorealistic image generator known as Aurora.

• Some users who reported having access to it later said their Grok had reverted to its older model, Flux.

• The release of Aurora comes after X made Grok free for all users.

Streamline your development process with Pinata’s easy File API

• Easy file uploads and retrieval in minutes

• No complex setup or infrastructure needed

• Focus on building, not configurations

Try today!

Google
⛅️ Google says its new AI weather forecast model is too good

Google Deepmind’s new AI model GenCast is accurate enough to compete with traditional weather forecasting, according to the researchers. GenCast is a diffusion model, similar to those you may have seen in AI image generators. However, this one is tuned specifically for Earth’s geometry.

Insights for you:

• Google’s DeepMind team unveiled an AI model for weather prediction this week called GenCast.

• DeepMind researchers said they found that GenCast outperforms the European Centre for Medium-Range Weather Forecasts’ ENS, the world’s top operational forecasting system.

• Google says GenCast is part of its suite of AI-based weather models, which it’s starting to incorporate into Google Search and Maps.

AI Research
🎮️ AI models stumble in complex game tests, scoring low across the board

A new testing platform called BALROG shows that even top AI language models fail when faced with complex gaming challenges. OpenAI's GPT-4o, the best performer in the tests, only achieved 32% possible scores across all games tested.

Insights for you:

• Researchers have developed BALROG, a benchmark platform that tests large language models (LLMs) and visual language models (VLMs) in a variety of game environments, from simple tasks to complex games such as the NetHack Learning Environment.

• The test results show clear limitations of current AI language models: even top performers such as OpenAI's GPT-4o scored on average only 32% of the possible points.

• In complex games requiring long-term planning, all models failed almost completely. The deficits in image-based decision-making were particularly striking: When the language models were presented with visual representations of the game environments, they performed even worse than with purely text-based input.

Google
👓️ Google releases PaliGemma 2 open-source vision language model

Google has released the next generation of its open-source vision language model, PaliGemma 2. This model offers scalable performance across various tasks, enhanced image descriptions, and a wide range of applications.

Insights for you:

• Google has introduced PaliGemma 2, which comes in various sizes (3B, 10B, 28B parameters) and supports multiple image resolutions (224px, 448px, 896px).

• PaliGemma 2 integrates the SigLIP-So400m vision encoder with the Gemma 2 language models, enabling it to recognize not only objects but also actions, emotions, and narrative contexts within images.

• According to Google, PaliGemma 2 performs well in diverse domains, including recognizing chemical formulas, interpreting musical scores, and analyzing X-ray images.