In just six months, the number of searches for “artificial intelligence” on Google has increased fivefold. ChatGPT — launching November 30, 2022 — already has tens of millions of users. And Sam Altman, CEO of OpenAI, which developed ChatGPT, has already appeared in court to explain and answer questions about the impact of AI. By comparison, it took Mark Zuckerberg 14 years to go to Washington to talk about Facebook’s role in society.
Altman, oddly enough, has been very outspoken about the technology his company produces. “My biggest fear is that we could do some serious harm to the world … I think if this technology doesn’t work, it could go very wrong,” he said. said as he testified. However, some analysts believe that his supposed “fear” language is a more rigorous move to deter the rise of competitors to OpenAI, which already holds a dominant position in the space. They say it may have been carefully calculated to encourage regulation.
Heavy and bombastic phrases about AI explosions have already spawned their own memes. The term “critical hype”, coined in 2021 to define criticism of new technology, has become popular thanks to ChatGPT. A pioneering example of “critical hype” is the case of Cambridge Analytica, where the company was accused of collecting Facebook data to understand and influence voters during the 2016 presidential election.
The culmination of these statements was the departure of Jeffrey Hinton, known as the “Godfather of AI,” from Google. He left the company so that he could speak freely about the dangers of AI. “According to what we know so far about how the human brain functions, our learning process is probably less efficient than that of computers,” he told his ELPAÍS. Interview after leaving Google.
Meanwhile, the outgoing UK government’s chief scientific adviser just said AI “could be as big as the industrial revolution”. There are already groups out there trying to organize so that their deals aren’t swept away by this technology.
The predictions and concerns about AI are too numerous to list. However, the impact of this technology may actually be bearable. What if everything went slower than expected and there was less social and economic turmoil? While it’s hard to deny the impact of AI in many areas, changing the world is not so easy. While previous revolutions have changed our way of life, humans have managed to adapt without much disruption. Could AI ultimately be a subtle revolution, too?
“at least, [AI has caused] It’s a big structural shift in what you can do with software,” says Benedict Evans, an independent analyst and former partner at Andreessen Horowitz, one of Silicon Valley’s leading venture capital firms. “Maybe a lot of new things will be possible. That’s why people compare it to the iPhone. It can be more like a “graphical user interface” that allows you to interact with your computer.
These new AI and machine learning (ML) technologies are clearly of great importance in the tech world. “My concern is not that AI will replace humans,” says Meredith Whittaker, president of Signal, a popular messaging app. The danger is not that AI will replace workers’ jobs, but that the adoption of AI by employers will be used to make those jobs worse, further exacerbating inequalities. ”
It should be noted that new forms of AI still make many mistakes. Jose Hernandez Orallo, a researcher at the University of Cambridge’s Leverhume Center for Future Intelligence, has been studying these so-called “hallucinations” for many years. “at this point, [AI is] Know-it-all brother-in-law level. However, in the future [it may be] Perhaps an expert on some subjects more than others. This is what causes us anxiety. Because I still don’t know which subject is what subject. [the AI] is the most reliable,” he explains.
“It is impossible to build a system that never fails, because we are constantly asking more and more complex questions. ..they are very unpredictable, ”he added.
But if the technology is not so mature, why has it had such a sudden and widespread impact in the last few months? There are at least two reasons, Hernández Orallo said. increase. One is commercial pressure. “The biggest problem arises because there is commercial, media and societal pressure that these systems must always respond to something, even if they don’t know how. These systems fail less often if is set, but since there are thousands of ways to summarize text, most of the time the answer is “I don’t know.” ”
The second reason, he points out, is human perception. “We have the impression that the AI system has to be 100% accurate to him. It’s like a calculator mixed with an encyclopedia.” But it’s not. “For language models, it is easy to generate plausible but incorrect text. The same thing happens with audio, video and code. Humans do it all the time. responds with phrases that are good but may not make sense to children, just say, “That’s weird,” not go to the pediatrician and say, “My son has a lot of hallucinations.” yeah. For both children and certain types of AI, [there is an ability] It’s about mimicking things as much as you can,” he explains.
If it becomes clear that there are things AI is not completing properly, the major impact on the labor market will fade. Similarly, disenchantment sets in when AI is asked a question and we are unsure of the answer it provides. For example, if a student asks a chatbot about a particular book they haven’t read, they may have a hard time determining it. If the synopsis is completely trustworthy. Sometimes even the slightest suspicion is unacceptable. Perhaps in the future, humans using AI will assume (and accept) that the technology makes certain errors. But despite all the hype, we’re not there yet.
Long-term recognition of the limited impact of AI does not remove the main concern that AI is superior to human intelligence. In the collective imagination, this horror resembles the notion that machines control the world’s software and destroy humans.
“People use this concept for everything,” Hernández-Olaro shrugs. “question [that really need to be asked when thinking about] How much capacity should a general-purpose system like GPT-4 have? Should it be more powerful than humans? And what kind of person is the average person, the smartest person?For what specific tasks is it used? [the answers to these questions] Very poorly defined at the moment. ”
“Ever since we imagined machines to replace humans, we now fear them,” said Matt Bean, a professor at the University of California, Santa Barbara. We have strong evidence of how we rely not only on imagination and assertiveness, but on criticism and fear.”
The most recurring emotion regarding this issue is fear. “We seem to be in a kind of trance about these things.” [AI] The system communicates our experience to these machines,” Whittaker says. “Reflexively, we think of them as humans…we start to think they are listening to us. Looking at the history of systems that predate ChatGPT, these systems were far It is worth noting that, despite their lack of sophistication, the responses were often the same: when people used these systems, they found themselves in a vicarious intimate relationship with them. And then, just like today, “experts” believed that these systems could soon (always “soon,” never “now”) completely replace humans. I expected it to. ”
please sign up weekly newsletter To get more news coverage in English from EL PAÍS USA Edition
