- Inclined.ai
- Posts
- 🤖 Stability AI Is Fighting Meta
🤖 Stability AI Is Fighting Meta
PLUS: DreamGPT Is Reimagining Creativity
What's up? You're reading Inclined AI. If you are celebrating today, please do so responsibly.
Here are the headlines, dude:
Stability AI launches a new LLM
DreamGPT tries to add variety to AI
Google employees dislike Bard
Drama with the AI Drake song continues
STABILITY AI CONTINUES TO OPEN ACCESS TO AI MODELS
Stability AI is putting the “open” in “open AI.”
They launched StableLM yesterday for commercial and research purposes to compete with Meta’s LLaMA for commercial and research purposes.
Their context lengths match ChatGPT (4096 tokens instead of LLaMA’s 2048) and are under a lax creative commons license so that people can use the two versions of their model without stressing over permissions.
However, their fine-tuned model is limited to researchers, so that’s still a speed bump.
You’re getting a lot of random bits tossed out at you; let’s narrow it down.
Stable Diffusion is the most popular model from Stability AI. Their new Large Language Model is the Stable Diffusion for conversational AI. You can’t run a personal clone of GPT-4 on your laptop, but you can use StableLM.
Plus, it will perform better than the current options. Alpaca and LLaMA are confusing messes for licensing.
Announcing StableLM❗
We’re releasing the first of our large language models, starting with 3B and 7B param models, with 15-65B to follow. Our LLMs are released under CC BY-SA license.
We’re also releasing RLHF-tuned models for research use. Read more→ stability.ai/blog/stability…
— Stability AI (@StabilityAI)
3:12 PM • Apr 19, 2023
StableLM is the Toyota Camry of chat AI—dependable and obtainable.
Hours after launching the GitHub repository, open-source AI developers found ways to help people access the tool with a few clicks of their mouse.
The AI community is fighting to keep up with the profit machines, and more models becoming available is always a good thing. We need transparent research and development.
An open-source future is how we approach AGI faster.
Want to create stunning emails that wow your subscribers?
Meet Mailmaker by Mailscribe, the ultimate email creation platform, with a new tool that makes email marketing a breeze.
With Mailmaker’s new AI Assistant, you can:
Type in your email topic and let the AI Assistant generate a draft for you
Use the AI Assistant’s suggestions to improve your grammar, tone, and style
Design, edit, and send gorgeous emails with drag and drop that the AI will then translate for you instantly
Mailmaker is easy, fun, and powerful. It helps you grow your business and connect with your audience on a deeper level.
Don’t miss this chance to take your email marketing to the next level.
Sign up for Mailmaker and access the best email creation platform instantly. You can get 50% off your first purchase using our code: inclinedai50
WHAT DOES AI DREAM OF?
Prompt: Electric sheep hopping over fence Style: Children’s Book
A Redditor posted a new concept called “DreamGPT” that they coded using GPT-4 as an assistant.
The goal is to inspire creativity by giving Large Language Models a dream state. The idea sounds like something from an Isaac Asimov book.
Dreams are a tricky concept to grasp when it comes to AI. For instance, Ask ChatGPT if it dreams, and it will say that creating outputs for stories and fiction is a form of dreaming.
Even the hallucinations and gibberish you sometimes get from conversational AI are a form of “AI dreams.”
The term is not new in the field of AI research.
But DreamGPT is different. It separates the ideation phase into two other concepts: the “Awake” and “Dream” phases.
When the model is awake, it examines topics and searches for information. Three agents are debating and digesting this information to mimic human consciousness.
Think of this stage as putting a puzzle together. Together the agents make a complete picture.
Then the AI smashes the puzzle and the pieces scatter. The model then finds new, unique ways to connect those inputs.
DreamGPT is looking for hidden, unassuming patterns.
The hope is these two steps work together to give you new, fresh outputs that show more reasoning.
It’s wild to think that “dreams” are already a part of the process and tie into all the models we see today.
Is DreamGPT the next logical step? Time will tell.
Quick Nuggets
📚 5 articles to read so you can explore the effects of generative AI
🤔 Changes? AI may alter society as we know it, or we’re falling for another hype cycle
🤫 Secret websites that helped make GPT-4 so smart
🤯 10 hours is all it took to launch and exit a Chrome extension using ChatGPT
👾 ChatGPT NPC is able to have novel conversations with players
🐦⬛ Quora’s CEO sat down with Semafor to talk about Poe and the impending shift in AI usage
🙊 Bard’s a pathological liar, according to some Google employees
👵 Grandma exploit is a fresh way to break chat AI tools
🦠 Malware can be spread by ChatGPT and other chatbots
🦜 Parrot search: LLMs struggle to identify other LLMs
📹 The University of Toronto created Latent Diffusion Models to generate high-resolution video
🛑 The FTC is worried that ChatGPT can lead to more fraud
❓ That AI Drake song is getting weird. Apparently, Ghostwriter might’ve pretended to get struck by copyright claims
🐻 Xi Jinping is struggling to control AI without crushing the country’s research
🔥 Fresh Products
Human or Not? - you chat with a human or AI, but which one? (link)
remoteBase - develop AI tools with the help of AI experts (link)
CollovGPT - inspire your next interior design makeover (link)
ChatGPT-2D - shows outputs in 2D map (link)
MagicSlides - creates professional slides in a jiffy (link)
My AskAI - add your own chatbot to your website (link)
Schrodi - make a personalized short story for anyone (link)
DeckAssistant - get an extra hand with your stream deck (link)
Searcholic - search ebooks and documents through AI (link)
Preppally - adds AI-assisted career path guidance (link)
Thunderclap - adds some oomph to your Twitter game (link)
Good Content, Otterman Empire
It’s pictures like this that make me happy that AI art exists.
That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.
Check Out Our Sunday Edition
Every week you can dive into more AI news and topics with us by subscribing to our premium edition.
We’ve written about:
If you’re not already subscribed, that’s okay. We’re offering a free 7-day trial so you can read this one. That’s how excited I am to post it.