ChatGPT is an innovative technology that has taken the world by storm. Many tech enthusiasts described this Artificial Intelligence software as the most disruptive technology of the century. While ChatGPT seems capable of incredible feats of intelligence, it also has remarkable lapses and hallucinations. There are plenty of limitations since ChatGPT is still in the pilot process. So far, the main question is how this well-sophisticated technology can change the world for businesses and societies, and it is clear that we are far away from having an answer.

Building upon the successes of its predecessor, ChatGPT takes AI capabilities to a whole new level. It possesses an enhanced understanding of context, improved reasoning abilities, and a broader knowledge base. But what makes ChatGPT so different is that the user’s conversation with this AI seems more natural and human-like.

Based on what we know so far, ChatGPT can perform human-level in many professional and academic exams. ChatGPT has made impressive strides over its predecessors but continues to struggle in certain subjects. Some users also report that ChatGPT shares fake news and misleading information or that its responses need to be more accurate.

Is AI truly intelligent?

There have already been several extensive discussions about whether Artificial Intelligence is genuinely… intelligent. Recently, in Finland, the much-discussed ChatGPT took an IQ test. Usually, these types of tests are intended for adults. The researchers, therefore, used the tools that assess the intelligence of adult humans to measure the IQ of the robot. Not so absurd, considering how many articles have been written extolling the “smartness” of ChatGPT.

Psychologist Eka Roivainen doesn’t think a chatbot’s excellent text capabilities are necessarily a sign of intelligence. Therefore, she believes only an IQ test could allow us to reconstruct our opinion.

During the test, ChatGPT did not experience stress or other concentration problems. In humans, these emotional factors can influence the outcome of an IQ test, indeed, any difficulty. In addition, Eka Roivainen chose the 3rd edition of the Wechsler Adult Intelligence Scale (WAIS) dating from 1997, which is the benchmark for measuring the intelligence of adults and youth over 16 years of age.

The test includes 11 basic subtests divided into verbal and performance scales. However, while artificial intelligence is often compared to human intelligence, the two are very different and cannot be trained similarly. Thus, the ChatGPT was tested on five verbal subtests: vocabulary, comprehension, information, arithmetic, and similarities. However, the last secondary test, digital memory, was not placed on the machine.

How smart is ChatGPT?

According to Eka Roivainen, ChatGPT scored 155 on verbal IQ. That performance is higher than 99.9% of the 2,450 people who took the same test in the study. As for the other tests, the scores were also very high. Indeed, an IQ test can be considered a reliable way to assess a person’s intelligence. Does the same apply to machines?

Whether ChatGPT is smart was also the focus of the Dublin Tech Summit. A group of global experts discussed conversational AI to explore what this technology is capable of, its current limitations, and its possible future. Adam Cheyer, Dr Patricia Scanlon, Lasse Rouhianen, and Dr Oisín Boydell shared their insights on this controversial topic.