Hi folks!👋🏻 This is The Prompt! We're like the Swiss Army Knife🇨🇭 of AI. We've got everything you need - news, research, tools, and more - all in one place.

Let's get it

PhotoGuard: Guard your photos against deepfakes 🛡

ML photo editing tools like StableDiffusion have made it easy for bad actors to create high-quality fake images.

This new technique, Photoguard, will make it impossible for AIs to manipulate your pictures. 

How it works: The code targets a specific element of the ML model that alters the photo, causing it to produce a photo that is either unrealistic or unrelated to the original image.

I am guessing that social media companies will soon adopt similar techniques.

Here are the links:

Github implementationPaperHugging face demo

 💥 The Latest

-

OpenAI will become extremely cautious with the deployment of its models. This means that users/use cases might become strongly monitored. Customized ChatGPTs are part of their roadmap as well.

Lots of AGI talk lately. Models are becoming more powerful by the day.

-

Meta released their GPT-3 competitor, LLaMa 🦙

Meta launched a group of models, but the largest one has 65b parameters which is about 1/3 of the size of GPT-3. This is a big advantage as it means 3x faster inference and 3x less memory. LLaMa outperforms other models on most benchmarks.

However, in their benchmarks, they don't mention GPT-3.5, so not sure how will it perform against the latest OpenAI models.

-

Instagram co-founders launch AI-news platform  

Back in the startup world, the cofounders of Instagram launched Artifact, an app that is using AI to curate news for its users. The app was inspired by Tiktok, and the AI will be optimized to measure how long users spend reading about a topic, rather than what generates the most clicks. 

📚 Learning lounge

• [Educational article] Everything you need to know about red teaming LLMs

• [Twitter thread with examples/code] How to improve accuracy of the "hands" with ControlNet and poses

• [7min video] Anime made with Stable Diffusion, Dreambooth, and Davinci 🤯

🛼 Toolbox

• Joblist.AI: Find ChatGPT job opportunities

• Powerpresent AI: Generate presentations on the go

• LatentSpace: Podcast for AI builders

• GPTIndex: Connect your LLM’s with external data without worrying about tokens

•  MemeMorph: Turn yourself into a meme

🤓 Latest AI papers

• MimicPlay: Robot Imitation Learning by Watching Human Play

• ZoeDepth: Zero-shot Transfer by Combining Relative and Metric Depth

• MUX-PLMs: Language Models with Data Multiplexing

• Modulating Pretrained Diffusion Models for Multimodal Image Synthesis

• Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedback

📸 AI Photo of the day

Midjourney version comparison using prompt: "photo of a fat alpaca"

❤️ If you like The Prompt, and want to support my work:

• Share The Prompt with a friend, and invite them to subscribe here.

• Book an ad in The Prompt (reply to this email if you’re interested)

• Check out my latest tool: VoicePen.ai (audio to blog post)

Thank you for reading! ✌🏼