GTP-3, the AI that texts back.

A neural network capable of coding, translating, summarizing, and generating text poses urgent questions on the definition and role of human language, creative skills, and inherent biases.

At the end of 2020 OpenAI released GTP-3, the 3rd generation of a neural network capable of coding, translating, summarizing, and generating text. This machine learning model was trained using so far the biggest set of data containing various blogs, Wikipedia entries, and digital books. As a result, it is the closest we’ve reached to creating machine learning which understands human language. 

GTP-3’s capacity to use natural language can be seen as a stepping stone to AI understanding other human skills, but it shows that communication with AI is not as far-fetched as previously thought. The importance of being able to ask “Why?” is increasing as we turn over our responsibilities in fields such as healthcare and transportation to AI. GTP-3 and further developments with it would change how we interact and trust technology, and what we consider exclusively human abilities.

Although far from flawless and still quiet niche, GTP-3 is helping power a wide range of tools and services - from automatic news article summaries to new computer languages. However, as artists and designers explore the possibilities of this new machine learning model, we encounter its limitations, inherent prejudice, and hate. After all, the model is trained on uncurated content from the Internet, resulting in the generation of new texts with old racial, gender, and cultural biases.

What does that mean to design and development? GTP-3 has already proven itself useful in the world of product design and coding. Feeding the model with text descriptions alongside matching Figma code resulted in a specialized design tool, which can code based on plain English descriptions. With a couple of tweaks, the code is complete. However, this shouldn’t be seen as a threat to careers, only as a way to remove tedious work and allow developers to focus on the creative and innovative part of their jobs. Although GTP-3 is not yet open for public use or purchase, it is a development worth following for individuals from all fields.