Wednesday marks a teacher strike day across the UK, placing parents in the role of superintendent the familiar pandemic-inspired homeschooling teachers have for their children. Except this time, there’s a magical automated assistant ready to help.
Educators have cautiously praised ChatGPTExtremely sophisticated chatbot from openAI, saying it could revolutionize education. One homeroom teacher in the UK said it made them rethink homework, while another teacher in Oregon used it to create lesson plans and study guides.
The tool’s personalized responses are what makes it so appealing as a full-blown digital tutor. I used it recently to learn about enzymes, when my 12 year old had questions that I had no hope of answering. When ChatGPT offered a dense technical explanation, I asked it to provide simpler terms and an analogy.
“Sure!” it answers. “Think of a lock on a door. The padlock is like an enzyme and the key is like the substrate molecule…” It extends the analogy further to describe the enzyme’s active site as the keyhole.
These are notable answers. We could dig deeper into every aspect of biochemistry if we wanted to. Unlike a human tutor, ChatGPT can be questioned for as long as you like.
This has huge potential for personalized, independent learning… except that ChatGPT is often flawed and it hides that very well. When I checked one of my daughter’s English homework questions on the tool it came up with a list of eloquent examples, upon closer inspection there was one that was extremely incorrect. . The bot says the protagonist has a turbulent relationship with his parents, even though the character’s parents die throughout the book.
On another occasion, I used this tool to create some linear equations for my daughter to practice. She was confused when I asked the tool to generate different answers than the one she calculated. I asked ChatGPT for an explanation and once again it broke down its method in simple terms that sounded as authoritative as any real math tutor. But when I double checked the answers on Google, it turned out that ChatGPT’s answer was wrong and my answer was correct. Thus, her nightmare of failing her little math exam as well as much of my initial enthusiasm for ChatGPT was over.
New York City’s public school system, the largest in the United States, has banned its students from using ChatGPT, in part because of concerns about the “accuracy of the content.” That is why the recent comparisons between ChatGPT and “computer to write” are a confusing comparison, because computers are always right and ChatGPT is not.
How is it incorrect? An OpenAI spokesperson said the company has updated ChatGPT over the past few months to improve its actual accuracy, but they don’t have statistics to share. The tool also warns users when they open it for the first time that it sometimes makes mistakes.
Will it get more accurate? Yes, but it’s hard to say how much. The large language model that underpins ChatGPT is made up of 175 billion parameters, which are the settings used to make the model’s predictions, compared with the 1.5 billion that its predecessor GPT-2 had. . Artificial intelligence has been accepted that the more parameters are added to a model, the more realistic the model becomes and the correlation is real for GPT. It becomes significantly more accurate when all those parameters are added. Rumor has it that the next version scheduled for release this year, called GPT-4, will be in the trillions of dollars.
The problem is, we don’t know if a big jump in parameters means a big leap in reliability. That’s why students should exercise caution when using ChatGPT, if at all, in the near future.
When I asked Julien Cornebise, professor emeritus of computer science at University College London, if he would ever trust it as a homework tool, he replied: “Absolutely no, not yet.” He pointed out that even if the system is improved, we still do not guarantee that it is honest.
Students should be familiar with corroborating any information the system shares with other online information or with an expert. Albert Meige, vice president of technology focus at consulting firm Arthur D. Little, says his teenage daughter used it to help her with her physics homework — but he was able to authenticate the facts. the answer thanks to his PhD in computer physics. He recommends using chatbots to help better understand the questions posed in homework. “She discovered that she shouldn’t ask a single question,” he said. “It’s an interactive process.”
Use it to get feedback, Cornebise concurs. “That’s what star students do.”
As a relatively small company, OpenAI can get away with presenting a quirky alternative reality. Google and Meta Platforms Inc. of Alphabet Inc. won’t be able to do the same. Google has its own very sophisticated language model called LaMDA, but is extremely cautious about integrating a similar chatbot into its own search engine, possibly due in part to accuracy issues. corpse. Three days after releasing an AI tool that can generate scientific papers, called Galactica, Meta took it down after academics criticized it for generating unreliable information.
OpenAI will be held to the same high standards as the synthetic AI arms race heats up and chatbot technology is integrated into search engines in the US and China.
Until then, use it with caution and a little reasonable skepticism, especially in the education sector.
© 2023 Bloomberg LP