What does AI think about love?
This is an article that’s enough to fill any avid reader with dread… the thought that at some point in the future, AI will replace our favourite author and be able to produce material worth reading that hasn’t had any human interaction or input at all. Is it a scary thought that future literary classics may indeed be penned by a machine?!
When you read the article though, it’s quickly apparent that it’s quite difficult to make head or tail of much of the content within it – which we’re secretly pleased about! Only after re-reading it 3 or 4 times is it clearer that the article represents the end result of 3 attempts by different versions of GPT-3 to write a love story. Safe to say, we’re not convinced that ending a paragraph with
He called the next day. We went out for lunch. We went out for dinner. We went out for drinks. We went out for dinner again. We went out for drinks again. We went out for dinner and drinks again. We went out for dinner and drinks and dinner and drinks and dinner and drinks and dinner and drinks and dinner and drinks and dinner and drinks and dinner
will the next Booker prize make!
However, we then stumbled on this article, also written using GPT-3. This one makes much more sense and is passable to read… so should we be worried? After being fed a few simple instructions and a short introduction, GPT-3 produced eight different outputs, or essays. According to the article, each was unique, interesting and advanced a different argument.
What is GPT-3?
Generative Pre-trained Transformer 3 (GPT-3) is an autoregressive language model that uses deep learning to produce human-like text. First released in June 2020, it’s recognised as being better at creating content that has a language structure – human or machine language – than anything that has come before it.
How does it work?
Give GPT-3 a prompt and it will analyse the language and then provide the most probable answer. To do this, all of the text existing on the Internet around the prompt is processed by the text predictor, calculating the most statistically expected output. The end result in many instances is blogs, poems, stories – produced almost instantly.
GPT-3 is revolutionary and heralds one of the biggest advances in AI yet.
It’s not the end of the story though…
While GPT-3 is indeed a huge leap in the ability of AI, it’s not quite there yet (as the first example of the love story clearly shows!) If a prompt is not carefully designed, the results given by GPT-3 will be of inferior quality so its use is still limited.
It may not be perfect, so for now at least, there’s still the need to cast a human eye over the work, even just for quality control services.
As we mentioned in our previous blog, this can be where FSTL steps in to help with AI! We provide a service that can save you time in “sense checking” work produced using AI. Whether that’s correcting reams of content produced without using paragraphs or correct grammar, or checking it for accuracy including references and names, FSTL can take care of it all for you.
Contact alex@fionashipley.com to find out how!