ChatGPT, artificial intelligence and the future of education

A few weeks ago, Ethan Mollick, a professor at Wharton, told his MBA students to play with Google Tags, a model of artificial intelligence, and see if the tech could write an essay based on one of the topics covered in his course. Granted, the mission was primarily a gimmick intended to illustrate the power of technology. Still, the algorithm-generated essays — albeit flawed and a little too reliant on the passive voice — were at least reasonable, Mollick recalled. They also passed another critical test: a screening by Turnitin, a popular anti-plagiarism software. The AI, it seemed, suddenly got pretty good.

This is certainly the case right now. Over the past week, screenshots of conversations with ChatGPT, the latest version of the AI ​​model developed by research firm OpenAI, have gone viral on social media. People have steered the tool, which is freely available online, to crack jokes, write TV episodes, compose music and even debug computer code – all things that AI also has to do. More than a million people I’ve played around with the AI ​​now, and while it doesn’t always tell the truth or make sense, it’s still a pretty good writer and even more confident bullshit. Along with recent updates to SLABOpenAI’s art generation software, and Lensa AI, a controversial Platform which can produce digital portraits using machine learning, GPT is a stark red flag that artificial intelligence is starting to rival human capabilities, at least for some things.

“I think things have changed very dramatically,” Mollick told Recode. “And I think it’s only a matter of time for people to notice.”

If you are not convinced, you can try it yourself here. The system works like any online chatbot, and you can just type and submit any question or prompt you want the AI ​​to address.

How does GPT work? At its core, the technology is based on a type of artificial intelligence called a language model, a prediction system that essentially guesses what to write, based on previous texts it has processed. GPT was built by training its AI with an extraordinarily large amount of data, much of it from the vast supply of data on the internet, along with billions of dollars, including seed funding of several prominent tech billionaires, including Reid Hoffman and Peter Thiel. ChatGPT has also been trained on examples of two-way human conversation, which helps it make its dialogue much more human, like a blog post published by OpenAI explains.

OpenAI is trying to commercialize its technology, but this current version is supposed to allow the public to test it. The company made headlines two years ago when it released GPT-3, an iteration of the technology that could produce poems, role-play and answer certain questions. This latest version of the technology is GPT-3.5, and ChatGPT, its corresponding chatbot, is even better at text generation than its predecessor. It’s also good enough to follow instructions, like, “Write a short frog-and-toad story where the frog invests in mortgage-backed securities.” (The story ends with Toad taking Frog’s advice and investing in mortgage-backed securities, concluding that “sometimes taking a little risk can pay off in the end”).

The technology certainly has its flaws. While the system is theoretically designed not to cross certain moral red lines – he is adamant that Hitler was bad – it’s not hard to trick the AI ​​into sharing tips on how to engage in all sorts of evil and nefarious activities, especially if you tell the chatbot they’re writing fiction . The system, as other AI models, can also say biased and offensive things. As my colleague Sigal Samuel said Explainan earlier version of GPT generated extremely Islamophobic content and also produced some rather concerning talking points about the treatment of Uyghur Muslims in China.

GPT’s impressive capabilities and its limitations reflect the fact that the technology works like a version of Google’s smart writing suggestions, generating ideas based on what it’s read and processed before. For this reason, the AI ​​can come across as overly confident without displaying a particularly deep understanding of the topic they are writing about. It’s also why it’s easier for GPT to write about commonly discussed topics, like a Shakespeare play or the importance of mitochondria.

“She wants to produce texts that she thought were plausible, given everything she’s seen before,” says Vincent Conitzer, computer science professor at Carnegie Mellon. “Perhaps it sounds a little generic at times, but it is written very clearly. He will probably pick up on points that have often been raised on this particular topic because he has, indeed, learned the kinds of things people say .

So for now, we are not dealing with a know-it-all bot. AI-provided answers were recently banned from the StackOverflow coding feedback platform because they were very likely to be incorrect. The chatbot is also easily tricked by riddles (although its attempts to answer are extremely funny). Overall, the system is perfectly comfortable making stuff up, which obviously doesn’t make sense after human scrutiny. These limitations can be comforting to people worried that the AI could take their jobor possibly ask a threat to human safety.

But the AI ​​is getting better and better, and even this current version of GPT can already do some tasks very well. Consider Mollick’s mission. While the system certainly wasn’t good enough to earn an A, it still worked pretty well. A Twitter user said that during a mock SAT exam, ChatGPT Mark around the 52 percentile of candidates. UNC computer science professor Kris Jordan told Recode that when he gave GPT its final exam, the chatbot received a perfect score, much better than the median score of humans taking its course. And yes, even before ChatGPT went live, students were using all kinds of artificial intelligence, including earlier versions of GPT, to carry out their missions. And they probably aren’t reported for cheating. (Turnitin, the maker of anti-plagiarism software, did not respond to multiple requests for comment).

Right now, it’s unclear how many enterprising students might start using GPT, or if teachers and professors will find a way to catch them. Yet these forms of AI are already forcing us to ask ourselves what kinds of things we want humans to keep doing, and what we’d rather technology discover instead.

“My eighth-grade math teacher told me not to rely on a calculator because I won’t have one in my pocket all the time when I grow up,” said Phillip Dawson, an expert who studies cheating. exams at Deakin University, Recode. “We all know how it happened.

This story was first published in the Recode newsletter. register here so as not to miss the next one!