ChatGPT is a language model developed by OpenAI that uses natural language processing technology to generate consistent and natural responses to questions and text commands.
It trains on large amounts of text data to learn to understand and generate human languages.
It is capable of answering questions, generating texts, translating languages, and performing many other language-related tasks.
It is a very powerful tool for applications such as chatbots, virtual assistants, and content generation systems.
Attention: the previous paragraph was written with the ChatGPT system according to the slogan: "Write a short and attractive paragraph explaining what ChatGPT is".
We will talk about this system because it is the star of the moment, but it is to be expected that fierce competition will break out in a very short time.
To demystify: machine learning implementations are based on simple and limited models of the functioning of neurons to be simulated many times on a computer.
These networks of simulated neurons undergo a training process with a known set of input numbers and a known set of output numbers.
At input or output, sets of numbers can represent text, images, sounds, etc.
In essence they are boxes of many numbers that are adjusted.
The adjustment is by gaining experience: reward/punish the model if it succeeded or not what it should have produced.
These training for elaborate tasks require very high computational resources.
Classic examples of use: recognize handwritten characters and thus read documents;
or identify cats, dogs, etc.
in an image.
A model can be trained so that texts enter as a slogan and other texts exit.
ChatGPT falls into this category and, as you can imagine, its internal workings are very complicated.
The quality of the model depends on the reference data set to train it on and the care with which that collection was put together.
Once trained, the system responds based on its experience.
ChatGPT's crystal ball reaches 2021 because the huge databases used to train it currently reach that year.
Is this true intelligence as we normally understand it?
The answer is: definitely not.
In some respects the fruits of human intelligence (infinitely richer in general) and the new artificial fruits may come to resemble each other, but in others - most - they will undoubtedly be very different.
This great sensation of the moment comes with ethical problems that are no longer trivial at all.
Abstracts of papers generated by ChatGPT were recently submitted to various scientific journals, and 32% were considered “real” by experts who read them.
In a tragicomic extreme, there are already conferences on artificial intelligence that have issued bans on this type of contribution.
Questions are already arising about the intellectual property of an artistic work -literary, visual or musical- coming out of this type of system.
And what about legal liability for a failed computer program or business strategy if it was the work of these models?
Sooner or later we will have to establish certain contours to live with these dilemmas and many others: the storm is coming.
That is the challenge for the future and the germ of new ethical problems that we should already be debating as a society.w