Hello beautiful readers 👋
It’s time for your injection! 🧠
How does ChatGPT actually work..? 👀
ChatGPT is now one of my closest, most-trusted friends.
So I wanted to get to know her a lil’ better.
Here’s a very quick, simplified summary from Steve Wolfram (aka Steve-0 or Wolfy), computer scientist, physicist, and founder of Wolfram Alpha…
[skip ahead if you’re not a n*rd]
Reasonable Continuation
That’s “all” it is.
ChatGPT finds the reasonable continuation of a sentence, given what’s already been written.
In other words, given the text so far, ChatGPT asks itself: “Statistically, what should the next word be?”
And then adds that next word. And it just keeps doing that, again and again until you’ve got an essay fit for an Oxford professor.
For example, if the text so far is
The best thing about AI is its ability to…
Chat finds the statistically most common next words based on an analysis of the entire internet…
And then picks the most probable word^
Except it doesn’t.
Chat actually introduces some element of statistical randomness.
Instead of always choosing the most probable word and turning into a textbook, Chat randomly chooses less common words every now and then to make the answer feel more human and interesting.
Things get much more advanced from here - e.g. looking at most probable pairs of words instead of just individual words, looking at most probable strings of words, various parameters for randomness, neural nets, convolutions, and so on. But the above is the gist of the algorithm.
Humans
And then Chat got tested. By humans.
It received non-stop feedback from real humans, who judged whether or not the answer was “good”.
That’s why ChatGPT cares little for the truth and frequently hallucinates facts, even going as far as to invent journals, academic papers, and entire theories.
Because humans don’t care.
Stats, facts, sources. No one cares about that stuff. We care about whether something “makes sense” - does it sound reasonable.
If yes, we assume true. If no, or if it conflicts with an existing belief, only then do we scrutinise.
So ChatGPT has grown up into the world’s best people pleaser, trained on trillions of data points to tell you exactly what you want to hear.
Something to make you think 🤔
Be your own prompt engineer
An increasing number of losers on the internet (myself included) now spend most of their life not coding, not writing, not going outside, but prompt engineering ChatGPT to do it all for them.
Here’s a weak prompt:
Now here’s a stronger prompt:
That’s all “prompt engineering” is.
But you can see the change.
Recently I’ve started “prompt engineering” humans: myself and my team.
For example, we wanted to come up with a gangster af social media strategy. None of that lame “here’s some platitudinous advice”, filmed in your bedroom.
No, some serious drive-by Mafia dopeness.
So here’s how I prompted myself and the team:
Immediately the team flipped in thinking.
Instead of “Oh, maybe when we get more followers we could bend to our knees and beg for a collab with Ali Abdaal”, we were talking about:
Flying to Africa to document a day-in-life with the Masai tribe
Creating a full-length movie about downloadable personalities
Flying a blimp across every major city in the world, showering down fake monkey money which links to our website and ridicules Treadmill Capitalism
Of course, these ideas are not as simple as just messaging Ali Abdaal, but they are significantly more interesting.
And with your full creative genius unleashed, you can then think sensibly about which of these ideas make sense, etc.
I use prompt engineering all the time with myself & former clients from my founder training days:
If you were a billionaire, what would you do? No, not what would you buy. What you would you do, day-to-day, what would your life actually look like minus random superficial things like a Ferrari full of prostitutes.
This was my game-changer for lifestyle design.
(Thanks to Dave Erasmus for the prompt.)
It made me realise that I could actually live my “billionaire lifestyle” literally right now. I didn’t need any more money to make it happen.
One idea that came out of it was hiring a fun secretary.
Someone who works for me and my friends to organise super-fun events every week. Initially I thought:
“omg yh so cool but can I really afford a full-time salary just for fun? ugh no i guess not, i guess i just w8 to get rich”
But reflecting on it I realised: I could just hire that person now on an hourly basis, split the cost with my friend group and it would work out at perhaps £5/week lol.
Here’s one more:
You’re on your deathbed, writing your memoir. What are you final few lines?
Here’s what I wrote:
And so here I am. And with my dying breaths, I think back on my life.
Billions raised from poverty. Billions raised from suffering. Trillions donated to charity.
From the darkest depths of hell to the alluring mist of justice and love.
The eternal pursuit of perfection.
What a f*cking life.
Yes yes, very pretentious, very ambitious. But gets you thinking about what really matters to you.
So, be your own prompt engineer
Hit reply with what you come up with! 👇 I read & reply to every single one 😄
Ask Me Anything 👋
What are your struggles/goals? What’s holding you back? Any questions - let me know and I’ll answer personally.
Andrew ✌️
The billionaire and death prompts got me thinking. The billionaire one made me realise I'm already pretty content with my life - what I'd do with my time wouldn't change that much, I pursue my hobbies now, I would do so then as well. The death note however got to me. There's a lot of things I want to do and I already feel like I don't have the time for it. It's good to write it down though, to remind myself, so I will.