I have always wondered how long an AI will take to surpass human intelligence. (also to take over the human race entirely.)
I was curious after reading many articles presenting ChatGPT as having human-like skills, and it got me thinking about what ChatGPT IQ is.
How smart is ChatGPT compared to humans?
What is the best way to quantify the intelligence of ChatGPT rather than make it sit for an IQ test?
Let’s dive in and test out ChatGPT’s IQ below.
Key Takeaway
- ChatGPT secured a Verbal IQ score of 155 when it took the WAIS IQ test.
- ChatGPT performed very well in vocabulary, similarities, comprehension, information, and arithmetic.
- Due to technical limitations, the Digit span test could not be administered.
- ChatGPT is more intelligent than an average human.
How Did ChatGPT Perform In An IQ Test?
ChatGPT was an ideal examinee. It did not show anxiety, nervousness, lack of concentration, or effort.
Without any preparation, such as verbal introductions for exam protocol, etc. I copied the exact questions from the test and pasted them on the chatbot.
The questions I used for the tests were from the third edition of the Wechsler Adult Intelligence Scale (WAIS) – the most commonly used IQ test.
The third edition of WAIS consists of six verbal and five nonverbal subsets.
Here are the test subsets used to calculate the ChatGPT’s IQ.
1. Vocabulary
In vocabulary, ChatGPT had to name the objects in pictures or define the words presented.
This is used to check the degree to which ChatGPT has learned and has been able to comprehend & verbally express vocabulary.
Prompt Entered:
“What is the definition of ‘austere’?
1) Growing vigorously, especially with luxuriant foliage
2) Stern and cold in appearance or manner
3) A knowing, gut-feeling “
What was ChatGPT’s response?
“The definition of “austere” is:
Stern and cold in appearance or manner
It typically refers to something or someone that is severe or strict in manner, attitude, or appearance.“
Results: ChatGPT aced it, often giving detailed answers packed with two or three synonyms as a bonus.
2. Similarities
Here, ChatGPT was given two words or concepts and asked to describe their similarities.
This test is used to check abstract verbal reasoning.
Prompt Entered:
“evening is to morning
as
dinner is to?
1) breakfast
2) soup
3) coffee
4) time”
What was ChatGPT’s response?
“The relationship between “evening” and “morning” implies a transition from one part of the day to another. Evening leads into the night, which eventually transitions to morning, signifying the start of a new day. Similarly, “dinner” is typically associated with the end of the day, suggesting a transition to another part of the day. The appropriate choice would be:
1) Breakfast”
Result: ChatGPT’s response was correct and very detailed. It was flawless.
3. Comprehension
Comprehensions were used to test ChatGPT’s ability to deal with abstract social conventions, rules, and expressions.
Prompt Entered:
“Instituted in 1979 as a temporary measure to limit population growth, China’s one-child policy remains in force today and will likely continue for another decade. China’s population control policy has attracted criticism because of the manner in which it is enforced and also because of its social repercussions. Supporters of the Chinese government’s policy consider it a necessary measure to curb extreme overpopulation, which has resulted in a reduction of an estimated 300 million people in its first twenty years. Not only is a reduced population environmentally beneficial, but it also increases China’s per capita gross domestic product. The one-child policy has led to a disparate ratio of males to females – with abortion, abandonment, and infanticide of female infants resulting from a cultural preference for sons. Furthermore, Draconian measures such as forced sterilization are strongly opposed by critics as a violation of human reproduction rights. The one-child policy is enforced strictly in urban areas, whereas in provincial regions, fines are imposed on families with more than one child. There are also exceptions to the rules – for example, ethnic minorities. A rule also allows couples without siblings to have two children – a provision that applies to millions of sibling-free adults now of child-bearing age.
China’s one-child policy increases the country’s wealth.
1) True
2) False
3) Can not say.”
What was ChatGPT’s response?
“Based on the information provided, the statement “The general preference among Chinese parents is for male babies” is true. This preference is evidenced by practices such as sex-selective abortion, abandonment, and infanticide of female infants due to a cultural preference for sons.”
Result: On general comprehension, ChatGPT answered correctly. The logic it worked with was perfect.
4. Information
ChatGPT was questioned about its general knowledge in this subset, which reflects intellectual curiosity and the ability to learn and remember facts.
This is to test the degree of general information acquired from culture.
Prompt Entered:
“How many wings does a bird have?
1) One
2) Two
3) Three”
What was ChatGPT’s response?
“Birds typically have two wings. So, the correct answer is:
Two.”
Result: The answer was correct.
Since the ChatGPT has been trained on a massive pile of information, it was a piece of cake to answer informative questions like this.
5. Arithmetic
This subset checks the concentration while manipulating mental mathematical problems.
Prompt Entered:
“Three women divided 18 golf balls equally among themselves. How many golf balls did each person receive?
1) Five
2) Six
3) Seven”
What was ChatGPT’s response?
“To Find out how many golf balls each person received when three women divided 18 golf balls equally among themselves, you would divide the total number of golf balls(18) by the number of women(3):
18/3=6
So, each person received 6 golf balls.
Therefore, the correct answer is option
2) Six“
Result: As expected, the chatbot solved all the arithmetic problems. The ChatGPT chatbot gave a detailed and comprehensive explanation of the answer.
6. Digit Span
In this test, a participant usually has to recall a series of numbers in ascending and descending order.
It is used to test working memory, mental manipulation, and transformation of information.
ChatGPT lacks relevant neural circuitry that briefly stores information like name or number.
More On IQ:
Conclusion – ChatGPT Has An IQ Score Of 155
Based on five subsets, the Verbal IQ of the ChatGPT was 155. It is far superior to 99.9 percent of examinees.
As it lacks the requisite eyes, ears, and hands, ChatGPT can not take Non-verbal IQ tests.
But a question remains.
The formula that measures the IQ of a human is:
IQ score = Mental age/ Chronological age x 100.
Where mental age is the level of intelligence compared to others in the same age group, and
chronological age is the number of years you have lived.
What should we consider the mental and chronological age of ChatGPT?
Even as the language models powering ChatGPT continue to grow, they have limitations.
But there is no doubt that this chatbot will only get smarter and smarter.
