GPT-3 and the Rise of Human-centric Adaptive Computing
Introduction
💡 This article is part of a series: Intro, Part 1, Part 2, Part 3
The release of GPT-3 to the world
In July 2020, OpenAI released GPT-3, the most powerful language prediction model ever.
GPT-3 stands for Generative Pre-trained Transformer 3 and its model architecture is a transformer-based neural network. Practically, GPT-3 provides a general-purpose "text-in, text-out" interface, that takes any prompt in human language and returns a text completion, matching the pattern provided during the training phase.
Immediately after release, the first demos of what GPT-3 is capable went viral on Twitter, "shocking" everyone: just by providing few "English-in, English-out" training examples and little-to-none fine tuning of the model, GPT-3 started to write press releases or technical manuals, summarize articles, let you have conversations with historical figures, generate short stories or songs written based on the style of a specified artist, write entire emails based on few bullet points given.
Like anyone else, I was of course impressed by these "English-in, English-out" examples, but one less obvious use of GPT-3 model really, truly, blew my mind: the "Human Language-in, Machine Language-out", where computers were “controlled” by user’s natural language.
The interaction between humans and machines has been a passion, almost an obsession, for me, as I believe it's still an unsolved, yet fundamental, challenge for technology. While building successful consumer and B2B apps used by millions, I first adopted mobile interfaces, attracted by their remarkable ease-of-use "by design and constraint", and then experimented with conversational and voice interfaces as the next step in my quest for delivering the ultimate user experience.
Based on those experiences and research, I believe that GPT-3 can be the missing piece to revolutionize the way humans interact with machines, finally democratizing technology and making it accessible and usable by everyone on the planet.
As such, this article focuses on GPT-3 not as a writer, poet, or a famous figure, but as the enabler for a new generation of Human-centric Adaptive Computing.
The article is structured in 3 parts:
Part 1 describes how the "Human-in, Machine-out" GPT-3 Interface can finally change the game in human-to-machine interaction
Part 2 introduces the concept of Adaptive Software enabled by GPT-3, by analyzing the architecture of a GPT-3 application
Part 3 explains the fundamental importance of coupling GPT-3 Adaptive Software architecture with a Human-centric design
Note: throughout the article I generically use the "product designer" or "designer" terminology to describe the person in charge of designing a product; by design, I mainly mean how a product works, not necessarily how it looks.
💡 Part 1 describes how the "Human-in, Machine-out" GPT-3 Interface can finally change the game in human-to-machine interaction