AI News: We're One Step Closer To AGI This Week!

1. OpenAI outlines five levels towards AGI.

🥇92 00:00

Levels progress from chatbots to organizations, with current focus on human-level problem-solving.

• Level one involves chatbots and conversational AI like ChatGPT.
• Level two focuses on human-level problem-solving.
• OpenAI aims to advance through each level towards achieving true AGI.

2. Concerns arise over OpenAI's business practices.

🥈85 03:36

Whistleblowers allege OpenAI restricts employees from discussing issues and removes rewards for whistleblowing.

• Allegations include illegal practices related to employee rights and equity.
• OpenAI refutes claims, stating they have policies to protect whistleblowers.

3. OpenAI introduces Strawberry, a new reasoning technology.

🥈89 05:06

Strawberry aims to autonomously navigate the internet for deep research, potentially bridging towards AGI.

• Strawberry's goal is to plan ahead and perform actions for deep research.
• It may represent a significant step towards AGI capabilities.

4. Andre Karpathy launches AI education company, Eureka Labs.

🥈88 08:43

Eureka Labs aims to revolutionize education by combining subject matter experts with AI teaching assistants.

• The model involves scaling teaching through AI assistants fluent in multiple languages.
• Teachers design courses, which are then supported and scaled by AI assistants.

5. Anthropic's CLA now available on Android.

🥈82 10:23

Anthropic's conversational AI app expands to Android, offering an alternative to other AI chat apps.

• Users can now access CLA on Android devices for voice conversations with AI.
• Provides a choice for users seeking AI chat functionality on Android.

6. Google introduces AI-powered video creation app, Google Vids.

🥈87 11:20

Google Vids integrates AI for creating slide-style videos, enhancing workspace productivity.

• Users can prompt the app to generate videos based on selected styles and content.
• Designed to streamline video creation for work purposes within the Workspace Suite.

7. YouTube launches AI-generated conversational radio.

🥈86 13:07

Users can create custom radio stations by describing preferences, showcasing AI's evolving capabilities.

• AI-generated conversational radio allows users to request music based on descriptions.
• YouTube continues to innovate with AI-driven features for enhanced user experience.

8. Apple clarifies not using Pile data in their AI model.

🥇92 13:47

Apple confirms using Pile data for research but asserts their AI model is not trained on it, maintaining data integrity.

• Apple's AI model does not incorporate Pile data, ensuring data separation.
• Maintaining transparency in data sourcing enhances trust in AI applications.

9. Microsoft's Designer platform expands across various apps.

🥈88 14:45

Microsoft's Designer platform integrates into multiple Microsoft apps, facilitating easy creation of diverse visual content.

• Co-pilot feature enhances design creation efficiency.
• Mobile app availability enhances design flexibility and accessibility.

10. Mistral introduces Codestrol Mamba for code generation.

🥈89 16:08

Mistral unveils Codestrol Mamba, a code generation model with open-source access and high token capacity, appealing to developers seeking efficient coding solutions.

• Codestrol Mamba offers a substantial token capacity advantage over existing models.
• Fast response time enhances coding productivity and effectiveness.

11. Amazon launches Rufus AI shopping assistant within its app.

🥈87 16:50

Amazon introduces Rufus, an AI shopping assistant embedded in the Amazon app, providing shopping and political information directly to users.

• Rufus offers personalized shopping recommendations and political insights.
• Integration within the Amazon app enhances user convenience and engagement.

12. Meta restricts multimodal models in the EU due to regulatory uncertainties.

🥈86 17:25

Meta limits multimodal model availability in the EU, citing challenges in complying with GDPR and uncertain regulatory landscape.

• EU regulations pose obstacles for Meta's AI model deployment.
• Differences in regulatory environments impact AI model accessibility.

13. Innovative image adjustment using knobs demonstrated.

🥈88 18:22

A creative approach using knobs to modify image parameters showcased, offering a unique and interactive image editing experience.

• Stable diffusion technique utilized for image adjustment.
• GitHub availability allows for experimentation with image modification.

14. AI transforms selfies into 3D-printable models.

🥈85 18:56

AI-powered app converts selfies into high-quality 3D models suitable for 3D printing, showcasing advancements in personalized digital creations.

• App enables easy creation of personalized 3D models from selfies.
• Potential applications in personalized merchandise and digital art.

15. AI achieves high accuracy in sex determination from dental X-rays.

🥈89 19:31

AI attains 96% accuracy in identifying sex from dental X-rays, demonstrating potential forensic applications in skeletal remains analysis.

• Forensic identification benefits from AI's accurate sex determination capabilities.
• Limitations in accuracy noted for younger individuals.

16. OpenAI launches GPT-40 Mini model for enhanced efficiency.

🥇92 20:26

OpenAI introduces GPT-40 Mini as a faster and smarter model, supporting text, vision, and various inputs, aiming for improved performance and cost-effectiveness.

• GPT-40 Mini offers enhanced text and vision processing capabilities.
• Comparative evaluation highlights GPT-40 Mini's competitive performance.