Now that the bubble of Metaverse is deflating, a new word is taking its place in social networks, blogs and videos related to technology. In this post we explain what ChatGPT is and what it is for.
To put it briefly, the restless Mr. Elon Musk founded an artificial intelligence research lab called OpenAI. That laboratory created a model that allows you to interact with artificial intelligence as if you were conversing normally. But, long before that, Musk was already gone.
What is it and what is ChatGPT for?
In the own words of the project website:
We have trained a model called ChatGPT that interacts conversationally. The dialog format makes it possible for ChatGPT to answer follow-up questions, admit mistakes, challenge incorrect premises, and reject inappropriate requests.. ChatGPT is a sister model to InstructGPT, which is trained to follow an instruction or question and provide a detailed response.
What makes ChatGPT different from other chatbots is that does not work on the basis of a set of rules or commands as the responses to users arise from machine learningeither. This learning is based on an enormous amount of data from online forum threads, blog articles, social media posts, and many other sources. This information helps ChatGPT increase its understanding of natural language allowing it to better identify and answer the purpose of the question. Twitter is no longer among those sources of knowledge. In his social network account, the South African millionaire clarified:
Not surprising, as I just found out that OpenAI had access to the Twitter database for training. I put it on pause for now.
I need to understand more about the governance structure and revenue plans (Of OpenAI) in the future.
OpenAI started as open source and not for profit. Neither is still true.
A conversational bot (Chatbot) is a software application that can interact in a human-like conversation based on user input. In the case of ChatGPT, its developers promise that it is capable of simulating dialogue, answering follow-up questions, admitting errors, challenging incorrect premises, and rejecting inappropriate requests.
However, this ability is not limited to deep questions about intriguing scientific enigmas. You may also be asked about topics such as sporting events or your opinion on the weather.. Possible uses include content creation, real-time adaptation of digital marketing campaigns, customer service, and finding and correcting errors in computer programs.
However, not everything is as perfect as it sounds. The developers themselves detected a tendency to respond with "answers that sound plausible but are incorrect or nonsense". Also, like some human experts, he seems to talk too much.
The model is often excessively verbose and overuses certain phrases, such as reaffirming that it is a language model trained by OpenAI. These problems arise from biases in the training data (trainers prefer longer answers that appear more complete) and well-known over-optimization issues.
I wanted to personally test ChatGPT to include examples in this article, however, after asking for my email and password, verifying it, putting my name and phone number, writing the code that was sent to my phone, and explaining my purpose, I was asked to send an email telling them why I wanted access. We will have to settle for the examples of the project website.
It is too soon to know if anything will come out of ChatGPT and Artificial Intelligence will keep its promises. I have been waiting since 1985 for the disappearance of the keyboard and the definitive replacement of paper in offices and procedures. As the twitterer @OrwellGeorge defined it very well:
When I was a kid they made me draw about 2022 and I made flying cars. Tomorrow in the 2022 census they are going to come to ask me if I have a toilet.
Anecdotally, that census went wrong due to a failure to collect data and will probably have to be repeated next year.