A new study conducted by UCLA has proven that AI is more intelligent than humans. How? Researchers have found that the autoregressive language model “Generative Pre-trained Transformer-3 (GPT-3)” is the most compatible program for human intelligence.
UCLA researchers put it to the test by running a test to compare that knowledge to the average college student. The program passed all tests with flying colors. By incorporating deep learning, it can also generate human-like text. Amazingly, the program has 175 billion parameters, the largest of its kind.
So how did the program compete with human intelligence? It was tested using three fronts. First he had SAT scores, then general knowledge, and finally IQ. Humans could only achieve the 35th percentile on the aforementioned plane, but the program was able to secure the 80th percentile on all tests.
GPT-3 outperformed human participants in all question types in terms of both generation accuracy and multiple-choice accuracy. pic.twitter.com/KQge8cDQzh
— AI Breakfast (@AiBreakfast) December 27, 2022
GPT-3 performs complex tasks that are not easy for AI, such as induction of abstract patterns. Most sites use abstract patterns like this to filter out bots, but this development forces sites to come up with better ideas for keeping AI out of their territory.
This may have its drawbacks, but the benefits of having such a system outweigh the drawbacks.This program can be used to develop better technology as well as contribute to the Metaverse.
It’s this human-like process of acquiring knowledge that worries researchers a bit. GPT-3 is a program that forces information through a mechanism rather than training the system to learn. This can make it too stiff or confusing.
Experts also found that GPT-3 definitely excels when solving problems from scratch and when solving MCQs. This means that the chances of error are relatively low, and you may breathe a sigh of relief.
The only problem in developing such complex and advanced AI systems is the ideology of “robots taking over the world”. The theory seems highly unlikely, but with such new technology, I’m not sure it’s no longer achievable.
GPT-3 has made great strides in the world of technology, but we are still waiting for our research to be approved by our fellow pals so that GPT-3 can finally make its way into the world in earnest.
READ NEXT: Google Assistant, Siri, Alexa, and Bixby – Here’s a list of the world’s best and worst voice assistants