What's up? You're reading Inclined AI. We’ve been so focused on AI news that we might’ve missed some things. For instance, did you know a new Indiana Jones came out??? Something to do with crystal skulls…

Here's what belongs in a museum today:

• Google is showing off its robot training

• The content outcry is getting louder

• Prompt engineering is all about not asking silly questions

• Semafor dives into everything ChatGPT

Be a sponsor of our newsletter.

GOOGLE REMINDS US WHY THEY’RE AI LEADERS

ChatGPT is cool, but it can’t take out the trash, walk a dog, or physically pick out an object from a lineup. We tend to get excited about all the new Gen AI tools, but that’s only half the puzzle.

That’s why you need to get hyped for Google DeepMind’s new robot friend, RT-2. It’s a robot they’ve trained by integrating principles learned from LLMs.

RT-2 is a Swiftie, and that’s saying something.

Okay, move passed that bad joke. My point is that this robot is more competent than earlier PaLM tests in ways that count.

For example, it moved a Coke can toward a picture of Taylor Swift with just a prompt. And when researchers asked it to select a beverage for an exhausted person, it grabbed a Red Bull.

It sounds like Google trained a robot to understand advertising, but there’s more to it.

RT-2 succeeds without the need for complex instructions. That’s new and promising.

See, the world is confusing.

You make thousands of micro-decisions every day. What are the odds a robot can muster that same kind of output without a drop in performance? Even your Apple remote has trouble understanding you.

Getting a robot to walk across the room and grab a broom must be simple. Otherwise, we’ll never get AI across the AGI chasm. 

We can’t automate jobs and reach singularity if our AI friends fail to interact with the physical world. Integrating LLM research into robotics is as promising a step forward as we will get right now.

So, yeah, it’s silly to celebrate a robot being able to identify Taylor Swift, but we’ll take it (even if it failed some other tests, but who’s counting).

Life-Changing Concepts → gives you an unfair advantage in business using mental models

Master ChatGPT → check out the ChatGPTricks e-book that takes you from zero to hero

Inclined AI Premium Newsletter → get a free 7-day trial on us when you subscribe today

WHO CREATES THE CREATORS?

Let’s put a spin on the content compensation controversy. We’ve exhausted the legal debates and know that more lawsuits happen every week.

The main concern is that governments will remove models that haven’t compensated creators for their work from the internet. A decision like that would set AI research back years, sure.

However, that’s all conjecture from pessimists.

The fair use doctrine will protect AI companies. They aren’t selling copies of John Grisham novels. All they did was scrape websites that illegally posted those books.

The output from ChatGPT is a collage of things. No one sues Elvis impersonators in Vegas. So don’t panic and hit that red button that says, “Ditch the hype train.”

Everything is going to be alright.

Do you hear me? Do you feel me? Because when push comes to shove, this is closer to the lawsuit against EA for their NCAA video games or music sampling.

In a worst-case scenario, AI companies will need to pay for licenses in the future, but courts aren’t about to scrub the internet of every AI model out there.

We should worry more about disgruntled creators.

Data fuels Large Language Models and keeps them learning. Every research lab knows one truth: more data = better performance. 

But a myriad of mishaps is reshaping how companies acquire data. AI threatens to replace the people who fed it the info, which, if left unchecked, becomes a snake eating its own tail.

Without new, novel content to consume, these models will not learn. 

ChatGPT is writing SEO, empty-calorie blogs for companies. Social media is a bunch of prompt templates and scheduled, rehashed threads.

No one wants to create things if it means losing that ability in the future.

That means we must appease the writers, artists, and creatives who make things worth feeding a model.

If that means licensing material and compensating the under-compensated, go for it.

You don’t need to flip out about these lawsuits, but you should worry about the subtext. If we can’t get enough new content for the next generation of models, then an AI winter will happen.

No one wants that.

Quick Nuggets

🤡 AI tries jokes for this group of Edinburgh performers

🤔 Here’s a take that suggests that AI has no reason to harm us

🧑‍💻 Prompt engineering tips: learn to stop asking bots silly questions

💼 AI in work is helping companies scout, hire, and train new workers

📌 Code assistants are changing the way developers experience their job

❓ The big mystery in this AI boom is what parts of the black box are actually working

🪐 An MLM was trained by researchers, but the novelty is it happened in outer space

⭐️ Semafor dives deep into the hidden secrets behind ChatGPT

🔥 Fresh Products

• Likewise Learn - build social media engagement (link)

• Ever Efficient - AI solutions for optimal efficiency & growth (link)

• Synthical - discover, learn & share research (link)

• Lolo - AI chatbot to track your nutrition & calories (link)

• Noon - talent sourcing on autopilot (link)

• Journal BuddyAI - assistant to make journaling impactful (link)

• WordSeek - rank higher in search by including missing website keywords (link)

• Dezee - ultimate AI creative partner (link)

Good Content, Wes Trek

Give me this movie tomorrow, please. If any director can pull off a new spin on Star Trek, it’s Wes Anderson. They could go so well together.

_{- That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.}