Hi folks!👋🏻 This is The Prompt! We're here to deliver freshly wrapped AI news and tech. With extra quac on the side!🥑

Let's get it

FEATURED

Create videos with your vision 🛡

What you see is what you get - literally.

This model can reconstruct videos from brain activity. 

It has two modules: The first module learns general visual features from brain activity, while the second module fine-tunes these features to generate high-quality videos with accurate semantics.

Last year, the same authors released a model that reconstructs images from vision, and this is the next phase of their work.

And the results are unbelievable, see for yourself 👇🏻

NEW TECH

• A model that does multi-subject image gen right. 

  Current challenges with Stable Diffusion/Midjourney and alike is that it’s hard to generate two different subjects in one image, as they often blend features among subjects FastComposer makes it possible for a model to differentiate two subjects, by using subject embeddings.

• GPT detectors are biased against non-native English speakers. The authors say that their use for the evaluation of plagiarism in schools should be reconsidered.

• A new model that has similar accuracy as GPT-4. LIMA is a 65B LLaMa language model fine-tuned on only 1k curated samples w/o RLHF. The evaluation seems a bit shaky though, trained on 300 examples out of which 230 were written by the authors themselves.

WHAT ELSE IS GOING ON

🦙 Neeva is shutting down its AI search engine. Why? Google/Bing/ChatGPT Browsing competition. They’re pivoting to building LLMs for enterprises.

👀 Apple is hiring experts in gen-AI like crazy. Especially in the gen-image space. They pre-announced the personal voice app, but maybe more is coming. They also banned ChatGPT.

🏋🏻‍♀️  Meta revealed their custom AI chip. They’re promising more computing power and efficiency than standard CPUs.

RESOURCES

The best resources we came across lately that will help you become better at writing prompts & building AI apps.

📚 Stable Diffusion database [ list of every plugin, tool & app for SD]

👋🏻 From Sketch to Photo with Midjourney [Twitter thread step-by-step tutorial]

🎥 AI-generated horror movie [ Twitter thread with step-by-step tutorial]

💰 Tips for optimizing Pinecone costs [ Twitter thread ]

TOOLBOX

The latest AI tools to use or get inspiration from.

• Schrodi: Create children's stories

• Chatable: Coaching co-pilot

• MangoSEO: AI-powered programmatic SEO

• Puddl: OpenAI usage analytics and tracking tool

• Trible: AI + no-code web builder

• Yoodly: Speech coach

PROMPT OF THE DAY

TOOL

Midjourney

PROMPT

A hyper - realistic GoPro selfie of Nefertiti in an Egyptian palace --ar 3:2‘

RESULT