ChatGPT’s ‘limited’ intelligence

ChatGPT's limited intelligence - Photo 1.

ChatGPT is considered a smart AI with a huge amount of knowledge – Photo: decoder

ChatGPT is a large language model trained by the American artificial intelligence development company OpenAI, with capital from technology billionaires Elon Musk and Microsoft. It is considered as one of the intelligent automatic response systems (chatbots) with a “huge amount of knowledge”, the ability to chat and create content that can make writers unemployed or make schools worry. afraid because students ask to write essays for them… Many people even think that this new generation AI can replace humans in many fields.

ChatGPT is very smart

Many experts who have tested ChatGPT are amazed. Researcher Omar Al-Ubaydli asked this chatbot to “write a poem about electrical engineering in the style of Wilfred Owen” and was amazed by the excellent response.

One executive officer (CEO) even treats ChatGPT like an employee. Jeff Maggioncalda, executive director of online course provider Coursera, told CBS that he asked the AI ​​to check for blind spots and biases in his thinking and got the answers “very, very good”.

Mr. Maggioncalda said the tool helps me think more clearly and see the problem from the vantage point. For example, while attending the World Economic Forum in Davos (Switzerland), Maggioncalda asked ChatGPT: “What should I consider when addressing the prime ministers in Davos?”.

Not only ChatGPT, about 30% of professors in the US say they have used AI in their work and industry experts also rate AI as a “game-changing” innovation. Unfortunately, the administrative and writing skills of chatbots are threatening semi-professional jobs such as composing letters, documents, writing basic copy…

Mihir Shukla, founder of AI and robotics company Automation Anywhere, predicted at Davos that “between 15-70% of all the work we do in front of computers can be automated”.

Columbia Business School professor Oded Netzer, an expert in text mining techniques, says that ChatGPT is a revolutionary step forward in artificial intelligence. “It’s really an incredible leap in technology and innovation,” he told CBS.

ChatGPT's limited intelligence - Photo 2.

ChatGPT can do basic administrative tasks well – Photo:

…But Smart ChatGPT “has limits”

CNBC writer Ryan Ermey recently experimented with ChatGPT to write an article on a financial topic to compare with his recent article. As a result, Mr. Ermey was quite surprised that ChatGPT only took a few seconds to write a complete article with basic information, although a bit “mechanical”.

But when Mr. Ermey asked ChatGPT more about the relevant information needed to include in the article, the AI ​​started to “stuck”.

“AI can be a powerful tool, its users still have to know their content to get really useful results for the reader,” he commented.

Ermey asked ChatGPT if he wanted to do his job, and this chatbot replied: “Artificial intelligence has not developed enough to completely replace human journalists in tasks that require creative thinking, analysis. critical thinking as well as the ability to understand and explain complex social and political issues”.

Yann LeCun, Meta’s artificial intelligence scientist, also admits the company’s Blenderbot AI is “boring” than ChatGPT but safer. According to Mr. LeCun, Blenderbot will be more cautious in religious matters, but ChatGPT can comfortably chat about misconceptions in the Quran, write a prayer for a rabbi to send to Congress. ..

Writer Wayne MacPhail also commented in the newspaper The Global and Mail that ChatGPT can make us think it thinks like a human but it can’t make sense of the world.

“ChatGPT is not thinking at all – and certainly not thinking like a human. What it’s doing is searching, at breakneck speed, through the trillions of linguistic connections it makes by scanning rows. mountains of human-generated content,” MacPhail wrote.

Worse, it doesn’t even care whether the answer it provides is correct. “I know this because I asked ChatGPT so,” he said.

“ChatGPT is extremely limited, but good enough at a few things to give the wrong impression of how awesome it is,” OpenAI executive director Sam Altman said on Twitter in December 2022, saying “it’s true. erroneous” when relying on AI on important issues.

Should ChatGPT be banned in schools?

Some schools were quick to ban ChatGPT because of its ability to write essays, solve math, science, and even write computer code. This alarmed the educators. However, it is also suggested that it would be a great educational tool, not to mention banning it would be ineffective. Cherie Shields, a high school teacher in Oregon (USA), had students use ChatGPT to frame their essays and then write them directly in class.

“Any tool that allows students to filter their thoughts before class and practice ideas will only enrich the discussions,” history teacher Jon Gold said in the newspaper. New York Times.

OpenAI also recently announced that it has created a tool to help distinguish human-written content and ChatGPT-generated content. However, this tool is still not accurate, so far only recognizing about 26% of content written by AI.

Leave a Comment