Hi folks!👋🏻 This is The Prompt! We're your favorite AI newsletter that's like a freshly baked batch of cookies - we serve up the perfect mix of AI news, warm from the oven and ready to satisfy your craving!

Let's get it 🍪

FEATURED

🐶 Bark: Highly realistic text-to-audio model

Bark is the latest text-to-audio model that can create multilingual speech.

And I can’t stress how realistic the outputs are!

It can also create music, background noise, and simple sound effects. Just add music notes around your prompt, and the ♪ model will “sing your text prompt ♪

Plus (and this is very interesting), it can generate nonverbal communication like laughing, sighing, and crying if you add a metatag in the text like [laughs], [sad].

Wanna see for yourself? See these examples, try the demo, or get the code.

NEW TECH

New model scales LLMs to 1M tokens and beyond 🔥

This new language model can remember up to 2M tokens from input (in comparison to GPT-4 which can only take 32K).

You know what this means right?👀

Let me give you an example: The entire Harry Potter series is ~2.5M tokens.

With this model, you can create your own fantasy novel series.

Isn’t that amazing or what?

WHAT ELSE IS GOING ON

🦙 ChatGPT can forecast stock movements. Researchers took financial covers from October 2021 (which is where the knowledge cutoff for ChatGPT is), and prompted the AI to make predictions. Results: ChatGPT was able to guess the stock moves. Crazy times.

👀 Musician Grimes is willing to “split 50% royalties” with AI-generated music. She posted on Twitter that she will treat AI as any other artist that she collaborates with.

🏋🏻‍♀️  ChatGPT on TedX = AI goes mainstream? 😅 OpenAI’s co-founder Greg Backman talked about the inside story of ChatGPT and the potential it has.

RESOURCES

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

📚 Dataset and pre-trained models for GPT4Tools [ Model release ]

👋🏻 How to trail your own language models [ well written explanation ]

🎥 Faster Whisper transcription code [ Github specs]

TOOLBOX

The latest AI tools to use or get inspiration from.

• Existential: Career recommendations for your interests & skills

• Gems: AI knowledge assistant

• Objaverse: 800K+ Annotated 3D Objects

• BeatlyMusic: Discover trending AI music of your favorite artists

• Redense AI: Build newsletter using AI + links

• Gnbly: Automate your calls: reservations/ service calls

• JailbreakChat: Collection of jailbreaks for ChatGPT

• GPT-Driver: AI-powered QA bot

PROMPT OF THE DAY

TOOL

Midjourney

PROMPT

high luxury symmetrical close - up shoot of Lamborghini, anamorphic lens, ultra realistic, hyper detailed --ar 16:9 --q 2 --v 5

RESULT

LATEST PAPERS

• CLaMP: Contrastive Language-Music Pre-training for Cross-Modal Symbolic Music Information Retrieval

• Factored Neural Representation for Scene Understanding

• Scaling Transformer to 1M tokens and beyond with RMT

• Emergent and Predictable Memorization in Large Language Models

• Think Before You Act: Unified Policy for Interleaving Language Reasoning with Actions

• Multi-aspect Repetition Suppression and Content Moderation of Large Language Models

ch