OpenAI’s API is incredibly adept at picking up patterns in text. From determining the format to the context, it only takes a little bit of input to get amazing results. In another post I showed how the API was able to complete lists of fictional characters, determining that it’s a list, a list of DC characters and even the gender. I also demonstrated how the API can add to scripted dialogue between characters, quickly grasping the personalities and the points of view of each one.
I thought it would be fun to show some other examples of what the API can do, from both my experiments with it and the examples OpenAI provided. This is just a small fraction of what’s possible. The API was never trained to do any of the previous examples or the ones I’m about to show. That means that there’s an entire universe of potential out there if you can figure out how to ask the right question. In this post I thought I’d demonstrate its ability to summarize content in a way that was never possible before.
Simplifying information
The API gets the context of text. It’s capable of taking a large passage and summing it up into a smaller one, and even using easier-to-understand language. This isn’t a typical key-phrase summarization where it looks for repeated phrases. This is honest to goodness getting what the text is saying and then breaking it down to me like I’m five summarization.
I can’t overstate how impactful this could be in education, research, training, self-learning and any other aspect of human life where we want to learn new things. Let me show you from an example in the OpenAI API Playground.
If we start with this phrase:
My second grader asked me what this passage means:
"""
And insert below it the first five paragraphs from Wikipedia about quantum computing (about 600 words of text.)
My second grader asked me what this passage means:
"""
Quantum computing is the use of quantum-mechanical phenomena such as superposition and entanglement to perform computation…
(I’ve truncated the rest)
"""
I rephrased it for him, in plain language a second grader can understand:
“””
The API will respond with an answer like this:
"Some people think that if you had one hundred people working on one problem for one hundred years you could not figure out how to solve it with a regular computer but if you had one hundred people working on one problem for one hundred years using a special kind of computer called a Quantum Computer you could figure out how to solve it."
Remember, this is explaining a quantum computer to a second grader.
What if we change the initial request to this?
My eighth grader asked me what this passage means:
"""
This is the response:
“When we say quantum computing, we mean computing based on quantum physics. With classical computing we perform calculations one bit at a time. With quantum computing we try to take advantage of a superposition between 1 and 0. This is what we mean by qubits. This allows us to make connections between pairs of bits (qubits) using quantum logic gates. Quantum computing allows us to perform computations faster than classical computing.”
This explanation is more sophisticated than the one provided to the second-grader, but easier to understand than the original Wikipedia explanation. The API can not only summarize, it can summarize appropriately for a particular audience.
Applications
This has tremendous potential from education to research. One of the ways to tackle a difficult subject is by reading introductory information. Unfortunately, that’s not always possible. With OpenAI’s API, anyone can take their content and make it accessible to a wider audience.
You could use the API to take difficult concepts and make it easier for people to work their way up from a simple to a more complex understanding of a topic. You can also create this content at the same time by asking the API to provide varying degrees of sophistication in the explanation.
In my AI Channels app I’ve been using this to break down news stories and as a general explainer that makes it easier to understand just about any topic, but that’s only scratching the surface of what’s possible.
Some future applications could include:
Education
Electronic textbooks that adjust to the reader. You could start by taking a quiz and the textbook would adjust the content based on what you know and then adapt as you understand concepts and bring you up to speed. I think this alone could revolutionize education. Too many people are held back from their potential, not because they’re not smart enough, but because they just haven’t had the right instruction.
Research
There’s an incredible amount of useful information locked up in research papers that are only comprehensible to people within that field. That means that an implant surgeon might never know there’s a new polymer that has medical applications or a biologist unaware that data scientists have a new technique that could radically change the way they conduct research. If research papers were easier to search and read across a variety of disciplines, there’s an infinite amount of potential for new discoveries.
Smart closed-captioning
The API could be used to power YouTube and other video content by providing additional context. If someone mentions the Maldives or the Crimean War, the API could provide added context informing the viewer that the Maldives is a place and who the belligerents were in the Crimean War.
Wearables
Extending the smart closed-captioning concept, you could use this technology in a kind of wearable like an AirPod that will whisper something to you if you need it explained in greater detail. This could either be done on request or using an algorithm that has an understanding of what you might want more information about.
Health
Reading instructions from doctors or medication can be extremely difficult. We’re often too afraid to ask for more information because we don’t want to look dumb. But if instructions could adapt to our understanding level or we had easy-to-use tools like a camera app that could break things down, we could be better informed.
Closing
These are just some of the applications that are made possible by the OpenAI API. I’m excited to see what other people come up with.
One more thing…
For fun I had the API summarize this post for a second grader. This is the result:
“This means that computers can read long passages like books or news articles and then summarize them into shorter passages with less complicated words so that they are easier for people who don't know much about computers or programming languages or math or whatever subject matter they're reading about."
I couldn’t agree more.
