ChatGPT reveals strengths and weaknesses when put to the test to complete a university Physics degree
Credit:Shutterstock:egaranugragh
Two researchers tested ChatGPT‘s ability to complete a university-level Physics degree, revealing both impressive insights and significant limitations in AI’s academic capabilities.
What happens when you pit OpenAI’s ChatGPT against a university-level Physics degree?
Two researchers, L.J. Morrell and K.A. Pimbblet, decided to find out. In a quirky yet revealing experiment published in the ‘European Journal of Physics’, they tested whether the AI could graduate in Physics and, if so, with what kind of grades. The results? A fascinating blend of brilliance and blunders. While ChatGPT showcased moments of scientific savvy, it also exposed the limits of current AI technology. This unconventional challenge sheds light on the exciting potential—and glaring gaps—of AI in tackling complex human disciplines. Could an AI ace your toughest college courses? Read on to find out!
ChatGPT excelled in subjects like programming and maths calculations
The experiment followed an approach known as “maximum intelligence,” showcasing the model’s optimum capacities. This ranged from simplifying questions, breaking more complex tasks into smaller chunks and providing detailed prompts. In this ambience, ChatGPT fared well, achieiving exceptional results in subjects including Numerical Modeling and Simulation, and shining in programming and mathematical calculations.
However, artificial intelligence fell short in some areas which required more practical skills. In laboratory modules, inability to perform experimental measurements and the need to manually input data made ChatGPT’s expertise more doubtful. Just goes to show that human skills are still needed in areas where physical interaction is essential!
Chat GPT capable of achieving a 2:1 degree but fails at abstract reasoning
What about written theory? We all know ChatGPT is a whizz at writing, yet it didn’t make the top spot when it came to written assessments, scoring an average of 65 per cent – an upper second class degree in the British system. This is obviously not bad for a machine, but clearly AI can only work its real magic so far in scenarios where there are clear concepts and predefined solutions. Abstract reasoning on the other hand is not ChatGPT’s strongpoint, nor could it cope with advanced reasoning like the calculation of the gravitational acceleration on the International Space Station.
In the grand experiment of ‘AI vs. Academia’, ChatGPT passed the Physics 101 hurdles with a solid 2:1 degree-worthy performance. But when the equations got twisty and the concepts turned cosmic, our friendly AI found itself stumped, proving it’s no Einstein—yet. So while ChatGPT might not ace abstract reasoning or crack the secrets of the universe, it’s still the study buddy you’d want for tackling the bulk of a physics syllabus. Who knows? With some upgrades, it might just graduate to star pupil someday!
Read more about Technology
