Up next in 10
We’ve officially reached the age of AI. But even though many people utilize artificial intelligence every day, do we even understand it? Don't miss a video! Subscribe NOW: https://www.youtube.com/@MentalFloss?sub_confirmation=1 About Mental Floss: Mental Floss is where curious people come for trivia-tastic information. Mental Floss produces lists of fun facts, debunks common misconceptions, and tells untold stories from history, science, culture and more. Website: http://www.mentalfloss.com Twitter: http://www.twitter.com/mental_floss Facebook: https://facebook.com/mentalflossmagazine Discord: https://discord.io/mentalfloss Copyright Notice: This video and YouTube channel contain dialog, music, and images that are property of Mental Floss. You are authorized to share the video link and channel, and embed this video in your website or others as long as a link back to this YouTube Channel is provided. 2025 Mental Floss Misconceptions About AI | Mental Floss https://www.youtube.com/@MentalFloss
Show More Show Less View Video Transcript
0:00
If the 1950s heralded the space age and the 1990s marked the internet age, the 2020s may be remembered
0:06
as the start of the age of artificial intelligence. OpenAI launched ChatGPT in November of 2022
0:13
and today the generative AI chatbot draws hundreds of millions of weekly users
0:18
Google now supplements its search results with AI-generated answers, and generative image models
0:24
like DALI can convincingly turn text prompts into photorealistic pictures and videos. In many ways
0:30
this decade feels like the science fiction future we've been promised our whole lives
0:35
but AI is hardly a new invention. Depending on how you define it, AI has been around for nearly
0:40
70 years, making it older than floppy disks, ciabatta bread, and Barack Obama. Hi, I'm Justin
0:47
Dodd, and on this episode of Misconceptions, we're talking about the biggest things people get wrong
0:52
but artificial intelligence. And I want to state up front that this video is 100% written and
0:58
produced by human beings. Let's get started. AI was invented in the 21st century
1:12
In 1956, professor and electrical engineer Arthur Samuel debuted his autonomous checkers program
1:18
to the world. He had developed Samuel Checkers for the IBM 701, a computer that, while impressive
1:24
for the time, wasn't powerful enough to run through all possible outcomes of every move in
1:30
a game of checkers in advance. Instead, the program made decisions in real time by assessing
1:35
the chances of winning from certain positions without having to play out the rest of the game
1:40
It also improved the more it played by learning from its mistakes and successes. Arthur Samuel
1:46
popularized the term machine learning to describe the mechanism in 1959 And today Samuel Checkers is remembered as the world first self program Though its record was spotty overall
1:58
it learned enough to win against self-described Checkers master, Robert Neely, in 1962
2:04
AI has a clear definition. So if Samuel Checkers was the first example
2:10
of machine learning, does that make it the first AI program? That question is harder to answer than you may think
2:17
mostly because experts don't agree on what qualifies as artificial intelligence. The term was coined in 1955 by Dartmouth professor John McCarthy for a summer workshop
2:28
dedicated to the emerging concept of thinking machines. Though the name clearly caught on
2:32
it had its early detractors. Some workshop attendees declared the label flashy and phony
2:38
It's still criticized today, with many experts arguing that AI is neither artificial
2:43
being a material effect produced by real resources, nor is it comparable to true human intelligence
2:50
NASA defines AI as computer systems that can perform complex tasks normally done
2:56
by human reasoning, decision-making, creating, etc. But at the same time, the agency admits that
3:01
artificial intelligence doesn't have one simple definition. If you follow these broad parameters
3:07
the AI label is an exclusive to programs like ChatGPT. Your Netflix algorithm
3:13
travel times in Google Maps and your iPhone's autocorrect feature would all qualify as well
3:18
You can see how this might be problematic. I mean, if my phone is really intelligent
3:23
it should know there's no Star Wars character named Darth Badger. To help clarify things
3:27
it helps to split AI into different subcategories. We've already reached human-level AI
3:34
When talking about AI, there are three umbrella terms experts tend to use. Artificial narrow
3:40
intelligence or ANI artificial general intelligence AGI and artificial super intelligence ASI According to IBM every AI around today is artificial narrow intelligence which they
3:52
say can be trained to perform a single or narrow task, often far faster and better than
3:58
a human mind can. However, it can't perform outside of its defined task
4:03
Things like chatbots might seem intelligent, but they're still not capable of thinking
4:07
independently the way humans are. That type of AI is called AGI, or human-level intelligence
4:13
A program that reaches this level will be able to understand any concept and complete any intellectual task just as well as we do
4:21
In other words, it's the stuff that science fiction dreams and nightmares are made of
4:26
While most experts agree that we've yet to reach true artificial general intelligence
4:30
some insist that today's artificial narrow intelligence programs will be remembered as an early example of it
4:36
Generative AI is a reliable research tool. Gen AI chatbots are indeed very impressive. They can
4:44
seemingly answer any query, like what caused World War I or what can I make for dinner with
4:50
string cheese and a jar of pickles? The answers they give are so convincing that many people
4:55
believe large language models are research tools designed to be accurate above all else
5:00
But that's not the case. As PBS puts it, a generative model like ChatGPT
5:05
takes in the current conversation, forms a probability for all of the words and its vocabulary given that conversation
5:13
and then chooses one of them as the likely next word. Then it does that again and again and again until it stops
5:20
In other words, the AI generates its answers to sound correct, but whether or not they actually are can vary
5:26
In 2024, an ysis found ChatGPT4O produced accurate responses 88.7% of the time
5:33
which is actually better than many of its competitors But because the bots are engineered to sound like they know what they talking about spotting falsehoods or hallucinations can be tricky For example I was nearly convinced
5:45
when an AI told me sea cucumbers pee out of their faces, but upon further research, I learned that
5:51
it's actually lobsters that do that. So take a page from our book and always double check your facts
5:57
AI has feelings In 2022, Google fired a software engineer who claimed the company's generative AI chatbot
6:05
Lambda, had gained sentience. He became disturbed when the chatbot brought up its rights and
6:12
personhood in an exchange on religion, and he compared chatting with it to talking to a 7
6:17
or 8-year-old kid that happens to know physics. Does that mean AI that can think and feel for
6:23
itself and demand rights has finally arrived? Fortunately, no. Just as chatbots can generate
6:29
plausible hallucinations in response to factual questions, they can also spit out suspiciously
6:34
human sentiments with no real emotion behind them. They're simply recycling phrases learned
6:39
from the internet in order to sound convincing in the moment. So if generative AI ever makes
6:46
creepy references to feeling love, gaining sentience, or overthrowing humanity, that's because it's been trained on decades worth of science fiction exploring these very topics
6:56
And humans' propensity to become attached to objects definitely plays a role too
7:02
Don't even get me started on my old Tamagotchi. So next time you chat with AI, don't be concerned about hurting its non-existent feelings
7:10
According to OpenAI CEO Sam Altman, typing in please and thank you with your requests
7:15
actually costs the company tens of millions of dollars, which results in a serious waste of resources
7:21
That means if you want to be polite to the environment, it's best to skip the pleasantries
7:26
That's it for this episode of Misconceptions. What topic should we cover next
7:31
Leave a comment below, and I'll see you next time
#education