Good morning human brains, welcome back to your daily munch of AI news.

Here’s what’s on the menu today:

• OpenLLaMA 🦙 

  A new player has entered the stage in the ongoing saga of open-source ChatGPT competitors.

• Text generation in AI art 🎨

  Stability.AI unveils an AI art generator called Deep Floyd that has the capability to generate text inside images.

• Your Daily Munch 😋 

  Fakespot acquired by Mozilla, humans as lightbulbs, and an AI-powered website builder.

Sponsor Bot Eat Brain | New here? Subscribe!

APPETIZER

OpenLLaMA: An Open Reproduction of LLaMA

A new player has entered the stage in the ongoing saga of open-source ChatGPT competitors.

Enter stage-left: OpenLLaMA, “a permissively licensed open source reproduction of Meta AI's LLaMA large language model”.

OpenLLaMA is trained on the RedPajama dataset which we wrote about a couple of weeks ago.

So, is it just a RedPajamaDolly clone? So what?

This model is trained on the same data as RedPajamaDolly (yea, we know the names are getting ridiculous), but it outperforms it on several benchmarks.

Your takeaway: The proliferation of open-source LLM’s (here’s yet another) is making AI cheaper and easier to access for everyone and preventing big-tech monopolies.

BUZZWORD OF THE DAY

FROM OUR PARTNERS

Your own fully customizable ChatGPT

Make your own fully customizable ChatGPT, that knows only your content, then deploy it everywhere: your website, Slack, Zapier, or just a share URL

My AskAI is the easiest (entirely no-code and we have a 78-year-old user…) and quickest (you can go from sign-up to ask <30 seconds if you’re fast) way to start getting high-quality, accurate answers on your own content (website, PDFs, Word docs or CSVs).

Make it your own before sharing it with friends, users, and colleagues with custom CTAs, colors (everyone loves a gradient), answer style, and referencing.

Put an end to Ryan from operations, Ahmed the new joiner, or your favorite customer Gertrude asking that question you have answered a billion times already.

MAIN COURSE

AI image generation with TEXT is here 💬

Stability.AI, the company behind the famous stable diffusion AI model, just announced the research release of “Deep Floyd”, an AI image generator that can do... TEXT!

Here’s a taste of what Deep Floyd can do:

But that’s not all…

Floyd is not just about text, Floyd is actually multiple specialized neural models chained together giving it the ability to generate different texts, styles, and textures, understand spatial relations, and fuse concepts together.

How to get access today:

DeepFloyd is not yet available in DreemStudio. For now, a live-hosted demo of DeepFloyd IF is available on Hugging Face here.

Researchers, here are the GitHub repo and the model weights.

Why it matters:

Text generation has been a major weakness of AI-generated tools. You could generate eye-catching images, but still needed a graphic designer if you wanted your images to say anything. Until now.

Expect to see an explosion of AI-generated graphic design, logo-generators, and ads as creators adopt Floyd into their workflows.

MEMES FOR DESSERT

YOUR DAILY MUNCH

Think Pieces

Thomas L. Friedman: We Are Opening the Lids on Two Giant Pandora’s Boxes. [non-paywall link here]

David Marchese: How Do We Ensure an A.I. Future That Allows For Human Thriving? [non-paywall link here]

Startup News

Cancer Detection Startup Closes Nearly $59M This Week

Alleywatch: The 11 Largest NYC Tech Startup Funding Round of 2023 So Far

Fakespot, a startup using AI to detect fake reviews and scams, was acquired by Mozilla

Research

Humans as Light Bulbs: 3D Human Reconstruction from Thermal Reflection

Tools

Pineapple - simple ai-powered website builder.

MagicBlog - high-quality long-form blog content in just one click.

Mentioned - influencer outreach on autopilot.

AskJack - an AI assistant for all your company docs and tribal knowledge.

Video

TWEET OF THE DAY

Big tech is making big shifts to their workforce in anticipation of AI:

AI ART-SHOW

Until next time 🤖😋🧠