What's up? You're reading Inclined AI. You can tell a lot about a person by whether they’re more excited for Ted Lasso or Succession coming back.

Here's what else to watch:

• A mayor in Australia might sue OpenAI

• Meta unleashed a big computer vision project

• Discover why AI does not know what to do with hands

• Snap adds more guardrails for their chat AI

Be a sponsor of our newsletter

THIS GUY IS NOT HAPPY WITH AI

ChatGPT pissed off the Mayor of Hepburn Shire (not a place in Middle Earth, I checked), and he’s threatening legal action against OpenAI.

Brian Hood’s supporters told him that ChatGPT claims Australian officials arrested him on bribery charges, but that’s a lie. 

It turns out Hood was working at Note Printing Australia during a massive financial scandal in the early 2000s but was the person who notified the authorities.

He was never arrested, according to his legal team.

Hood’s lawyer sent a letter to OpenAI on March 21 demanding they fix the error or Hood will sue them for defamation. It’d mark the first lawsuit of this kind caused by artificial intelligence.

But Hood’s argument is weak.

You can fact-check this on ChatGPT. When we tried replicating the complaint, GPT-4 and GPT-3.5 did not know who Hood was. 

Also, he became mayor last November, so we shouldn’t expect any AI model to know that.

We even asked ChatGPT about the bribery scandal in Australia, and it did not name Hood as a notable figure. So where is his team getting their information?

Also, Hood is now considered a public figure. In the United States, he must prove actual malice to succeed in a defamation lawsuit.

How do you prove a large language model acted in malice?

Our take is that the lawsuit has weak grounds to carry on, and the evidence is thin. The whole story feels more like PR fodder than a real threat.

A SPONSOR TO HELP YOU GROW AND RELAX

Sick of hustle culture?

The Boonly is a gift every Sunday for anyone craving a spark of wholesome curiosity.

Reading it gives me the permission to rest, relax and enjoy my Sundays.

They always deliver:
✅ Curiosity-sparking content
✅ Witty insights that nourish your mind
✅ Inspiring ideas to promote your personal growth
✅ Refreshing perspectives on well-known concepts

You can make all those frustrating bits of life easier with some help from The Boonly. They manage to mix winding down from a long week with wise, professional content.

I’ve added it to my Sunday digest and you can too by subscribing today.

You’re just one click away.

META IS STILL DREAMING ABOUT AR/VR TOOLS…

…BUT, the new Segment Anything Model they released is also a massive boon for AI development.

Here’s the skinny:

1/ The project allows for better, more effective image tagging and segmenting.

It is no small task to get a model to separate the different objects it’s envisioning. However, the new concepts this team of researchers introduced is foundational to getting computers to see better.

2/ Better tagging = better learning.

Multimodal Large Language Models are one track toward superintelligence; however, if they cannot see things correctly, they cannot learn.

You know the difference between an orange and an apple because of all the senses you use daily. Better segmentation heightens the visual acuity for future models.

3/ It’s the most extensive data set of its kind, AND it’s public.

The dataset is free for non-commercial use, which will give many academic labs a chance to use this to build groundbreaking visual AI models.

Foundational models are a big deal; this one being free for commercial use will matter in the long run.

Okay, that’s the scoop, but what’s the big “why.”

Meta likely did this for AR/VR development. In the simplest terms: if you want to grasp an object in virtual reality, the computer must understand the thing.

There are also augmented reality use cases. Picture your own set of glasses with Terminator vision. This project is a leap closer to that future.

Reading the GitHub repository, research paper, or blog post will probably confuse you. So instead, try out the demo and get hands-on evidence for why this is a significant moment in AI.

Quick Nuggets

🌲 Stanford’s 386-page report on AI is a huge read, so check out TC’s takeaways 

🎞️ AI videos are freaky right now, but where are they headed in the future?

📝 ChatGPT replaced his job for a bit. This journalist details the results

😱 This ChatGPT-powered Furby is all of my nightmares come true

🇮🇳 India is not going to abide by AI regulation and the problems it imposes

🔥 Hot new job skill: here’s a hint, it rhymes with BatHeTee

🔬 Researchers examine why plagiarism tools struggle to identify ChatGPT

👻 My AI on Snapchat is getting more safeguards

🇨🇳 U.S. VCs are financially backing China’s new OpenAI competitor

🩺 AI might help doctors climb out of the massive paperwork hole they’re stuck in

🙊 Bard lies way too easily for some people’s comfort levels

🎥 LVDM is next-level for long video generation

🤌 Hands: you have them and AI struggles to understand them. Here’s why

🐨 Koala is a new dialogue model used for academic research

✈️ Expedia adds GPT-powered search into their app

🦿 Humanoid robotics is not catching up to AI chatbots right now

🐕 A Hong Kong watchdog is monitoring privacy concerns in ChatGPT flagged by Italy

🎨 The online image market is getting turned upside down by generative AI

🔥 Fresh Products

• BanterAI - speak w/ celebrities (link)

• Nitro - wants to amplify the personal use of AI (link)

• June - shows you the right product metrics (link)

• LLMParser - an early playground to help anyone extract text from structured data (link)

• Invicta AI - a community of uniquely built models (link)

• Chappy AI - a nice chatbot for fun facts (link)

• Plask - use models to help AI generate images (link)

• SiteGPT - add ChatGPT to every website (link)

• Greeting Card Generator - so we can start doing that more often (link)

• oragic - an AI-enabled diary to share your feelings (link)

• Nureply - automate your email campaigns for cold outreach (link)

• EmulateMe - make your first AI avatar (link)

• OurBabyAI - have a diffusion model generate photos of your fake baby (link)

• Thumbly - optimize your YouTube thumbnails for more clicks (link)

Good Content, Bull Market

Why does this feel like something the Wolf of Wall Street would’ve done in his heyday?

_{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:

• Here Is The Work AI Will Replace Soon

• Get to Know ChatGPT Through Interactive Games

• The Difficulties in AI Alignment

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.

Claim the offer below to support our writing 🙂