An IQ Showdown Between ChatGPT and A Data Scientist — Whose Intelligence Reigns Supreme?
I discovered some insights and surprises about artificial intelligence and human intelligence when I experimented to compare ChatGPT’s intelligence with mine.
Table of Contents
• Understanding IQ Tests: What Do They Measure?
• Setting Baselines and Experimental Procedures
• Analyzing ChatGPT’s Test Performance: Confidence and Accuracy
• Insights into ChatGPT’s Intelligence: Strengths and Weaknesses
• • • IQ’s Real Value
• Appendix
I’ve stumbled upon these goofy YouTube videos in my recommendations where folks rate each other’s smarts and then dive into IQ tests.
This got the data nerd in me thinking about where would ChatGPT land on this scale, and, more crucially, could ChatGPT outsmart me?
Honestly, I really don’t know anything about IQ tests. From what I’ve seen, these tests typically involve patterns and possibly shapes in a sequence, and your job is to choose which shape or pattern comes next.
Understanding IQ Tests: What Do They Measure?
Before we dig into the results, let’s see what these IQ tests are supposed to measure.
According to ChatGPT itself, an IQ (intelligence quotient) test measures a person’s cognitive abilities in relation to their age group.
More precisely, it’s designed to assess certain mental faculties, such as:
- Logical reasoning, which is the ability to solve problems through deduction.
- Mathematical ability, which comprises skills related to numerical problems.
- Spatial visualization, which refers to the ability to understand and manipulate shapes and spaces.
- Language skills, which encompass vocabulary, comprehension, and verbal reasoning abilities.
- Memory, referring to the recall of information after a given period.
These all seem like things that ChatGPT should handle well.
Setting Baselines and Experimental Procedures
To provide context, I’ll establish two baselines for comparison with ChatGPT’s IQ: the average IQ, which stands at 100, and my own performance.
To achieve this, I’ll be taking the IQ tests from Brainable and IQ Metaspace (linked in the appendix for your convenience) before inviting ChatGPT to take the same tests and answer the identical questions.
Just a heads-up, the Brainable IQ test sparked some skepticism in me due to its request for sign-ups and payments, which led me to doubt its score accuracy.
Consequently, I opted to take the IQ Metaspace test as well to double check and gather more comprehensive data.
Brainable also offered brain-training services. From a logical standpoint, it seemed plausible that their aim might be to downplay scores to foster a sense of needing improvement and to encourage further use of their services.
You know, pitting myself against ChatGPT and other people can be very humbling, and I’m a little nervous about what I’m going to find out here. I hope I won’t attach too much of my self-worth to my IQ scores on these tests.
Alright, without further ado, let’s jump in!
Analyzing ChatGPT’s Test Performance: Confidence and Accuracy
ChatGPT answered confidently — got to hand it that! In face, it was impressively confident about every answer that it gave.
However, during the process, I became increasingly convinced that many of its responses were incorrect.
It might be a little concerning that ChatGPT is so assured in the face of wrong answers.
ChatGPT also seemed to do better on questions where it explained its answers, which is consistent with the emerging academic research on how to get the most out of these large language models.
Comparing Our IQ Scores
The results of the Brainable test were somewhat unexpected. ChatGPT ended up with an IQ of 102, which is better than average.
But was it better than me?
I actually just edged it out by a bit with a score of 116.
In regard to the IQ Metaspace IQ test that we took, I absolutely crushed it. I achieved a whopping score of 138. You might as well start calling me Ken Genius now.
Could ChatGPT measure up to the benchmark I set? Interestingly, it fell short by a gigantic margin. ChatGPT scored 51 on this IQ test, which is also below-average compared to the initial benchmark.
Also, we can never have enough claps for the Medium algorithm. So if you find this article useful, feel free to drop some.
Insights into ChatGPT’s Intelligence: Strengths and Weaknesses
Given the above performance differences between me and ChatGPT across the two IQ tests, what could possibly account for this perceivable disparity?
The cool thing about all these tests is how they categorize all the questions based on the specific types of intelligence they assess. This breakdown helps you can identify which areas are your strengths and weaknesses.
Now, let’s analyze the aspects in which ChatGPT triumphed and struggled.
In the Brainable test, ChatGPT excelled in pattern recognition verbal intelligence, while performing reasonably well in mathematical ability. However, it struggled the most with spatial reasoning, an area where I excelled the most.
The IQ Metaspace test was largely focused on spatial reasoning. I think that’s probably why I did particularly well in this test compared to the Brainable test.
ChatGPT really performed quite well at numerical reasoning within this IQ test, yet encountered difficulties in all other evaluated areas.
As a human, I believe my abilities are a little bit more well-rounded than ChatGPT’s, considering how ChatGPT sharply excelled or failed miserably in specific areas.
Lessons Learned
More important than the scores were the lessons derived from doing this little exercise.
I think ChatGPT still has a long way to go for interpreting visual information.
Currently, it might be more practical to use it for math and logical reasoning, considering these are the areas where it performed most strongly.
Secondly, I must admit riding pretty high on my score from the IQ Metaspace test. Getting this high IQ number, which would fall into what could be termed a gifted category, made me feel quite good about myself.
That is until I actually looked up the research around IQ.
IQ’s Real Value
What I discovered is that IQ is not directly correlated with success in work, life, or even happiness (Cherry, 2022).
Actually, high-IQ people are more likely to be less happy overall (Eren et al., 2018).
Essentially, IQ tests are really good at evaluating how good you are at taking IQ tests.
Excelling in these tests isn’t a prerequisite for excelling in any career, particularly in fields like data science.”
Additionally, I learned something interesting from my younger days. Possibly during middle school, my parents had me undergo some diagnostic tests to understand my learning ability, and they mentioned that my IQ was nearly average.
Now, as I’ve grown and developed, my IQ score has actually gone up.
To me, that’s a function of my interest in solving these types of problems, my patience for sitting through an exam like this.
Moreover, I don’t view IQ as something that’s static over time. Therefore, I wouldn’t put a whole lot of faith in this number.
If you like solving problems and doing some of these quirky tasks in your free time, you’ll probably do well on IQ tests.
Again, it has no bearing on your overall success.
So, I hope you found this interesting and learned about ChatGPT’s intellectual capabilities as well as the limitations of IQ as a metric.
Until next time, good luck on your data science journey.
References
- Cherry, K. (2022). Are High IQ People More Successful? Verywell Mind. https://www.verywellmind.com/are-people-with-high-iqs-more-successful-2795280
- Eren, F., Çete, A. O., Avcil, S., & Baykara, B. (2018). Emotional and Behavioral Characteristics of Gifted Children and Their Families. Neuropsychiatry 55(2), 105–112. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6060660/
Appendix
- Click on these links, if you would like to try the Brainable IQ test or the IQ Metaspace IQ test.
If you enjoyed this article, remember to follow me on Medium for more content like this or subscribe to me via email. You can additionally share and recommend this article to your network that’s interested in data science!
If you like fun, informative videos on data science, machine learning, and AI, check out my YouTube channel, where I provide commentary, tutorials, and other educational videos.
If you interested in the unique stories of the people in the data and AI space about how they have approached the big decisions in their life that shaped their worldview and career, check out my podcast Ken’s Nearest Neighbors.
To get weekly updates on my content creation and on additional learning resources in the data science industry, sign up for my newsletter, the Data Dribble!
