Interact with GPT-3


Vol. 15 // 2020

Natural language processing concept, copyright Visual Generation via Creative Market

We are living in such exciting times! There are so many new scientific breakthroughs that are happening all around us all over the world. One of the hottest fields leading the way is AI. Scientists are writing faster and smarter algorithms; machines are learning and more accurate predictions are being made.

NLP plays a huge part in this endeavor. In the ideal scenario we would want to talk to a machine and have it understand and execute commands. All of this is made possible through NLP breakthroughs, the latest of which is GPT-3.

GPT-3 stands for Generative Pre-trained Transformer 3 and it is an auto-regressive model that uses machine learning to produce text. The model was trained on 175 billion machine learning parameters. Compare this to GPT-2’s 17 billion parameters.

What makes GPT-3 so amazing is the quality of the text it generates. It is so close to what a human would create that telling the difference can be very difficult.

All of this sounds wonderful but the question that comes first to my mind, and I am sure yours too, is what can it do right now? So for this issue of Cognitive Times we decided to show you some things that GPT-3 can do and the fun you can have with it. Some of these applications are mind-blowing. If language models continue to make such progress, we can be sure to witness even more exciting developments in future .

1. Sharif Shameem is using GPT-3 to build text to web applications in React. That is amazing as it means we are making strides towards text and speech based programming.

2. Matt Schumer is using GPT-3 to generate code for a Keras machine learning model. He just inputs a couple of lines of text to generate an impressive result.

3. Then there are some exciting creative projects!

4. For those of us trying to use LaTeX to generate the math for our papers, Shreya Shankar has created a text to LaTeX converter using GPT-3. I know this is one application many mathematicians and scientists will be grateful for.

5. Jordan Singer has created ‘Designer’, a Figma plugin that will convert text to UI in seconds.

All of these applications are already very ‘futuristic’ but they are just simple examples and a hint of what is to come in the future. The age of AI has much more in store for us over the next few years! What do you predict will be possible by 2025?

Related Articles

Turing Talk