By: Ben Xu
Technology has developed over vast distances. Humanity dives into a new period — a time when we create models of the human brain to create Artificial Intelligence. These models are called chatbots. They can write poems and answer questions, yet to this day, in July 2024, they still can’t do one thing: Math.
Math is strict. There’s typically only one answer to a math problem. That’s how it differs from language. Language allows flexibility and creativity. Computer scientists created these chatbots to be Large Language Models, which means their general purpose is to focus on language. Chatbots were not created to follow the strict rules of mathematical calculation. Kristian Hammond, a computer science professor and A.I. researcher at Northern University, told the New York Times, “A.I. chatbots have difficulty with math because they were never designed to do it.” Dr. Hammond also believes that chatbots were never really designed to do any sort of math.
Past computers, though, have actually been known for their mathematical skills because programmers from before created them to follow strict rules. Other creations, such as IBM’s Watson, were actually accurate at answering strict problems. Watson, a computer David Furucci built, defeated the world’s best human “Jeopardy!” players.
Later, chatbots and the new form of A.I. sprang into growth. After adapting technology to mimic human brains, many people have attempted to solve the math problem. Kirk Schneder, a high school math teacher, says, “[Chatbots are] usually fine, but usually isn’t good enough in math. It’s got to be accurate. It’s got to be right.” Mr. Schneder believes that chatbots must become more accurate.
Khan Academy, an education company, has attempted to solve this problem. Recently, they have made a huge difference to its A.I. chatbot, Khanmigo. That is, many numeric calculations are sent to a calculator to avoid mistakes by the chatbot. ChatGPT has been using a similar attempt. When ChatGPT has to multiply or divide large numbers, it will just send it to a calculator to solve the calculation.
Other attempts to increase accuracy have been for chatbots to view situations differently. Yann Lacunn, the lead A.I. scientist at Meta, believes that we should use “world modeling.” That means letting chatbots better understand the world. This would help chatbots to have better reasoning and common sense. Another approach would be to add on to the Large Language Models and later create Artificial General Intelligence, or A.G.I.. This new form of technology would be able to do everything the human brain can do.
Sources:
https://www.nytimes.com/2024/07/23/technology/ai-chatbots-chatgpt-math.html