“Can you imagine having at your disposal a virtual assistant that can understand your needs and answer your questions fluently and accurately? Well, that’s what ChatGPT offers, the natural language model developed by OpenAI that has revolutionized the way we understand the artificial intelligence. With its ability to generate coherent text in a variety of languages and styles, it will be a must-have for companies in all economic sectors.”
This paragraph has not been written by a journalist. It is also not an expert statement. It is content created by a robot. By ChatGPT itself. We gave him very basic instructions: write the beginning of a story for an economic newspaper explaining what ChatGTP is and its relevance, and in a matter of seconds he generated said text. The first version was somewhat boring and we asked him to redo it with something more hook to catch the reader. The style may be debatable, but it is undeniable that the text is coherent and passes for human-created content.
ChatGPT is the latest hit from Silicon Valley, so prone to generating over-expectations around new technologies. Many speak of a disruptive advance that will have important implications for a multitude of industries. After playing with it for several days, it’s easy to catch the collective fervor because its possibilities are astounding.
This natural language chatbot is a big step in the long road that generative artificial intelligence still has ahead of it, that is, that capable of generating original content. “The collective enthusiasm is understandable. Both this chatbot and generative artificial intelligence are going to produce very important changes in practically all sectors. We are taking the first steps, but it is going to advance quickly”, defends Enrique Puertas, professor of artificial intelligence and big It dates from the European University of Madrid.
The model developed by OpenAI has shown us that we are on the brink of a revolution in the way we interact with computers. This chatbot, which is currently free, answers questions in writing and is capable of creating coherent and credible content that seems to be written by a human at amazing speed.
Five days after its launch at the end of November, more than a million people had already used it. He has been asked to write stories, children’s stories, poems, dialogues, school work, popular texts on scientific concepts, healthy menus, training routines, news, advertisements, lines of code, and much more. Some interactions from which ChatGPT continues to learn and improve to take the next step: its economic monetization. In fact, it has emerged that OpenAI is preparing a paid professional version.
ChatGPT is relevant because it has managed to bring the possibilities of artificial intelligence closer to the general public. Simply create an OpenAI account to start a conversation with the robot through a simple interface. “It is a technology that is very well trained to converse with us using natural language. They have given us an amazing toy and people are excited because it does not seem that we are talking to a robot,” sums up Juan Ignacio Rouyet, director of Delivery at Eraneos , formerly Quint.
The chatbot, based on the latest version (GPT-3.5) of the OpenAI natural language model, has been trained by feeding it millions of text inputs in various languages until 2021: all of Wikipedia, books, news, web pages, scientific articles and code repositories, among others. When asked a question, it is able to understand it and search its database for related information, as well as infer and generate answers based on this knowledge. In very simple terms, the AI model can predict the next word you need to type based on previous sequences. “He has been trained with millions and millions of texts so that he learns how language works in different languages and he has also learned the content,” says Puertas.
Since it is not connected to the Internet, it does not have information on current events. “It’s a matter of time before it connects to the web, for now they are interested not so much in validating the content but the form; testing the natural language model,” adds Puertas. For now, OpenAI is expected to announce a new version of its natural language model, GPT-4, this year, which promises to take a huge leap forward.
Obviously, the chatbot has limitations and raises concerns. OpenAI has put up barriers so that the chat does not generate violent content, hate speech or sexually explicit content. Also, avoid answering questions that could be illegal or with a high risk of being misused, for example on political or religious topics.
Furthermore, the robot’s knowledge is limited to the texts with which it has been trained and has biases derived from the data with which it was trained. He himself lists other limitations such as a lack of common sense, his inability to have a deep understanding of human emotions, and the risk that he can generate coherent but not necessarily true answers. In addition, he acknowledges that he has been trained with texts protected by copyright, “although only for educational and research purposes”, he justifies himself.
Its potential impact is undeniable. For example, it may change the way we search for information. Instead of consulting search engines like Google, which returns a list of links, generative artificial intelligence offers ready-made answers. Beyond the threat to Google -which, obviously, is also working on these technologies in projects such as LaMDA-, Rouyet points out that “there is an obvious risk that people believe that everything artificial intelligence says is true and correct. But We really don’t know what’s behind it, we don’t know how he’s been trained, what sources he’s drinking from. We lose the possibility of consulting the sources as we do now in the links that a search engine offers you.”
Will it also change the foundations of human learning? “Artificial intelligence can provide quick and easy answers, but it does not develop critical thinking and problem solving skills,” criticized the New York Department of Education a few weeks ago to justify the ban on using ChatGPT in schools.
ChatGPT is only a part of what generative artificial intelligence promises, which also applies, for example, to code development (Codex) or image creation with programs like Dall-E-2 (also from OpenAI), Midjourney or Stable Diffusion, “which doesn’t envy anything Illustrator can do,” says Puertas.
Until recently, machines were very good at analyzing data, but they couldn’t compete with humans when it came to creating content. “This is changing. Generative artificial intelligence is on the way to becoming faster, cheaper and sometimes even better than human creations,” the venture capital firm Sequoia says in a report.
The list of sectors that can be impacted is enormous because it covers all areas where humans create original work: marketing, media, advertising, architecture, graphic design, software development, video games, customer service, and many more. “We are talking about billions of workers. Generative artificial intelligence can make them at least 10% more efficient and/or creative and has the potential to generate trillions of economic value,” says Sequoia, which ensures that we are at a point of inflection and about to see a wave of revolutionary applications of this technology.
What’s more, they are convinced that by 2030 this artificial intelligence will be so advanced that it will generate creative texts and graphic works better than those a professional writer or artist can produce. “I don’t doubt it, and maybe we won’t have to wait that long,” says Puertas. Also, as the technology becomes smarter by incorporating industry- or company-specific knowledge, it could be applied to a multitude of economic sectors, Rouyet says.
The interest of Big Tech is evident. Google, Meta and Amazon are also working on generative AI, but they are cautious because any slip up could have a considerable reputational impact. Microsoft has just announced the public availability of OpenAI technology in its Azure cloud, so that companies will be able to address use cases with the artificial intelligence models of this company, including ChatGPT. In addition, it wants to integrate this technology into its Bing search engine and its office automation products.
In fact, Microsoft has invested 1,000 million dollars in OpenAI -founded in 2015 by entrepreneurs such as Elon Musk, Peter Thiel and Sam Altman- and it has emerged that it is negotiating to put another 10,000 million in the company, which also seeks to give entry to new investors to a valuation of 29,000 million dollars.
All this enthusiasm (in addition to OpenAI, there are other start-ups such as Jasper, Stability AI or Hugging Face) has infected venture capital, which is already in a race to take positions, which can lead to exorbitant valuations of companies that do not yet have a business model .
