• Not A Bot
  • Posts
  • 🏭️ Willy Wonka and the Stability AI factory

🏭️ Willy Wonka and the Stability AI factory

PLUS: Voice actors fight back against generative AI

New here? Welcome to NAB, and make sure to subscribe for more!

Happy Tuesday, fellow humans. 👋

Kylian Mbappe got offered $1.1B to play in the Saudi Arabia league for 1 year. Holy cow.

Is it too late to change professions?! Jk, I love my job (I think).

Let’s dive into it…

Have anything you’d like to share with over 40k subscribers? 

🧵 In today's edition:

  • 🏭️ Willy Wonka and the Stability AI factory

  • 💭 The Hype Around AI

  • 🍎 AppleGPT used for internal work

  • 🤑 AI Fundraising News

🤖 Top AI News

🏭️ Willy Wonka and the Stability AI factory

At the heels of Meta’s Llama 2, Stability AI and its research branch CarperAI have released FreeWilly1 and FreeWilly2 - “two powerful, new, open access Large Language Models (LLMs).”

  • FreeWilly1 is a fine-tuned version of the OG Llama’s 65B model

  • FreeWilly2 is a fine-tuned version of Llama 2’s 70B parameter model.

Both these models were trained through techniques inspired by Microsoft’s methodology, highlighted in their paper “Orca: Progressive Learning from Complex Explanation Traces of GPT-4.”

The main differentiator: Stability used a dataset with 600k data points - about 10% the size of the original dataset used in the Orca paper.

Quality over quantity.

The best part? Despite being 1/10 the size, FreeWilly’s performance was comparable to GPT 3.5 and GPT 4.

Here are some results taken from their blog post:

AGI Eval (all 0-shot)

Translation: FreeWilly 2 nearly outperforms GPT 4 in most tests, and is comparable to GPT 3.5

Overall: A big win for open-source projects. Excited to see how OpenAI responds.

💭 The Hype Around AI

Every major piece of technology that has existed to date has gone through the Gartner Hype Cycle.

Most recently, we witnessed it in the NFT/Web3 debacle of 2022.

Now, Duke University professor Sultan Meghji, who previously served as the FDIC's chief innovation officer, says AI might be next.

Meghji believes that much of the hype around AI is driven by marketing teams and not actual technologies.

While he believes AI will be transformative, he thinks the biggest changes brought about by AI will be less headline-worthy.

He expects AI to face challenges in appropriately curating data for large language models, as many of these systems are based on learning from what is openly available on the internet, which may not be of high quality.

He claims: "In AI, you have to be really careful with the data you use to train it and, at some point, you'll stop having positive returns. It'll stop getting smarter, so you need to stop and take a step back."

Stuff to think about.

Read more: Yahoo Finance

🍎 AppleGPT used for internal work

Credit: The Verge

It was reported that Apple had developed a chatbot late last year but disbanded operations over safety and data concerns.

According to Mark Gurman's newsletter Power On, Apple is now using its chatbot, AppleGPT, for internal purposes, such as:

  • prototyping future features

  • summarizing text

  • answering questions based on its trained data

What does this mean?

One possibility is that Apple might give the tool to AppleCare support staff to enhance customers' experience.

The company intends to take its time with AI as they are aware of the issues related to it and want to proceed deliberately.

Apple plans to make a significant AI-related announcement next year, and if history is any indication, we could be in for a treat 🍿.

Read more: The Verge

🤑 AI Fundraising News

Quench.ai raises $5M in Pre-Seed funding to provide an AI learning coach that helps individuals discover relevant content to support learning and upskilling, with an extensive library of indexed educational videos and the ability for companies to index their internal video resources.

Reworkd AI raises $1.25M in Pre-Seed funding to allow businesses to automate tasks using AI agents, streamlining manual procedures.

🗞️ AI Quick-Bytes

What else is going on?

  1. NY police used AI to track drivers on highways as attorney questions legality.

  2. Citi stays positive on the A.I. theme and lays out the key to finding winning stocks.

  3. Top news publishers are reportedly planning to sue AI firms

  4. SoftBank forms JV to build AI-powered warehouses with Symbiotic

  5. Microsoft brings Bing AI support to Google Chrome and adds dark mode too

  6. Sam Altman’s Worldcoin token soars on the first day of trading

  7. Marc Andreessen says AI will function as a lifelong 'partner' for kids with complete insight into everything they've done and want — and he thinks that's a good thing

  8. Voice actors fight back against generative AI

🐦️ Tweet of the day

Absolutely. Amazing.

Twitter has been on 🔥 lately.

What did you think of today's newsletter?

Login or Subscribe to participate in polls.

And that does it for today's issue.

As always, thanks for reading. Have a great day, and see you next time! ✌️

— Haroon: (definitely) Not A Robot and @haroonchoudery on Twitter

P.S.: Check us out on our socials - snacks on us if you also hit follow :)

Reply

or to participate.