Most likely: yes. And you’ll find out why in this very post!
Wait, don’t leave! This is a four-part series on how we can use artificial intelligence to our advantage!
Part 1 of this series, on Human Intelligence vs. Artificial Intelligence, is here.
Part 2, on Artificial Intelligence, Machine Learning and Generative AI, is here.
This is part 3, Is Artificial Intelligence Smarter Than I Am?
Part 4, How Can I Master Artificial Intelligence? is here!
Table of Contents
Part 3: So… Is Artificial Intelligence Smarter Than I Am?
Sorry, but the answer is almost definitely yes. In the ways we are able to measure its intelligence…. yes.
As your fellow human I’d really, really love to deny this. But I can’t. Here’s why.
We measure human intelligence with intelligence tests. The WAIS is the most commonly used test for measuring human intelligence.1 This test has been scientifically validated and used for many, many years.
In March of 2023, a psychologist wrote in Scientific American that he gave ChatGPT the WAIS-III2 — not the most recent version of the WAIS, which is up to version IV now, but reliable nonetheless. He tested the GPT 4 engine, which is the most recent architecture of ChatGPT and currently one of the most advanced Gen AIs.
He administered five of the 11 subtests of the WAIS-III. Some of the WAIS subtests are non-verbal, which makes them impossible for GPT 4 to take via chat input, and one of the subtests analyzes short-term memory and recall, which isn’t a useful test for GPT 4 given its lack of organic brain material.
Based on the machine’s performance on the five subtests he could administer, he reported GPT 4 scored 155 … out of a possible score of 155.
It’s admittedly an imperfect measurement as he couldn’t use six of the 11 subtests, but that’s somewhat beside the point. Here’s the point:
On the tests it could take, GPT 4 tested as more intelligent than 99.9% of all humans.
OK, but can we really compare artificial intelligence to human intelligence?
Well, artificial intelligence is exactly that: artificial. It’s programmed to respond probabilistically. We may not understand what exactly caused an AI to formulate a given response but we know it’s calculating something, or following some algorithm, to predict a response that will be successful. It’s a computer, after all. AI responds programmatically, based on a calculated prediction of success, not by thinking like a human does.
But couldn’t we also say that humans are programmed from birth to give answers and act in ways that have a high probability of success? We learn from experience what actions are socially appropriate and what answers are “correct,” and we refine our ability to perform in ways we predict will be appropriate. I’ve learned to chew with my mouth closed, to smile when I meet someone, to not talk when someone else is talking. I’ve learned how to give correct answers on tests by predicting that my answer will sufficiently address the question. I’ve learned how to draw (sort of) by predicting how to create an image that will be successful in meeting my goals.
Those things are programming too!
OK, but doesn’t artificial intelligence make mistakes?
Yes, absolutely!
It can “hallucinate,” such as in the NY Times report about Bing falling deeply in love (verging on harrassment) with the reporter.3
It can invent information, such as the recent example of the lawyer who was sanctioned for using ChatGPT to create a legal brief after it became clear the system had invented cases to back up its arguments.4
It can make basic arithmetic errors and misrepresent data. (To be fair, this is because Gen AI is a prediction engine and not a calculator. Its mathematical predictions are just not as accurate as calculations. And it’s just a matter of time before someone writes in a calculation engine as part of a Gen AI system so that its math becomes accurate. But for now it can be surprisingly wrong on easy math!)5
It can fail logic tests. If you ask ChatGPT “What is the first name of the father of Sebastian’s children?” (to which the answer works out to “Sebastian”) ChatGPT flounders.6 (However, if you preface the prompt with “Here’s a riddle:” it will answer correctly … and much faster than a human!)
So is ChatGPT always right?
Definitely not.
But 100% of humans are wrong some of the time.
Is ChatGPT intelligent?
Definitely.
If we measure its intelligence the way we measure a human’s intelligence, it’s smarter than 99.9% of us.
Here’s the best way I can explain it: Generative AI is not intelligent like a human, but it is intelligent. And in some important ways it can perform more intelligently than almost every human.
One important note on this: intelligence is not the same as consciousness. Intelligence is a measurement of the quality of output. Consciousness is a completely different thing: is something self-aware. Neither is necessary for the other to exist. Hoo, boy, this could be another, way longer series! But if you’re interested in a couple takes on this topic, go here and here once you’ve finished reading this!
Another important note: IQ is far from the only way we should measure the quality of a mind, and intelligence isn’t the only thing that makes us human. EQ, motivation, resilience, drive and plenty of other human qualities are just as important (perhaps more important) as IQ.
But still: when it comes to a strict definition of intelligence, the machines already have us beat.
Fine. I welcome my artificially intelligent overlords.
Not so fast! Artificial intelligence is a tool for humans, not a replacement for humans. We must master it so it doesn’t master us.
Allowing an artificially intelligent system to make decisions that influence humans, when it is not human and therefore can’t fully understand human life, could become exceptionally dangerous. Even existentially dangerous.
We must not put our heads in the sand on this one.
We must learn to use AI as a tool.
We must learn how control AI and validate its responses.
And ultimately, as a species, we must shape AI to our needs.
Because if we don’t, AI will shape us.
And we may not like how that goes.
Tune in for part 4 next week and learn how to master AI and supercharge your work!
Want to know how I can help you or your team learn? Go here!
Need more great AI thoughts right away? Our friends at BrandSwan are thinking about “How AI Tools Are Weeding Out Mediocrity”
Footnotes
- https://www.pearsonassessments.com/store/usassessments/en/Store/Professional-Assessments/Cognition-%26-Neuro/Wechsler-Adult-Intelligence-Scale-|-Fourth-Edition/p/100000392.html ↩︎
- https://www.scientificamerican.com/article/i-gave-chatgpt-an-iq-test-heres-what-i-discovered/ ↩︎
- https://www.nytimes.com/2023/02/16/technology/bing-chatbot-microsoft-chatgpt.html?unlocked_article_code=PMy4FLYjgPAGN7bxJR7ZDSdpGZxq4u3acocTrb0zFSVER93Yop00vbqnnRwPqgm7GMG5zduyKZ8fT2tSYD05fDUQIW3l5Bk6kXIhoZp_4yDsexgRE1TiChnoXzxmtyNt5aNbzkoop6LZTI3v7INV-EBpxm2syBydtis2-mic3B1K9Z2FIlo5qGiOq6iKd67tPc8uR61TYzVMUIbsCAd-RtmQF9t7FuCWSxq7EKQ51UTLxKMLd5GgzkAavl9Ll6sp_xT4WN6xnfINe9IO29V4wCDIvECOvLZ_052LNYSueUkc-RwZzasTFKzz0-HuQrpTAfjqxZkKTica8y4Skk8NR-_–qKRYDgbI-pN&smid=url-share ↩︎
- https://www.reuters.com/legal/new-york-lawyers-sanctioned-using-fake-chatgpt-cases-legal-brief-2023-06-22/ ↩︎
- https://community.openai.com/t/chatgpt-simple-math-calculation-mistake/62780 ↩︎
- https://www.scientificamerican.com/article/i-gave-chatgpt-an-iq-test-heres-what-i-discovered/ ↩︎
Leave a Reply