Published on: 04/12/2022 – 04:52 Modified on: 04/12/2022 – 04:50
SAN FRANCISCO (AFP) – California startup OpenAI has launched a conversational robot (chatbot) that can answer a wide variety of questions, but whose impressive performance is reigniting the debate about the risks of artificial intelligence (AI) technologies. ) .
Conversations with ChatGPT posted by intrigued netizens, notably on Twitter, reveal some kind of omniscient machine capable of explaining scientific concepts, writing a theatrical scene, writing a university thesis… or even perfectly working lines of computer code.
“His answer to the question + what to do if someone is having a heart attack + was incredibly clear and relevant,” Claude de Loupy, director of Syllabs, a French company specializing in automatic text generation, told AFP.
“When you start asking very specific questions, ChatGPT can answer right off the bat,” but overall its performance remains “really impressive,” with a “rather high level of language,” he says.
The start-up OpenAI, co-founded by Elon Musk in San Francisco in 2015 – the head of Tesla left the company in 2018 – received 1 billion US dollars from Microsoft in 2019.
It is particularly known for two automated generation software, GPT-3 for text generation and DALL-E for image generation.
ChatGPT is able to ask its interlocutor for details and “has fewer hallucinations” than GPT-3, which despite its capabilities can give completely different results, reports Claude de Loupy.
Cicero
“A few years ago, chatbots had the vocabulary of a dictionary and the memory of a goldfish. Today, they’re much better at responding consistently based on request and response history. They’re more goldfish,” notes Sean McGregor, a researcher at AI – related incidents in one database.
Like other programs based on deep learning, ChatGPT retains a major weakness: “It has no access to meaning,” recalls Claude de Loupy. The software cannot justify its choices, that is, explain why it put the words that make up its answers together in this way.
However, AI-based technologies that can communicate can increasingly appear to be really thinking.
Researchers from Meta (Facebook) recently developed a computer program called Cicero, named after the Roman statesman Cicero.
The software has proven itself in Diplomacy, a board game that requires negotiating skills.
“Unless he speaks like a real person – shows empathy, builds relationships and speaks the game properly – he won’t be able to form alliances with other players,” the social media giant said in a statement.
Character.ai, a startup founded by former Google engineers, released an experimental chatbot online in October that can take on any personality. Users create characters after a short description and then can “talk” to fake Sherlock Holmes, Socrates or Donald Trump.
“Simple Machine”
This level of sophistication intrigues but also worries many observers at the prospect of these technologies being misused to trick people, for example by disseminating false information or creating increasingly credible scams.
What does ChatGPT “think” about this? “There are potential dangers in building sophisticated chatbots (…) people might think they are interacting with a real person,” acknowledges the chatbot interviewed by AFP on the subject.
Companies therefore take security precautions to prevent misuse.
On the homepage, OpenAI clarifies that the chatbot may generate “incorrect information” or “produce dangerous instructions or biased content”.
And ChatGPT refuses to take sides. “OpenAI has made it incredibly difficult to get him to speak his mind,” says Sean McGregor.
The researcher asked the chatbot to write a poem on an ethical topic. “I am a mere machine, a tool at your disposal / I have no power to judge or to make decisions (…)” the computer answered him.
“Interesting to see people questioning whether AI systems should behave the way users want or the developers intended,” tweeted Sam Altman, co-founder and chief of OpenAI, on Saturday.
“The debate about what values to place on these systems will be one of the most important that a society can have,” he added.
© 2022 AFP