The Generative Pre-trained Transformer is far from perfect, but we need to prepare for its arrival and learn to exploit its language

What happens when software like GPT are they able, in a matter of seconds, to replace the hours of research and dedication that a writer puts into his work? What will happen when AIs are able to write a text as good as that of a junior author, if not even more sophisticated and better?

And if we told you that these software already exist, but above all they work really well?

I can't deny that the idea of ​​a robot capable of doing my job initially scared me quite a bit. And what will become of my wages, what will become of my work? As it happens for midjourneyHowever, the question is never that a job is destined to disappear. It's about ability to adapt. Work for professionals like mine will never disappear, but will transform. And without learning about GPT, we can never transform in the right way and at the right time.

But let's start from the beginning. Let's talk about GPT.

The surprise of Ignazio Cassis: "My speech written by AI"

What is GPT and why is its potential impressive?

GPT, e GPT-3 as the latest model, it's a OpenAI's Generative Pre-Training, and was officially released on June 10, 2020. You have to think of it as a language model that relies on transformers, a deep learning process imagined and programmed for the creation of human-like texts. By feeding this artificial intelligence with a huge body of data, between books, articles and websites, GPT has become able to produce texts that are more and more coherent with human language and, above all, endowed with impressive accuracy and quality.

Reading the results of some tests, you too would be amazed: in front of the potential of tools like GPT, the doors that open wide are unprecedented. In addition to being able to generate text in several languages ​​and handle tasks like answering questions, GPT's goal is to reduce the amount of data needed to get fantastic results.

The easiest method to use GPT-3 is your own ChatGPT, and while you can't tweak a huge amount of parameters, this tool alone is jaw-dropping at the quality of insights it can provide to a good author. But also for the potential it can give to those who are looking for personalized texts, capable of passing plagiarism tests, without having to hire a dedicated figure. But will it be true?

It doesn't stop there. With GPT-3 you can also program, because its purpose is to deal with everything structured as a language. So why not a good programming language too? GPT-3 does this but also writes essays, summarizes long texts, translates, takes notes and encodes. Where is your degree now?

Codex: the software that transforms language into code

The contraindications of the Generative Pre-trained Transformer

There are two main risks associated with GPT-3 and, in general, to automatic text generators and other writing products. The first is what is called “unsupervised learning”, i.e. unsupervised learning.

Some answers provided byAI they are not evaluated during training, and therefore remain conceptually wrong. This originates wrong information within the texts, or errors of concept that can cause confusion or even misinformation.

And by the way, specialists warn about the amazing skill of this tool produce fake news. You want because in fact the algorithm does not think, but associates elements and is fed by the content of the networks, often overloaded with disinformation.

That said, some experts have stated that just over 50% of readers recognize a text written with GPT-3, or otherwise not edited by a human being. This is not a small danger, because a significant part of the population could believe everything that GPT-3 declares (or whoever), without asking the right questions.

The future of these AIs lies in the how they were fed. Before imagining them in place of any junior copywriter, we need to understand how GPT and all that software captures information. Not only for a matter of potential information theft, but also for a reliability issue. The ethical question is far from closed.

GPT is still far from perfect, but anyone working in the sector must prepare for its arrival and learn to exploit it to their advantage, in order to make the most of this very powerful innovation.

The promising results of this tool suggest that we can consider it a competition, or a useful tool for one's daily work that enhances one's ability to associate facts and thoughts.

Which side will you be on?

GPT-3, the algorithm that writes like a human arrives in Italy