The Turing test is a subjective measure. It depends on whether the people asking the questions feel convinced that they are talking to another person when in fact they are talking to a device.
But whoever is asking the questions, machines will soon leave this test in the rearview mirror. (...)
ChatGPT, a bot released in November by OpenAI, a San Francisco lab, leaves people feeling as if they were chatting with another person, not a bot. The lab said more than a million people had used it. Because ChatGPT can write just about anything, including term papers, universities are worried it will make a mockery of class work. When some people talk to these bots, they even describe them as sentient or conscious, believing that machines have somehow developed an awareness of the world around them.
And yet these bots are not sentient. They are not conscious. They are not intelligent — at least not in the way that humans are intelligent. Even people building the technology acknowledge this point.
These bots are pretty good at certain kinds of conversation, but they cannot respond to the unexpected as well as most humans can. They sometimes spew nonsense and cannot correct their own mistakes. Although they can match or even exceed human performance in some ways, they cannot in others. Like similar systems that came before, they tend to complement skilled workers rather than replace them. (...)
“These systems can do a lot of useful things,” said Ilya Sutskever, chief scientist at OpenAI and one of the most important A.I. researchers of the past decade, referring to the new wave of chatbots. “On the other hand, they are not there yet. People think they can do things they cannot.”
As the latest technologies emerge from research labs, it is now obvious — if it was not obvious before — that scientists must rethink and reshape how they track the progress of artificial intelligence. The Turing test is not up to the task. (...)
ChatGPT is what researchers call a neural network, a mathematical system loosely modeled on the network of neurons in the brain. This is the same technology that translates between English and Spanish on services like Google Translate and identifies pedestrians as self-driving cars weave through city streets.
A neural network learns skills by analyzing data. By pinpointing patterns in thousands of photos of stop signs, for example, it can learn to recognize a stop sign.
Five years ago, Google, OpenAI and other A.I. labs started designing neural networks that analyzed enormous amounts of digital text, including books, news stories, Wikipedia articles and online chat logs. Researchers call them “large language models.” Pinpointing billions of distinct patterns in the way people connect words, letters and symbols, these systems learned to generate their own text.
They can create tweets, blog posts, poems, even computer programs. They can carry on a conversation — at least up to a point. And as they do, they can seamlessly combine far-flung concepts. You can ask them to rewrite Queen’s pop operetta, “Bohemian Rhapsody,” so that it rhapsodizes about the life of a postdoc academic researcher, and they will.
“They can extrapolate,” said Oriol Vinyals, senior director of deep learning research at the London lab DeepMind, who has built groundbreaking systems that can juggle everything from language to three-dimensional video games. “They can combine concepts in ways you would never anticipate.” (...)
The result is a chatbot geared toward answering individual questions — the very thing that Turing envisioned. Google, Meta and other organizations have built bots that operate in similar ways. (...)
Turing’s test judged whether a machine could imitate a human. This is how artificial intelligence is typically portrayed — as the rise of machines that think like people. But the technologies under development today are very different from you and me. They cannot deal with concepts they have never seen before. And they cannot take ideas and explore them in the physical world.
ChatGPT made that clear. As more users experimented with it, they showed off its abilities and limitations. One Twitter user asked ChatGPT what letter came next in the sequence O T T F F S S, and it gave the correct answer (E). But it also told him the wrong reason it was correct, failing to realize that these are the first letters in the numbers 1 to 8.
At the same time, there are many ways these bots are superior to you and me. They do not get tired. They do not let emotion cloud what they are trying to do. They can instantly draw on far larger amounts of information. And they can generate text, images and other media at speeds and volumes we humans never could.
Their skills will also improve considerably in the coming years. (...)
In the months and years to come, these bots will help you find information on the internet. They will explain concepts in ways you can understand. If you like, they will even write your tweets, blog posts and term papers.
They will tabulate your monthly expenses in your spreadsheets. They will visit real estate websites and find houses in your price range. They will produce online avatars that look and sound like humans. They will make mini-movies, complete with music and dialogue.
“This will be the next step up from Pixar — superpersonalized movies that anyone can create really quickly,” said Bryan McCann, former lead research scientist at Salesforce, who is exploring chatbots and other A.I. technologies at a start-up called You.com.
As ChatGPT and DALL-E have shown, this kind of thing will be shocking, fascinating and fun. It will also leave us wondering how it will change our lives. What happens to people who have spent their careers making movies? Will this technology flood the internet with images that seem real but are not? Will their mistakes lead us astray?
by Cade Metz, NY Times | Read more:
Image: Ricardo Rey
