Thread by @AliAbdaal

24 hours ago, I tried an experiment - I tweeted a thread with 15 productivity hacks. It’s become one of my highest performing tweets of all time, with over 1 million impressions and 23,648 engagements.

The truth is - the entire thread was written by an AI.

Let me explain… 1/

https://twitter.com/AliAbdaal/status/1582835494475698176

Two days ago, I saw a tweet about a new AI-enabled writing app called Lex, made by @nbashaw and @every. I asked them if I could try it out…

https://twitter.com/AliAbdaal/status/1582500170767687680?s=20&t=0OYHI7NC9OTKvoDCTQymhA

2/

The app lets you use GPT-3 by @OpenAI to beat writers block. You write as you usually would, but if you get stuck, you type +++, and GPT-3 generates the next few lines. I decided to try it for some productivity advice

https://twitter.com/AliAbdaal/status/1582830488552542208?s=20&t=0OYHI7NC9OTKvoDCTQymhA

3/

That got me thinking… what would it be like if I gave it the prompt of a Twitter thinkboi tweet - something like this 4/

I thought I’d get it started with a tip I actually use - the Daily Highlight, with a brief explanation.

I also gave it the start of a second tip “2 - Have an Hourglass on your Desk” and wondered how it would complete that tip.

This was all I wrote. Then the AI took over… 5/

The first thing it did was complete my “Have an Hourglass on your Desk” tip. Highlighted below is the text that GPT-3 generated. Not bad… 6/

At this point, I let the AI take over completely. These were the next three tips that GPT-3 generated, with zero human input. Genuinely, they’re all are pretty reasonable pieces of advice. Look at the likes on each of them as well 🤯 7/

And on and on it continued. Again, zero human input, I let the AI generate everything.

The tips are getting more generic now, but still sounding pretty good, and still getting decent engagement… 8/

Overall, the thread generated over a million impressions, 24k engagements, and got me 125 followers. It took me 2 minutes to think of the initial prompt, and it took GPT-3 no time at all to generate some very reasonable productivity advice that people loved… 9/

Firstly, this does make me a bit concerned. As these AI tools become more and more powerful, will we even be able to tell when writing, or art, or even videos, are actually human-generated or just AI-generated? 10/

What does this mean for the Creator Economy? What does it mean for the Twitter threadbois like me who tweet listicles of productivity tips in the hope of growing our audience, if an AI can do a “better" job than our own brains? 11/

Will the bar for what’s “good content” rise to the point where people stop engaging with generic stuff that could’ve been written by an AI? But also, the tips that it generated are genuinely quite good. Does the fact that an AI generated them make them any less useful? 12/

It’s interesting stuff to think about. I was a bit scared initially, but this experiment makes me even more bullish on the idea of personality and personal brands, where an AI can’t (yet) replicate the vibe we feel with other humans (especially through video) 13/

Somewhat reassuringly, the first tip which was “legit" (The Daily Highlight) that I wrote got the most likes compared to the 14 that the AI generated. But it was also the first tip in the thread so… 🤷‍♀️ 14/

The point of this experiment was to test whether an AI could generate the sort of content that works really well on Twitter, and whether anyone would be able to tell that it wasn’t really written by a human. I think the answer’s pretty clear 15/

Sorry if anyone saw this tweet thread and now feels duped. I hope you can see that this was an experiment for science. But also… AI tools are getting stupidly powerful. And if this is the stuff that works on Twitter, imagine how powerful it would be on LinkedIn 😉 16/

I promise that future tweets won’t be AI-generated so pls don’t unfollow / cancel me lol. But also - new tools come out all the time, that change the way we create and consume stuff. Is AI-assisted writing a Bad Thing? Or just a Tool Like Any Other? 17/

To learn more about the risks of AI and other long-term risks to humanity, check out moral philosopher @willmacaskill’s excellent book What We Owe the Future. Or alternatively, check out my brief summary of the book on my YouTube channel - 😀

And if you enjoyed this tweet, please consider liking / retweeting the first tweet in the thread - that’ll help more people see it, for whatever that’s worth lol

https://twitter.com/AliAbdaal/status/1583203039967850497?s=20&t=rpxdqQs0uTVMfTyS6fsHJA

Have a great day xx