I came across a tweet today where an individual talks about how he believes chatGPT is much smarter than the average person but that he immediately stops reading something after realizing it’s written by ChatGPT.
This made me ponder for about 30 minutes because we often talk about how smart AIs are, just carelessly throwing the word out there without much reasoning.
Also, it made me think about “smart contracts” and how I've personally come to be a critic of these poor terminologies and subsequent use across the cryptocurrency ecosystem.
But before I digress further, this piece is about artificial intelligence, not crypto or blockchain, so let's put a pin on that and circle back to the original topic.
Is artificial intelligence(AIs) smart?
To figure this out, we must first understand what being smart means.
Ironically, I asked AI(ChatGPT) what being smart means, alongside other terms often used interchangeably.
Not because I think it'd have the best answer, but because I've used it enough to know that it comes up with the best “reason-like” points to back a statement. This generally gives you more insights into how the AI understands your questions and the data(information) it has access to.
Smart
Practical, adaptive problem-solving and quick thinking, often in real-world situations.
Smartness is often situational and about using what you know effectively, rather than knowing everything.
Intelligence
The capacity to acquire, understand, and apply knowledge and skills.
Intelligence is more of a raw capacity for understanding and processing information.
Brilliance
Exceptional talent or intellect, often in a specialized area.
Brilliance is rare, intense, and often specialized. It dazzles.
Bottom line
You might be smart in navigating life.
You might be intelligent in understanding complex ideas.
You might be brilliant in transforming how others think.
I also straight up asked ChatGPT if AI is smart and this was the conclusion:
AI is not inherently smart. It can appear smart because it mimics aspects of human intelligence, but it lacks understanding, reasoning, and general adaptability that define true intelligence.
Notice how it ends with “true intelligence” not “smarts.”
The only logical reason this would happen is because these things overlap, even though they carry different connotations and subtleties.
Notwithstanding, when you look at the provided definition of intelligence and smart, you can tell that one is a very specific form(application generally) of the other. That said, when you do the same with brilliance and intelligence, you get a similar conclusion.
To spotlight my conclusion:
Intelligence would be the capacity of both the smartness and brilliance of someone.
Essentially:
— Being smart is a form of intelligence.
— Being brilliant is also a form of intelligence.
The difference is generally the degree of the “wow factor” when experiencing both and that supports the argument that intelligence is just the capacity.
If someone makes a smart move, you're impressed and you express this, but it's generally a low-leaning expression. But when someone does something absolutely brilliant, well, just picture anyone saying the word “brilliant” when amazed and you can immediately see how different the expression of impression is.
Now, what does this tell us about AI?
Smart or just efficient?
Remember that the AI, ChatGPT, happens to have said that being smart is generally situational and about using what you know effectively, rather than knowing everything.
Also, it's important to note that the “I” in AI stands for intelligence, so even though the ChatGPT doesn't think most AIs are smart, it's very definition of these terms highlights that AI in fact acquires, understands(learns by training) and applies knowledge, and most times situationally and at speed, effectively qualifying as a smart system.
So to address the question, it's a flawed question considering the context because being smart here translates to efficiency, which would mean that it is both.
It's not philosophically smart, it technically has no memory, it just sticks to maths, but these processes still afford it the basic definition of smart, even though it inherently doesn't care about being right, just sticking to the training data. If the training data affords it the ability to want to be right, then it will adjust accordingly.
So in essence, it's all only as smart as the data it has access to.
Now I lack any deeper knowledge of how these systems work beyond what's made available, so I'd say that by what's known, it pretty much does what it's programmed to do: mimicking human intelligence and effectively solving problems more efficiently.
Since it does that right, nothing else really matters.