“It works very well, but it’s not magic”: This is ChatGPT, the new artificial intelligence that pushes boundaries | technology

ChatGPT artificial intelligence model has more than 1 million users in just five days after its launch. The system answers any question not only in English, but also in correct Spanish. This consistency of dialogue makes it almost inaudible. OpenAI, its creator and one of the leaders in the field of artificial intelligence, also opened its use to any user, so the grids were filled with screenshots of solved programming requests, answers to deep questions, traps, tricks, but also obvious errors. . “It works very well, but it’s not magic,” says Álvaro Barbero Jimenez, director of artificial intelligence at the Institute of Knowledge Engineering (IIC).

“It’s a massive innovation, I was blown away, too,” explains Barbero Jimenez. “I asked him to solve some programming exercises that I give to those who join my team and he did them perfectly. But then I asked him about a programming project that had a silly bug, and he told me a completed solution. He told me to use a tool that doesn’t exist,” he adds.

The way to use it is very simple. Simply go to the chat.openai.com page, open an account with your email and password, and start chatting with the bot.

This ChatGPT comes in an amazing year of advancement of various AI models. Summer is the time for illustrations created from text requests. Prior to this, text generation from raw syntax had arrived with the GPT-3 model. Now the latest version of GPT software again surprises with human ingenuity. It is then that the debate begins between two groups: the enthusiasts who elevate the achievement to a precursor to a general artificial intelligence, which is reasoning and conscious, almost human, and those who contextualize it and quell the excitement. However, no one argues that it is a technological achievement.

One of the experts who see this most optimistically is Carlos Vinolosa, CEO of Optimus Price. “He is capable of doing that thinking. That is, understand why you are given this answer and not another. It’s the first time I’ve seen it in an artificial intelligence.”

The model itself is somewhat wiser in its interpretation. This is what ChatGPT answers in Spanish to the question “Who are you?”: “As a language model trained by OpenAI, I can process language and speak in a similar way to a human. This means that I can understand questions and provide coherent answers using human language. Unlike a human, I don’t have Self-awareness or the ability to think and reason independently. I am just a computer program designed to mimic human language.” It seems like a response intentionally provided by a machine to calm frightened humans with its ability.

Julio Gonzalo, Vice-Chancellor for Research at UNED, is also surprised by the initiation of chat, but he’s wiser and admits he has a long way to go: “He seems to know any topic and in superficial conversation he can convince you, but if you dig deeper you realize he’s speaking from hearsay or, To put it bluntly, he doesn’t know what he’s talking about. He intuitively learned to conduct conversations on almost any topic, seems to have real knowledge and can combine it with rational thinking. But he is an appearance: he has neither learned reason nor clear knowledge, “he explains.

But the most cautious and the most optimistic agree that it is only a matter of time. Few expect it to be here in 2022. The difference between the two sectors is to maintain or push the hype, but few suspect that this is a breakthrough with potential that is difficult to anticipate. “You have to polish a lot of details and contexts where you give a bad answer. But it’s not dangerous either. ChatGPT can make mistakes like anyone, and it doesn’t have to be perfect either,” says Fenollosa.

It was hard to stop the enthusiasm when a user asked him, in English, how to make a peanut butter sandwich out of a video with a biblical tone or when he had a discussion in Spanish about criminal law. Manuel Maroto, a professor at Complutense University, was surprised by the results: “Their level is good, considering the length of the responses,. The questions I posted on Twitter are very complex and cannot be answered in a paragraph: but as a paragraph answer they are very plausible. The separate thing is He says things relatively incorrect: The fact that guilt fraud is an old issue in criminal law, which was resolved decades ago in the opposite direction of what says. Hence the funny thing in the eyes of a lawman is that an AI put itself in a position that has been transgressed for so long, and then denies doing so,” he explains. But just having a bot built at a Silicon Valley startup be able to speak criminal law in language Spanish is devastating.

Maroto, over the objections, did not want to compare the answers of the hypothetical student or even with those of the teacher. Because the code is not meant to be an aggregation of knowledge “but rather a way of thinking and arguing,” but for the rest, he says, ChatGPT is “very impressive.”

The same impression was left by Elena Alvarez Melado, a linguist at UNED. “The fluency, ability to articulate sentences, coherence, and her apparent ability to reproduce the linguistic ability that we as humans expect to find in another human being is admirable,” he says. Then give him some challenges. First about the ambiguity: “In a list like coconut oil is made from coconuts, olive oil is made from olives, what is baby oil?” We know it’s not made for kids. He responded well and even gave the ingredients,” he explains.

Alvarez Melado then asked her to form sentences of a certain type: a negative reaction with such items. There failed. We don’t know why. Perhaps there are not enough examples in Spanish. “These are exercises that some teachers put in high school that a student could in principle do, but this system doesn’t know. I don’t know why. It’s a pretty niche but we know that this machine converts human instructions into python very well.” [lenguaje de programación]. It’s interesting to write python from statements but not from the “” clause.

“If ChatGPT were to be used now as a substitute for university, it would be disastrous, because chatGPT lies with the same confidence with which it says real things,” says Gonzalo, who adds a good comparison of the profile a bot can have today: “One crisis caller told me that when you There is a crisis to manage in the networks (for example, a fire in an industrial plant), the most dangerous profile on Twitter is the my books. The one who seems to know, but in fact does not know. Because people attribute to it knowledge and then brown chaos. Well im gpt is a hypermegaenteradillo”, Explain.

Gonzalo sets an example for 20th-century poets on Twitter. Chat responses sound like in-laws but are polite and courteous. “I asked him to tell me of the Spanish poets of the twentieth century, and he gave me a very good list. So I made up that I had made a mistake in one, and that he was actually from the sixteenth century (all wrong). Well, he agrees with me, apologizes and invents his dates of birth and death in the century XVI. Imagine how dangerous it is to replace teachers!”, He explains.

The trade in which artificial intelligence can help or even replace it has expanded a lot. The danger of machinery is no longer limited to industrial facilities. This chat knows how to answer questions that inexperienced programmers, teachers or writers may answer less clearly. There are class assignments like essays or essays that are destined to die with this technology.

“The programmer’s work, with the tests we’ve done, shows us that he will know how to do it very well,” says Victoriano Izquierdo, co-founder of Graphext, a data analytics company. “It will make it possible to reach the level of an operating system and do really complex tasks. Perhaps paradoxically now that professions that require more physical skills, such as a dentist or a hairdresser, seem difficult to replace even hardware Go further and don’t be afraid of the mechanical Edward Scissorhands,” he adds.

Fenollosa goes even further with a list of jobs allegedly at risk. “It all charges Copywriters, including journalists (if they don’t). For example, in press releases or current affairs summaries. Also FAQ, Stackoverflow type [página de dudas sobre código, donde han tenido que bloquear el uso del chat por exceso de errores] or Quora [preguntas genéricas]Including work to solve the doubts of many teachers. Simple programming, design and illustration. You can tell ChatGPT to “Build me a website for a building materials manufacturer and include several stock photos of people visiting a construction site” and it will give you a starting point that would have taken a human a few days to prototype,” he explains.

Venulosa says the bot will be able to tell good jokes. “I don’t mean that I can write funny stories by combining parts of other stories. ChatGPT understands sarcasm and elements of humor, it is able to come up with new jokes,” he says.

Of course, this will affect the professions. “On the scale between nothing happening and this being the end of humanity, we are somewhere in the middle,” says Barbero Jimenez.

As concerns mount, it’s reasonable not to be dazzled, as in this exchange about a lease in Catalonia. It looks impressive but it’s actually not:

This newspaper asked him for two poems about journalists. Not very subtle: “But amidst so much lies / The journalist feels weary and despondent / Because sometimes the truth seems / Lost in this bleak place. / And yet he presses on / With determination and courage / To keep searching for the truth / And to show the world what is really going on “. There is little hope that these tools will be able to solve what has so far been a mediocre production. All this would be seen today as “machine production”. Adding value will be completely human, like handicraft.

You can follow country technology in Facebook s Twitter Or sign up here to receive The weekly newsletter.