How do you know that you know what you know?

How do you know that you know what you know?

Watch over 2,400 documentaries for free for 30 days AND get a free Nebula account by signing up at https://curiositystream.com/upandatom and using the code “upandatom”. Once you sign up you’ll get an email about Nebula. If you don’t get one, contact the curiosity stream support team and they will set you up with a free Nebula account right away. The name of the documentary I mentioned is Can a Computer Write a Hit Musical?

– This video is sponsored by Curiosity Stream
get access to my streaming video service Nebula.
When you sign up for Curiosity Stream
using the link in the description.
Hi there, welcome to “Up and Atom” I’m Jade.
I’m in France at the moment
and I often use Google translate to convert stuff
I want to say in English into French.
(keys clacking)
This got me thinking.
Does Google Translate know French?
If I wanted to translate an English sentence into French
I would first need to look up the French words
in a dictionary and make sure they had the same meaning
as the words in English
I would need the rules of grammar to let me know
how the sentence should be constructed.
It would take a lot of time and effort,
but in the end I would be pretty confident
that what I want to say is being conveyed.
(speaks French) But this isn’t how Google Translate works.
Computer programmers have loaded the AI
with a huge number of samples of about 80 languages.
It also scans millions of documents
on the internet to determine how words
or sentences in one language are translated into another.
For example, Google Translates scans pages and pages
of text to determine what friend is usually translated
as in French.
It doesn’t know the meaning of the word,
it just uses statistics.
This made me wonder what does it actually mean
to know something?
Turns out I’m not the first person who ever wondered this.
The same question has been pondered
by philosophers for at least 2000 years.
One of the first philosophers to think
about knowledge is Plato who lived in ancient Greece.
In his dialogue, Theaetetus Plato explores the definition
of knowledge that seems to capture what we usually mean
when we say that we know something.
One way this definition comes out is true belief
with an account or rational explanation.
Some version of that definition has survived
through the centuries to today
and is usually worded as justified true belief.
Let’s step back and think about this three part definition.
Knowledge is a belief that is something happening
in your mind.
You have a view or opinion about something
that belief is true.
For a belief to count as knowledge,
it must reflect accurately what’s happening in the world.
If I believe that the moon is made of cheese
that belief does not count as knowledge.
The belief must be justified that as you must have come
to the belief because you had a reason
whether evidence or experience or testimony
or something else.
Let’s explore this definition with an example.
If I say that I know a sheep is in the field,
there must be three things going
on for me to have knowledge.
First, I must believe that a sheep is in the field.
Second, a sheep must actually be in the field.
And finally, I must have formed my belief
by looking outside and seeing a sheep
or hearing a sheep or something like that.
That last step is crucial because it ensures
that knowledge can’t just be a lucky guess.
If I were in my house and I couldn’t see outside
and my friend asked me, “Is there anything in the field?”
And I just guessed, “A sheep.”
And there was one, you wouldn’t say
that I knew there was a sheep in the field.
You would say that I just had landed
on the right answer by accident.
That for my belief to count as knowledge
there would have to be something connecting my belief
to the world, something that caused my belief.
This seems like a pretty solid definition
of what it means to know something.
And for a long time
the justified true belief definition was well accepted
among philosophers of knowledge.
Then about 60 years ago
an American philosopher named Edmond Gettier,
found some cracks in the definition.
Now there was trouble.
Gettier showed that a belief can be both true
and justified and still not count as knowledge.
What if when you looked outside,
you were actually seeing a wooly dog
that just looked a lot like a sheep.
And then what if, unbeknownst to you,
a sheep was indeed in the field,
but out of sight in some bushes, it sounds contrived.
But this case of the sheep
and the wooly dog shows how the traditional definition
of knowledge is not airtight.
We seem to have met all the criteria for knowledge.
You believe there’s a sheep in the field.
There is a sheep in the field
and you have good reason to believe
that a sheep is in the field.
But according to Gettier,
you still don’t know there’s a sheep in the field.
Your belief was founded on a mistake
and is only true by coincidence.
There is no link between your true belief
and the sheep in the field.
In fact, the problem that Gettier spotted was identified
over 1,000 years earlier by Dharmottara
a Buddhist philosopher who lived in what is now Pakistan.
In Dharmottara’s example, a person sees
what looks like smoke rising in the distance.
He thinks that someone must have built a fire
but actually it’s just a cloud
of flies hovering above some food.
However, just over the next hill
someone has built a fire and is starting to cook a meal
but the fire has just been lit, so there’s no smoke yet.
The person believes that someone has started a fire.
Someone has started a fire
and the person has good reason to believe
that someone has started a fire
since he saw what looked like smoke, but it wasn’t smoke.
So he was correct, but only by chance.
Dharmottara posed another example,
a person is walking in a desert
and sees what she thinks is a pool of water in the distance.
When she gets closer, she sees
that she had been looking at a mirage.
There was no pool of water after all, but at the same time
there’s a well hidden under a pile of rocks.
So the person’s belief
that there was water in the distance was true.
It was also based on a good reason, the mirage
but in reality, her belief was lucky.
She had no idea that the well was there.
Problems like these are called Gettier problems
even though Gettier’s original article was only two
and a half pages long, it sparked a period
of new and exciting energy among philosophers
as it showed that the traditional definition
of knowledge wasn’t quite right.
Philosophers quickly set about trying to mend these cracks
by coming up with new, airtight theories of knowledge.
Now, before we talk about what these theories were
it’s worth asking, why should we care?
Gettier cases are rare.
Most of the time,
the justified true belief definition works just fine.
Is it really that big a deal?
Well, from a practical standpoint,
as artificial intelligence becomes more advanced,
the question of what exactly counts
as knowledge becomes more and more pressing.
There’ll be more about this later on in the video.
But for the moment I want to talk
about the philosophical reasons this question is important.
Knowledge is the only way we know of getting at the truth
the analysis of knowledge aims to uncover exactly
what this getting at consists of.
Philosophers of knowledge or epistemologists
once you understand knowledge in all
of its possible ways and manifestations.
To do that, we need a rigorous, airtight definition
of what it means and what it takes to know something.
We thought we had one
but Gettier showed us that that wasn’t true.
Something is missing in the definition.
And if we can’t find out what that something is
we don’t really understand the nature of knowledge.
So how do we solve this problem?
How do we stop getting gettiered?
This is still an open question
and there’s no general agreement on the solution
but several people have made attempts.
The main tactic has been to take
the justified true belief definition
and add another condition.
Philosopher, Alvin Goodman proposed
that for a justified true belief to count as knowledge,
the belief must have a causal connection
to the thing the belief is about.
In Dharmottara’s example, the belief that there was a fire
over the next hill was not caused
by there actually being a fire.
It was caused by flies which were attracted to the food.
According to Goodman,
that’s why the belief does not count as knowledge.
Another option is specifying
that a true belief counts as knowledge only
if it isn’t true because of some accident or luck.
This proposal seems to address one
of the vulnerabilities exposed by Gettier.
But it then leads to the problem
of what exactly counts as luck?
Maybe luck is common
and Gettier examples just involve too much luck.
It’s hard to say.
This proposal just seems to add another layer of difficulty.
Yet another proposal focuses on making sure
that you don’t overlook possibilities
or facts that could make you reconsider
whether you in fact know something.
So if you’re trekking in a desert
and you think you see a pool of water in the distance
you ought to take into account the possibility
that you are seeing a mirage,
which would make sense since you’re in the desert.
But this option doesn’t seem very practical
as it’s impossible to know every single thing
about every situation you’re in.
So is it even possible to ever know anything?
What does that mean and what exactly counts as knowledge?
These questions are becoming more and more pressing
as artificial intelligence progresses.
Recent artificial intelligence programs are now able
to predict the motion of planets accurately
without having Newton’s laws of motion programmed into them.
In science, we usually expect that scientists gather data,
come up with a theory or law that describes that data
and then test that theory on new data.
But these programs don’t do that.
Instead scientists just feed data
about planet motions into the computer program.
And the machine doesn’t try to formulate a law
of nature or deeper meaning behind the data
but it can still make accurate predictions based
on the input data alone.
So while we might be able to say that by deriving his laws,
Newton knew something about the way the universe works.
Can we say the same thing
about the artificial intelligence programs?
They can sift through data and make accurate predictions
but do their predictions count as knowledge?
And what about knowledge that isn’t so scientific
like poetry, art and music?
AIs can now make music, write song lyrics and stories
but do they actually know what they’re doing?
Could they ever be as good
as humans at these extremely human activities?
Well, you can be the judge of that.
In 2016 scientists got an AI to write an entire musical.
The AI wrote the story, the music, the lyrics
it took months, but the musical was actually performed
in London’s West End with actors and everything.
The whole process was filmed and made
into a documentary called
“Can a Computer Write a Hit Musical?”
which you can watch for free on Curiosity Stream.
I have to say, this is actually one
of the best documentaries I’ve seen in a while.
It’s entertaining.
And you learn so much about the machine learning process
how much data the AI needs, how it learns
and teaches itself skills like writing music and lyrics,
how close and how far they are from getting it right.
This is just one of thousands of documentaries
on Curiosity Stream.
Curiosity Stream are currently offering
my streaming service Nebula completely free
when you sign up with them.
Nebula is a streaming platform made
by a bunch of us educational YouTube creators.
Where we can explore making different
and experimental content.
It’s been nominated for a Streamy award, which is cool.
I’ve made a feature film on there called “Is Math Invented
or Discovered?” which you can check out if you like.
There’s a promotion going on right now
where you get a 26% discount off the bundle.
So that’s two streaming services for just $14.80 for a year.
Sign up with the link in the description
or go to curiositystream.com/upandatom
and start discovering today.
I’ll see you in the next episode, bye.
(upbeat music)

 

 

Share This- How do you know that you know what you know?

Leave a Reply