It's been an absolutely insane week in the world of AI! OpenAI, Google, Microsoft, Meta, and all the major AI art and video generators have dropped new announcements and rolled out new features. In this blog post, we're going to rapid-fire through all the crazy AI news that happened this week without rambling too much. So let's dive right in!
OpenAI's AI Dev Day Highlights
AI Agents Coming in 2025
This week was OpenAI's AI Dev Day, and while most of the announcements were geared towards developers rather than end-users of ChatGPT, there were still some interesting tidbits. During a fireside chat, Sam Altman hinted that AI agents are on target to launch in 2025. These agents are independent AI models capable of performing a range of tasks without human input and could be available in ChatGPT soon.
Introducing Canvas
OpenAI rolled out the new Canvas feature, now live for all ChatGPT Plus subscribers. Canvas offers a complete overhaul of the UI inside ChatGPT, allowing for more interactive and editable responses.
- Text Editing and Formatting: Select text and ask ChatGPT to rewrite specific paragraphs or adjust formatting.
- Coding Features: Review code, add logs, comments, fix bugs, or port code to a different language directly within ChatGPT.
- Quick Shortcuts: Use quick commands like
/reason
,/search
, and/picture
to enhance your prompts effortlessly.
Meta's Llama 3.2 Release
Meta released Llama 3.2, a significant leap forward in AI technology that's open-source and can run directly on your device—even without an internet connection.
- Vision Capabilities: Larger models (11B and 90B) now understand both text and images.
- On-Device AI: Lighter models (1B and 3B) are perfect for on-device AI applications, even on mobile phones.
- Massive Context Window: Supports 128,000-token context windows, allowing for entire books' worth of information in a single conversation.
- Optimized for Mobile Hardware: Specifically optimized for Qualcomm and MediaTek hardware.
- Open-Source and Safe: Available for everyone to use and improve upon, with new safeguards like Llama Guard 3 for responsible use.
Microsoft's AI Announcements
Microsoft unveiled a slew of new features, especially for the new Copilot Plus PCs with Neural Processing Units (NPUs).
Recall Feature
Now available, the Recall feature lets you revisit any activity on your PC without sending data back to Microsoft. It's like an enhanced browsing history but for everything you do on your computer.
Click-to-Do Feature
When you click on images or text, new AI-related options appear:
- For Images: Visual search with Bing, blur backgrounds, erase objects, remove backgrounds.
- For Text: Rewrite, summarize, explain text, send emails, or perform web searches.
Enhanced Windows Search
AI-powered search can now find images based on context—even if the files are unnamed and you're offline.
Super Resolution in Photos
Upscale images directly within the Photos app using AI.
Generative Fill and Erase in Paint
Microsoft Paint now features Generative Fill and Erase, similar to Adobe Photoshop's capabilities.
Copilot Labs and Copilot Vision
- Think Deeper: Allows Copilot to reason through more complex problems.
- Copilot Vision: Understands the page you're viewing and can answer questions about its content.
Bing Generative Search Updates
Microsoft expanded Bing Generative Search to cover more informational queries.
- Deep Research Answers: Provides detailed explanations and helps solve complex problems.
- Sponsored Results: Ads will now appear in AI-generated search overviews, working with major publishers.
Google's AI Enhancements
Google made several updates to Google Lens and search:
Video Understanding
Google Lens can now understand and answer questions about videos. Simply ask, "Why are they swimming together?" while watching a video, and it will provide an AI-generated response.
Voice Questions
Speak to Google Lens to get answers about images or scenes you're viewing.
Shop What You See
Find where to purchase items you spot by simply taking a picture.
Song Identification
Identify songs through the new "Circle to Search" feature.
Organized Search Results
AI now organizes your search results for easier navigation and comprehension.
Ads in AI Overviews
Just like Bing, Google will show ads in AI-generated responses, partnering with major publishers to compensate them appropriately.
Advancements in Large Language Models
NVIDIA's NVLMD 1.72B
NVIDIA announced a new open-source large language model capable of vision tasks, rivaling leading proprietary models like GPT-4 Vision.
Google's Gemini 1.5
Google introduced Gemini 1.5 Flash 8B, a model that's 50% cheaper and offers higher rate limits with lower latency, aimed at developers using the API.
AI Imagery and Video Generation
Flux 1.1 Pro Release
Black Forest Labs released Flux 1.1 Pro, an upgraded AI imagery model with improved text and image quality. It's available for free on platforms like Glyph for a limited time.
Leonardo AI Updates
Leonardo AI rolled out new features:
- Style Reference Feature: Upload up to four reference images to direct the aesthetics of your output.
- Image-to-Image with Phoenix Model: Now supports the Phoenix preset for image-to-image generation.
Pika 1.5 Model
Pika released a new AI video generation model capable of creating meme-style videos where objects can be squished, exploded, or manipulated in creative ways.
ByteDance's New AI Video Generator
ByteDance, the company behind TikTok, revealed a new AI video generator said to rival state-of-the-art models. The tool can generate high-quality, 10-second videos with impressive realism.
AI Legislation Updates
California Governor Vetoes SB 1047
Governor Gavin Newsom vetoed a bill that would hold AI companies responsible for catastrophic harm caused by their models if misused by others. The AI community viewed this as a positive step toward more balanced regulation.
Judge Blocks Deepfake Bill
A California judge blocked AB 2839, which targeted distributors of AI deepfakes, citing freedom of speech concerns. The only part that stuck was the requirement to label deepfakes involving political figures.
Other Noteworthy Updates
Amazon's New Fire Tablets
Amazon is rolling out new Fire tablets equipped with AI tools like writing assistance, webpage summaries, and creating wallpapers from prompts.
Robotic Advancements
A quadrupedal robot capable of climbing ladders was introduced, potentially useful for high-risk tasks in dangerous environments.
Conclusion
That's all for this week's insane AI news roundup! With groundbreaking announcements from OpenAI, Google, Microsoft, Meta, and more, it's clear that the AI landscape is evolving at a rapid pace. Stay tuned for more updates, and don't forget to check out the latest tools and news on FutureTools.io.
Stay Ahead of the Curve
If you want to stay on the cutting edge of AI, make sure to like this post and subscribe for more updates. Let's stay ahead of the curve together!