Yann Lecun says his estimate for the creation of human-level AI is not that different from Sam Altman or Demis Hassabis - it is possible within 5-10 years
@ChatGPT_News_Ai
We're about to see a ton of new AI-enabled video game studios in the next couple of years.
The tech is already allowing game developers to operate at a massively higher output.
For example, Nvidia recently announced Edify 3D, which turns prompts into an entire 3D scenes that you can manipulate, pull apart, and edit.
The rundown of how it works:
—Uses 2 AI image generation models (diffusion guided by ControlNet) to create color and depth views from multiple angles
—Then uses a reconstruction model to convert them into 3D artifacts
—Then adds detailed 4k textures using physical based endering
—The result? A holodeck like scene creation in minutes.
An equivalent workflow using Epic Games’ Unreal Engine would require weeks of skilled 3D artist time, as the user would have to click through multiple layers of menus to make detailed edits.
@ChatGPT_News_Ai
Eric Schmidt says he has been shocked recently by how quickly China are catching up to the US in AI and that the "threat escalation matrix" goes up with each level of improvement in the technology
@ChatGPT_News_Ai
Runway just dropped a suprise AI image generation model called Frames — and the outputs are absolutely indistinguishable from reality.
Just days after the impressive video expansion tool launch, Runway is now cooking with a new image model.
@ChatGPT_News_Ai
Gemini Advanced now remembers your preferences for better answers tailored to you.
And you can manage or delete this anytime.
Similar to ChatGPT Memory feature
@ChatGPT_News_Ai
AI image editing just got another massive upgrade.
BlackForestLabs just dropped FLUX.1 'Tools,'which allow users to add way more control to creations.
The release includes four new features:
—Fill: Next-gen inpainting (including text) that beats Ideogram 2.0 and open-source alternatives
—Depth: Structure-preserving image edits using depth maps, outperforming Midjourney retexture
—Canny: Edge-based image control that outperforms current solutions
—Redux: Mix and recreate images with text prompts
Each tool was released with two versions: A full [Pro] model and an open-access [Dev] model for development purposes.
The [dev] versions outperform paid solutions, while [pro] sets completely new heights across benchmarks.
These type of tools are a Photoshop-type moment for AI image manipulation—but this time, it won’t be locked behind expensive software.
Plus, with Black Forest Labs’ integration with xAI’s Grok, millions of users are likely about to have access to easy-to-use, professional-g...
NVIDIA’s Project GR00T launches six workflows to boost humanoid robot vision and skills, including grasping.
@ChatGPT_News_Ai
CMU Researchers introduces "ManipGen"
Zero-shot robot that organizes and tidies from text commands, no demos needed!
@ChatGPT_News_Ai
ChatGPT can now use your Apps on Mac
This is (probably) a first step toward ChatGPT seeing everything on your computer and having full control as an agent
What you need to know:
It can write code in Xcode/VSCode
It can make a git commit in Terminal/iterm2
If you give it permission of course
Available right now to Plus and Team users
Coming soon to Enterprise and Edu
It’s an early beta
>The OpenAI team has been building towards this for almost a year
>The next step beyond this would be to allow ChatGPT to control see and control your desktop as an agent
>Bloomberg just reported that this agent,called the “Operator” will be released in January (covered in depth in The Rundown AI newsletter this morning)
>So App Usage could essentially be the OpenAI team doing bug fixes with humans in the loop feedback prior to Agent release
>Google DeepMind, Anthrophic,and several other startups (some which are in stealth) also reportedly have ag...
OpenAI introduces "Predicted Outputs"
It cuts GPT-4o latency by using a reference string, ideal for quick edits and updates.
@ChatGPT_News_Ai
Microsoft announces "Magentic-One"
Agent framework that autonomously tackle tasks by planning, acting, and coordinating.
In the demo, it shows how it orders a sandwich
@ChatGPT_News_Ai
Jensen Huang says there are 3 types of robots that can be manufactured at high volume: cars, drones and humanoids, and the leap from generative AI to general robotics is "very close"
@ChatGPT_News_Ai
NVIDIA introduces Edify 3D AI model
You can generate high quality 3D assets from text or image input under 2 minutes!
With PBR materials, quad-mesh topologies, and 4k textures.
@ChatGPT_News_Ai
NEW: ChatGPT’s ‘Live Camera’ video features were just found in the code of the latest beta.
Six months after their initial demo, OpenAI's visual AI could be ready for testing — and Advanced Voice Mode could soon be getting ‘eyes’
The code discovered in v1.2024.317 reveals:
—Live camera functionality
—Real-time processing
—Voice mode integration
—Visual recognition capabilities
The tech was initially showcased in an OpenAI demo in May, with Advanced Voice Mode interacting with a dog in real time
@ChatGPT_News_Ai
Jensen Huang says AI agents can't do 100% of any job, but they will do 50% of the work for 100% of people, and AI won't take your job; AI used by somebody else will take your job
@ChatGPT_News_Ai
Shoplexity is Here
Perplexity is taking on Amazon with its new Shopping experience.
It allows you to search for your plans (rather than just what you want to buy),and it will figure out what you need.
You can also take a picture of any item, find it online, and buy it all on Perplexity.
This essentially converts the entire world into a showroom and makes it shoppable.
This is a direct threat to Amazon.Some quick stats:
>54% of product searches in the US start on Amazon
>50% of all products sold on Amazon is sold by third-party sellers
>In effect Amazon leases out space on its website to external sellers, and provides various services including shipping, storage and money back guarantees.
>Perplexity is starting to create an Amazon like experience out of the open web
And also finally we see a path to revenue for Perplexity:
> That pair of Bose headphones you took a photo of is expensive
> Have you considered this pair of very well reviewed [No Brand] headphones?
>Ad revenue that doesn’...
Vinod Khosla says AI will do 65% of jobs in 20 years, but an increase in GDP from 2% to 5% will lead to an era of abundance where work isn't about servitude but about what people want to do
@ChatGPT_News_Ai
Hume AI introduces new app
AI assistants with unique voices, personalities, and enhanced AI tools like Claude 3.5
@ChatGPT_News_Ai
Decart AI dropped "Custom Worlds" for their real-time AI game engine Oasis
@ChatGPT_News_Ai
OpenAI's Kevin Weil says AI models "are going to get smarter at an accelerating rate" with voice and translation at a "magical" level already and vision to come
@ChatGPT_News_Ai
news: OpenAI's upcomning Orion model shows how GPT improvements are slowing down
It's prompting OpenAI to bake in reasoning and other tweaks after the initial model training phase.
@ChatGPT_News_Ai