GPT-3: Is this the beginning of Skynet?

Adriano Marques August 5, 2020

In this week’s Exponential Chats, some of the team members responsible for Amalgam’s development will have a chat about Open AI’s recently announced language model GPT-3. At this point, we have all seen several demonstrations of its rather impressive capabilities, and some were left wondering what this new development means for our future.

Before we go any further, let’s briefly cover what is GPT-3 and why it has raised eyebrows all over the world for the past few weeks.

GPT-3 is the newest Artificial Intelligence Model for Language transformations created by OpenAI and announced back in May. This model is not publicly available yet, but some researchers have gained early access to it and have published some videos of rather interesting uses of the new technology.

The first example to warm-up was created by Sushant Kumar. He primed GPT-3 with quotes and asked it to come up with more quotes. He built a simple website that shows one new quote everytime that the page is reloaded. Here are a few examples: (show some of the examples live) –

If you refresh it often enough you start to notice that these are computer generated quotes, but the quality is certainly unprecedented given that this is a just a computer using the english language to generate quotes that more often than not carry a meaning to us humans.

Now, I’m going to stop refreshing it before it starts spilling out racist quotes.

Let’s go to the next example: Sharif Shameem wrote an app that takes a description of a simple app and generates a fully functional react implementation. Here is a quick video demonstrating its capabilities:

It is very impressive that the model managed to understand all the unspoken requirements that entails “rolling a dice” which is the number of sides a dice has and the fact that when you roll a dice any of the sides can come up with a number at random. This example got a lot of frontend engineers thinking about their career choices and alternatives for early retirement. We’ll talk more about this later.

And finally, this is the last demo I want to show before we get into our discussion. This one got Camilla thinking hard about her future as Product Designer. Jordan Singer built a plugin for Figma that takes in a request for the design of an application in plain english and generates the wireframes automatically for you inside Figma. Here is the demo:

Quite impressive right? There are several other demos online. Later, do a search on youtube for “GPT-3 demos” and you’ll see several others. At first it seems that this is a General Artificial Intelligence because it manages to be very proficient in several different fields such as philosophy, programming, and design. But the truth is that GPT-3 is simply a Language Model that takes in a context, a request in natural language, and generates text back to the user that it thinks complies with the request within the context given.

In the first case, it was primed with quotes and asked to generate more quotes. In the second case it was primed with React code and was asked to generate code that implements a certain functionality, and in the last case it takes in figma designs in text format as context and is asked to generate a design that shows the functionality requested.

Exponential Chats is a live event conducted by our parent company, Exponential Ventures. In this event, our team members and guests have an in-depth conversation about Exponential Technologies, Entrepreneurship, and some of the world’s most challenging outstanding problems.

Here is the link to this week’s episode that will air this Thursday, August 6th at 1PM CST:


Stay In The Loop!

Subscribe to receive Artificial Intelligence content that will help you perform better.
Don't worry, we don't SPAM. Ever.