- FryAI
- Posts
- OpenAI introduces AI image detector
OpenAI introduces AI image detector
Hunter Kallay
May 09, 2024
Good morning! Are you hungry for AI updates? If so, you’re in luck, because we are serving up the latest on a silver platter for you to enjoy. 🍽️
🤯 MYSTERY AI LINK 🤯
(The mystery link can lead to ANYTHING AI related. Tools, memes, articles, videos, and more…)
Today’s Menu
Appetizer: OpenAI introduces AI image detector 🔍
Entrée: Microsoft to build $3.3 billion AI hub 💰
Dessert: AI-powered penguins are exploring the ocean 🐧
🔨 AI TOOLS OF THE DAY
📊 MockThis: Quickly generate mock data for various purposes. → check it out
🎨 Website Colors AI: Get the colors of your favorite websites using AI. → check it out
🎮 PS2 Filter: Give your photos and videos PlayStation 2 aesthetics. → check it out
OPENAI INTRODUCES AI IMAGE DETECTOR 🔍
Introducing a new classifier to help researchers identify content created by DALL·E 3.
We are also joining the Coalition for Content Provenance and Authenticity (@C2PA_org) Steering Committee to promote industry standards.
— OpenAI (@OpenAI)
2:05 PM • May 7, 2024
Are we in too deep with deepfakes? OpenAI doesn’t think so! 💪
What’s new? OpenAI has recently introduced new methods to help identify and verify content created with its DALL-E AI image generator, aiming to enhance transparency and trust in AI-generated content.
What are the new methods? OpenAI’s initiatives include a new image detection classifier and tamper-resistant watermarking methods.
Image detection: OpenAI’s new image detection classifier uses AI to predict whether a picture was created by DALL-E 3, achieving an accuracy rate of around 98%. Even if an image is cropped, compressed, or altered in saturation, the classifier remains effective. However, its accuracy in identifying content from other AI image generators, such as Midjourney, is significantly lower. With these foreign image generators, the model only detects AI generation with 5-10% accuracy. | Watermark techniques: OpenAI is expanding watermarking techniques for audio, tagging content from its Voice Engine text-to-speech platform with invisible signals. This aligns with the company’s ongoing commitment to transparency, including its new partnership with the Coalition of Content Provenance and Authenticity (C2PA). |
Why is this significant? Deepfakes are arguably the most dangerous consequence of generative AI to emerge over the past two years. Tools like DALL-E and Midjourney have become massively popular, and hundreds of other tools for image generation and manipulation have also arisen. This has led to widespread misinformation and misrepresentation in a variety of cases, such as celebrity impersonation, political deception, and in some cases harm to the youth. Although OpenAI’s new detector has only shown to be effective on their own tools, it represents a good start to deepfake detection and serves as a significant pioneering model for others to follow.
“Our efforts around provenance are just one part of a broader industry effort–many of our peer research labs and generative AI companies are also advancing research in this area. We commend these endeavors—the industry must collaborate and share insights to enhance our understanding and continue to promote transparency online.”
MICROSOFT TO BUILD $3.3 BILLION AI HUB 💰
Microsoft continues to pour money into AI development. 🤑
What’s up? Microsoft has announced a $3.3 billion investment to establish a data hub in Mount Pleasant, Wisconsin, focusing on training employees and manufacturers in the best use of AI.
Why? This investment will create 2,300 union construction jobs and 2,000 permanent jobs over time. Additionally, the center plans to train about 100,000 workers in AI and will provide support to local companies looking to adopt AI technology. This aligns with Microsoft’s commitment to making AI resources and skills training more widely available.
AI-POWERED PENGUINS ARE EXPLORING THE OCEAN 🐧
Image: EvoLogics
Q: Who is the penguin’s favorite relative?
A: Aunt Arctica. ☃️
What’s up? EvoLogics, an underwater tech company based in Germany, has built a robotic, AI-powered penguin to help with ocean research.
What does the robot do? The penguin-shaped robot, called the Quadroin AUV, is 44 inches long and weighs 55 pounds out of water. The robotic penguin can dive up to 492 feet and swim up to 11.5 mph, holding a 6-10 hour charge. While swimming through the ocean, the camera-equipped robot uses a proprietary AI recognition model to collect image data and recognize items. Once data is received, the robot transmits it via wifi through its multifunctional antenna. This allows researchers to identify underwater objects and explore unknown areas of the ocean in more precise and danger-free ways.
TASTE-TEST THURSDAY 🍽️
When you have a question, do you ask a chatbot or search the web?Is "Just Google it" being replaced by, "Ask ChatGPT"? |
HAS AI REACHED SINGULARITY? CHECK OUT THE FRY METER BELOW:
What do ya think of this latest newsletter? |