When I was a kid, I used to subscribe to a children’s magazine called Tinkle. It still is in wide circulation in India. The last page of the magazine contained letters of appreciation for the magazine. Children would post these letters of appreciation from across the world. I would pick a random name from this and write a long form letter to one of them. More often than not, I ended up getting a response. I developed connections – a penpal . It was exciting to write to people in Singapore, Malaysia and Canada. It was liberating to read about how they went about their lives, the cities they lived in, the people they lived with, etc. In the midst of moving homes and penpals changing countries, we all lost touch. In the mid 90s, when cable television was a rarity and there was no internet, you could talk to someone only through a traditional letter. It was inexpensive. There were phone calls, but they were even more expensive.
The Internet happened and it shrank the world and everyone was 2 screens apart. The advent of the mobile culture brought with it the 160 character limit for an SMS which is where we lost it completely. We mixed up English with vernacular languages. We came up with a mish-mash of words. Some even made their way into the dictionary. Nonetheless, human communication with words took a different trajectory. I’d given up hope that humanity would ever go back to long form communication.
Lo and behold, last week I heard about this instance of Artificial Intelligence called GPT-3. In its most basic terms, GPT-3 simply generates a continuation of any text that you give it. GPT stands for Generative Pretrained Transformer. I first heard about it from a tweet by Keagan Stokoe. I dug a little deeper and saw Tinkered Thinking was playing with it. Tinkered Thinking wrote a great post about GPT-3 here, it even reached the front page of Hacker News!. It’s a clear explanation of what GPT-3 is and roughly how it works. Be sure to read up on the conversations between Tinkered Thinker and GPT-3.
In a nutshell, GPT-3 is an AI that utilizes a neural network trained on language from the internet. It has been trained with about 175 billion parameters. It appears to intuit the context of what you feed into it and the content it throws out is remarkable. Close to what a well read intellectual would say.
Many swooped down to warn the world about the evils of GPT-3. The simplest example would be one where vile media outlets use it to smear dirt on one other, etc,. But I believe in the positive potential a technology offers. With GPT-3, the breakthrough would come in the form of better writing outlets? – for professionals, for me, and for you. Every time an author is stuck with writer’s block,GPT-3 has the capability of generating possible continuations from which an author can reference for further inspiration. (Tinkered Thinking did this here) Would it help create better stories? Would it help make better authors? Or would GPT-3 realise it is the better author? Would GPT-3 become a ‘crutch’ for writers? Would authors still be motivated to write if a machine can generate a story at the push of a button? Right now, GPT-3 is here and looks like it’s not going away anytime soon.
I went one step ahead to think about what would happen if we integrate GPT-3 into conventional operating systems? The predictive text options we see beneath what we’re typing (Phone, tablet or even Gmail) is a Neanderthal version of GPT-3. But with GPT-3, this game rises to a whole new level. Will we finally come out of – “IDK”, “Wr r u?”, “U 2” – symbolisms that have come to dominate instant messaging? Will we revert back to the long form method of communicating with each other? Everyone who has interacted with me knows my disdain for these short-forms. I will respond to you without eating up alphabets; no matter how short or long the response is. I love using long form words and expressions. Some of the letters I have written have ended up being longer than posts here in my newsletter and on my blog.
If you’re wondering whether integrating a tool like GPT-3 in your personal device would lead to everyone talking and writing the same way, “Hold your horses boys! You ain’t seen nuthin’ yet.” GPT-3 interacts with you based on what you feed it and where you lead it. With Billions of people having interesting thoughts, I think there’s a low probability that you’re going to be writing something that is similar to your next door neighbour. In fact, if what we have seen is true, GPT-3 will respond in a different way, when you repeat the same question. Tinkered Thinking, and many others have already tried this.
Communicating with fellow beings is the part of GPT-3 that excites me the most. It is in human nature to want to seek connections – it doesn’t matter where the connection comes from. We want to reach out – we want to be reached. The soul yearns for connection.
With GPT-3, I have renewed hope. If we are able to leverage a simple version of this to improve communication, the world will be a better place.
I’d love to get a response to this! I’d love to respond to your response, because in the medium of exchange we create new ideas, bonds, and relationships.
Special thanks to Tinkered Thinking for reading drafts of this post and guiding me on improving portions of it.