OpenAI GPT-3: The new AI that will blow your mind… might also be a little overrated
A few days ago, OpenAI, an artificial intelligence research laboratory, launched the closed beta of GPT-3, a.k.a. the most powerful language model ever built, and the Internet went crazy.
In plain English, GPT-3 is a set of machine-learning algorithms where you can interact in natural language and, in certain controlled environments, produce exceptional results.
Maybe because it’s one of the times that Artificial Intelligence is available for end-users in such a simple, yet powerful and interactive, way, or maybe because the outputs are stunning, but the hype generated around GPT-3 is very reminiscent of the boom that Bitcoin had some years ago (which is both good and bad).
Nowadays, AI is part of almost every tool/service/app we use daily. Think of Instagram’s filters as an example. There’s a lot of AI there. But, with GPT-3, it’s unique because the AI makes you feel as though there’s a really smart human on the other side following your directions or answering your questions.
So, how does it work? GPT-3 is a model that claims to have 175 billion parameters, and it has practically processed all the text available on the Internet. That’s why it can solve a variety of tasks like creating computer programs, composing poetry, designing interfaces, or even writing a full article.
GPT-3 in Action
Some of the examples generated by devs that got access to this service are astonishing. This shows that a well-trained model performs incredibly accurately. Let’s look at a few.
One initiative is trying to build a software around GPT-3 that uses an individual’s personal writing style to answer emails. Each person can tell the software the key points they want to get across, and it will produce the output.
Some people gave the algorithm parts of essays or memos, and GPT-3 wrote the next paragraphs. The results were mind-blowing.
Next, there is a plugin that, based on the user’s input, generates the user interface for a mobile app.
Last, but not least, this curious example shows how a user can virtually ask anyone questions. The software learns the writing style and what that person is supposed to know and answers based on that.
In the words of one of its creators, GPT-3 “still has serious weaknesses and sometimes makes very silly mistakes.” It performs really well under certain conditions, but it also fails when faced with some simple tricks or questions.
The hype around GPT-3 also reminds us of the hype around Bitcoin a few years ago from nearly the same group of people. This is a BIG red flag. Luckily, in just a matter of days since its launch, we see more realistic examples compared to Bitcoin.
As you can see, GPT-3 is far from perfect, and it’s not going to replace the human workforce anytime soon. But, it does show us that it’s just a matter of time before we see a more refined version that potentially could. As AI continues to evolve, it will have more potential to replace those jobs where creativity does not play a major role.
We are not saying anything new when we declare that AI is changing the world right now, but this will continue to increase in the near future. Being prepared for those changes and ready to adapt our businesses are critical strategies to put into practice today. And, part of that is being aware of initiatives like GPT-3, watch it closely, track how it evolves, and determine how to take advantage of it.
More articles of interest: