The nerd part of me had the best weekend ever as I finally got access to participate in GPT 3 Beta. I have been mentioning about GPT 3 and its language capabilities in almost all my conversations. I build a React app to capture information and parse through the output from GPT 3.

This was also the weekend when Elon Musk was on Saturday Night Live and Mother’s Day !!

What is GPT 3 ?

In non-technical terms, our friends at Open AI showed a machine how normal humans write and communicate ideas based on the internet content. Yes same as teaching a small baby how we speak! The machine after having learned the human way can interact with other humans based on detecting patterns , identifying ideas etc. That amazing machine which is a black box to the rest is accessible via Open AI’s Beta program as APIs. They can be used to build demos just like this one or build apps.

In technical terms, Open AI trained a machine on a large amount of text in English language. The output was a language model with more than 175 billion parameters making it potentially indifferent from human-written sentences, paragraphs, blogs etc.

Calculate like a human

The first thing that comes to your mind must be calculators, but GPT # has more to offer than that. I trained GPT 3 on how we speak and think quadratic equations and voila ! If you notice closely it follows BODMAS rule automatically every time without having very limited prior experience. So beyond calculations, GPT 3 helps you understand the math.

Help me with my emails?

It’s not that I hate writing emails, but often my emails are nothing more than one-liner requests or questions, which I really hate. They are often essential due to scenarios such as business travel. So I asked GPT 3 to write an email to get approval for my upcoming travel and look what it did !

Personalization & A/B Testing

Okay lets take things to the next level! One of the key challenges that is extremely simple yet difficult to do is experimentation testing. There has been so many instances where I sat with the marketing team to develop A/B version of the content to test effectiveness. It becomes more complicated as you have more elements or components on the same page, re-writing the text into multiple versions becomes an endearing task.

Lately I am getting a lot of ads about Linked In Learning. So I trained GPT 3 about LinkedIn Learning using text from the web, Wikipedia etc. along with some examples of A/B content. Now if I have to run any A/B testing campaigns about Linked In Learning, I just have to provide the control text (version A) and GPT 3 would provide the challenger (version B)

Marketing Emails

It appears that GPT 3 knowns a lot about LinkedIn Learning based on previous iterations. So I asked GPT 3 to write an email about LinkedIn Learning to AWS professionals in Sydney. Now at this time my web app is throwing some weird errors so back to good old Linux CLI to run this and look at the content that comes out. Yes if you’re in the marketing team you cannot use this right away. But the content gives you ideas to start with…

The fun does not stop here! I am looking at creating more demos using GPT3 and updating this page. Strongly recommend that you subscribe to this post to stay updated with the latest.