An AI Wrote This Blog Post
I’ve recently been tinkering with an AI platform called GPT-3 (available here to try for free), which predicts text based on what you write. If your phone or computer have ever suggested the next word or phrase for you in a text or email, or a web search has filled in what you’re searching for as you type, that’s exactly what GPT-3 is doing. It takes a ‘prompt’ - something you’ve written - and tries to complete it, taking a guess at what’s most likely to come next.
Unlike the tech in your pocket or home computer, GPT-3 can predict not just the next word but whole paragraphs that might come next. This is because of how it’s been ‘trained’. Text prediction is only based on how you personally communicate, and search prediction is based on how many other people search the same things as you. GPT-3, in contrast, has basically read the entire internet. It’s using every website, forum, blog post, social media account and more to learn how people speak and write in a variety of different situations.
This means it can replicate the tone for whatever you need. You can ask it ‘write a script’, or ‘write a standup comedy routine’, or ‘write the code for a webpage’, and it will understand how to present and format each of those things.
To demonstrate this, I thought I’d let GPT-3 speak for itself. I gave the AI a detailed prompt on a blog post I wanted it to write that would explain how it works. The rest of this blog is what it returned.
The below is content generated by GPT-3
It is speculated that, in the future, developers won't be writing code but writing prompts. A prompt is a short piece of text that describes what you want an AI to do. For example, a prompt could be "Write a blog post about the future of AI."
The AI would then take that prompt and generate a blog post about the future of AI. The blog post would be based on what the AI knows about the future of AI (which could be quite a lot, if the AI is given access to the right data sources).
GPT-3 is a new AI platform that is said to be the most powerful AI platform ever created. It is speculated that GPT-3 will be used to create even more powerful AIs in the future. GPT-3 is different from other AI platforms because it uses a technique called transfer learning. This means that it can learn from different tasks and then apply that knowledge to new tasks. For example, if GPT-3 is trained on a task like writing a blog post, it can then apply that knowledge to other tasks like writing code.
One potential use for GPT-3 is to generate code for new applications. The developer would provide a prompt describing the desired functionality, and the AI would generate the code. This could be used to create new applications much faster than traditional methods. There are many potential uses for GPT-3, and it is likely that we will see more and more uses for it in the future. As the AI gets better at understanding and responding to prompts, it will become more and more useful for a variety of tasks.
This blog post was written entirely using GPT-3. The prompt was:
"Write a blog post about AI and GPT-3, starting with the sentence 'it is speculated that, in the future, developers won't be writing code but writing prompts'. Use non-technical language so that it is easy for anyone to understand. Explain what prompts are and some examples of how they could be used, and some thoughts on how GPT-3 could be used in the future. Explain that this blog was written from this prompt and quote this prompt in the blog."
What do you think?
So, what does this mean for the future of AI? It is hard to say for sure, but it seems likely that AIs will become even more powerful and intelligent than they are today. They may even become able to write better code than human developers.
Only time will tell.
Photo by Alex Knight from Unsplash