You've probably seen GPT-3 pop up in your social feed a lot recently. Let's dive into the basics around what it is and why you shouldn't be scared of it... 👇
What is GPT-3?
The Generative Pre-training Transformer (version 3) is an extremely sophisticated language predictor created by @OpenAI. It can be accessed via an API. So a human gives the model some text and GPT-3 provides back its best guess around what text should follow.
The Generative Pre-training Transformer (version 3) is an extremely sophisticated language predictor created by @OpenAI. It can be accessed via an API. So a human gives the model some text and GPT-3 provides back its best guess around what text should follow.
How does it work?
As an AI language model, it has analyzed almost all the text on the internet. It then takes this data and makes predictions by identifying patterns. The output is based on the surrounding text and it has zero understanding of words or the output
As an AI language model, it has analyzed almost all the text on the internet. It then takes this data and makes predictions by identifying patterns. The output is based on the surrounding text and it has zero understanding of words or the output
Why are people going crazy over it?
For a couple of reasons:
1. It's the most powerful AI language model that has been released
2. It's currently in closed-access and limited releases create hype and FOMO
3. People are sharing some very cool things they've created with it:
For a couple of reasons:
1. It's the most powerful AI language model that has been released
2. It's currently in closed-access and limited releases create hype and FOMO
3. People are sharing some very cool things they've created with it:
@paraschopra created a fully functioning search engine:
@jsngr built a Figma plugin that designs for you:
@zebulgar showed how it could generate a proper business doc:
Should I be scared? Does this mean we never need coders, designers, or business people to work ever again?
No. With zero comprehension, GPT-3 does not fact-check, tends to lose coherence over long passages, can contradict itself at any time, and cannot create unique thoughts.
No. With zero comprehension, GPT-3 does not fact-check, tends to lose coherence over long passages, can contradict itself at any time, and cannot create unique thoughts.
Instead of thinking of GPT-3 as a competitor, think of it as tool. You use tools constantly to make your work easier (Gmail, Excel, a calculator...). Consider this the same way.
You can use GPT-3 to save time on the boring or monotonous tasks, which leaves you more time for the important and fun parts of your work.
I automate people's work tasks all the time and explain the same thing: I'm not coming for your job. I'm coming to make sure you're being utilized to your full potential.
My goal is to provide tools that make it easiest for you to spend time on tasks you love and are great at.
My goal is to provide tools that make it easiest for you to spend time on tasks you love and are great at.
Embrace the potential and be excited for what else will come
