Openai: A Game changer for AI Language, and What Do We Need to Learn About It? An MIT Assistant Professor’s Answers on GPT-3.5
Over the past week, a lot of people have fallen under the spell of the freechatgt, which can answer all sorts of questions with stunning and unprecedented eloquence.
The method used to help the bot answer questions, which Openai has shown off before, seems to be a huge step forward in helping the bot handle language in a more personable way. The technique is impressive, despite the fact that he thinks it may make his job more complicated. It has made me think of what my courses will look like if they require short answers on assignments.
Openai has not released a full accounting of how it created a naturalistic new interface for its software, but the company did reveal some information in a post. It says the team fed human-written answers to GPT-3.5 as training data, and then used a form of simulated reward and punishment known as reinforcement learning to push the model to provide better answers to example questions.
Jacob Andreas, an assistant professor who works on AI and language at MIT, says the system seems likely to widen the pool of people able to tap into AI language tools. He tells us that it’s a familiar interface that makes us apply a mental model we used to apply to other agents.
Others disagree that ChatGPT is such a game changer, noting that students have long been able to outsource essay writing to human third parties through ‘essay mills’. “It doesn’t necessarily add much functionality that wasn’t available to students already if they knew where to look,” says Thomas Lancaster, a computer scientist and academic-integrity researcher at Imperial College London.
The end of essays is looking to be an assignment for education at the moment, according to a law student who studies innovation and society at a UK university. A student at Arizona State University would have gotten a good grade if Dan Gillmor had asked his students a homework question that he fed them and then asked them to write a response to it.
Nature wants to discover the extent of concerns about the impact of artificial-intelligence tools on education and research integrity, and how research institutions are dealing with them. Our poll is here.
Creating Artificial Intelligence in Chatbots: What Will They Have to Say About Us if We Don’t Have an Educator?
Lancaster acknowledges that something is put into a free package. But he thinks that ChatGPT-generated essays will out themselves more readily than the products of essay mills, by including quotes that weren’t actually said, incorrect information created through false assumptions, and irrelevant references.
“Despite the words ‘artificial intelligence’ being thrown about, really, these systems don’t have intelligence in the way we might think about as humans,” he says. They are trained to create a pattern of words based on what has happened before.
How necessary that will be depends on how many people use the chatbot. More than one million people tried it out in its first week. Although it is free, the current version is not likely to be free forever and some students might be a bit uneasy about paying.
She’s hopeful that education providers will adapt. “Whenever there’s a new technology, there’s a panic around it,” she says. It is the responsibility of academics to have a lot of distrust but I don’t think it is a problem.