✅ 1300+ experts say AI is GOOD for humanity!

+ what happens if you train AI on AI? 😂

Matt V
July 19, 2023

This is Wednesday’s Mindstream. And hopefully, we remembered to change the email subject today. Whoops.

If you thought we sent the same email out twice, we didn’t, we just forgot to tick one small box to update the subject. Doh!

So please read yesterday’s email if you didn’t open it.

but anyway…

We make AI simple and deliver it straight to your inbox, EVERY DAY!

Everything you need to know about AI today in 5 minutes. You can even read it while you watch Mamma Mia.

TODAY IN MINDSTREAM:

✅ 1300+ experts say AI is GOOD
📚 Expert prompt tips
😂 Why you shouldn’t train AI on AI-generated content
💡 AI-nspirational quote of the day
🎁 Refer Mindstream, get free sh*t
📸 AI image of the day

✅ 1300+ EXPERTS SAY AI IS GOOD

Finally, some news that counters all the AI doomerism we’ve been seeing!

Over 1300 experts have signed an open letter saying that AI is a force for good, and not a threat to humanity. The letter was organised by BCS, the Chartered Institute for IT, to counter "AI doom". Thanks, guys. Doing god’s work, truly.

The letter states: "The UK can help lead the way in setting professional and technical standards in AI roles, supported by a robust code of conduct, international collaboration and fully resourced regulation". Nice.

graphical representation of mainstream news coverage on AI.

The experts came from a range of backgrounds - business, academia, public bodies and think tanks. Showing that there is a cross-section of society that sees past all of the clickbait headlines!

Richard Carter was one of those who signed, having founded an AI cyber security business. He feels the dire warnings are unrealistic: "Frankly, this notion that AI is an existential threat to humanity is too far-fetched. We're just not in any kind of a position where that's even feasible".

📚 EXPERT PROMPT TIPS

Get better at using the AI tools that matter with our regular prompt tips.

Get Verby.

Lean on doing words to deliver clear instructions to ChatGPT. Try:
Interpret, Apply, Demonstrate, Illustrate, Elaborate, Justify and Validate, if you’re feeling clever.

😂 WHY YOU SHOULDN’T TRAIN AI ON AI-GENERATED CONTENT

AI models need to be trained on massive amounts of data. But what if that Data was AI-generated itself!? Could we have figured out how to make an AI that gets better and better forever?

As it turns out, no. A university department tested training AI models on “synthetic data” in an experiment.

Each AI was trained exclusively on its self-generated content. The results are probably not that surprising: within a few generations, severe artifacting began showing up in the outputs.

Imagine all of the small weird errors that AI image generators produce, emphasised 1000x over.

When the scientists cherry-picked the images that were artifact-free, the result was an AI that could produce a shockingly small amount of new content, with many outputs looking identical due to the limited data set.

So, don’t train your AI on AI, kids.

Read about the experiment in more depth in New Scientist (paywalled)

💡AI-NSPIRATIONAL QUOTE OF THE DAY

Don't be afraid to take risks and step outside your comfort zone. It is within the realm of the unknown that you discover your true potential.

Your daily pearl of wisdom from ChatGPT.

🎁 **REFER MINDSTREAM, GET FREE SH*T**

❤️ Love Mindstream? ❤️

Then share it and get free sh*t!

We just added some crazy referral gifts! 😮

Here’s what you’re missing out on.

1 Referral - Ultimate ChatGPT Guide - take your prompting to the next level.
3 Referrals - A custom Midjourney portrait of yourself in any style!
10 Referrals - Call with the Mindstream team! Talk about AI or your favourite sandwich, we don’t care.
25 Referrals - Your very own Mindstream mug! Drink from it while you read Mindstream!
50 Referrals - A Mindstream t-shirt! Hell yeah!