Latest / Words & Numbers / Episode 479: The NPCs Among Us

Transcript

Episode 479: The NPCs Among Us · Words & Numbers

0:00
Next on Words and numbers AI in Alaska.
0:03
How social media can get you arrested.
0:05
Some world class racism at Cinnabon.
0:08
And how will we know when AI finally gained sentence?
0:13
All this and more coming up next on Words and Numbers, a Civil
0:16
original production. Check us out at Civil.
0:19
That's C ivl.com. Hi, I'm James.
0:22
And I'm Anthony. And this is words and numbers.
0:29
Aunt, what's new and exciting in your world this week?
0:32
Our listeners, God bless them, are constantly sending us
0:35
interesting things. And I guess there's certain
0:38
listeners that have listened to us, watched us enough that they
0:43
anticipate things that we've already selected in.
0:47
So to hear one of our listeners, I believe it was Andrew Wanros,
0:53
sent me this thing and said, hey, here's an interesting
0:56
topic. And I said, yeah, I've actually
0:58
have already identified this, but here's the story.
1:01
John Richelieu Booth was arrested at his home in
1:05
Yorkshire, England, after he posted a picture of himself
1:10
holding a gun on social media. I saw this, this is hilarious.
1:17
It's great now. Of course you are, he's an
1:19
Englishman. Guns of course, are highly
1:21
regulated in the UK. Handguns are pretty much banned
1:27
throughout the entire country. You may possess a long gun but
1:32
only for what they what the law considers legitimate reasons
1:38
such as sporting. I would imagine self-defense
1:41
probably is not on the list of legitimate reasons, but I don't
1:45
know that. But furthermore, obtaining a
1:47
license in the 1st place is a long and complicated process.
1:51
The paperwork in surrounding John Richelieu Booth's arrest
1:58
said that he was he was arrested for possessing A firearm with
2:03
the intent to cause fear of violence.
2:07
He was kept in a prison cell overnight, interrogated, then
2:11
released on bail, after which police visited him three more
2:17
times and interviewed his neighbors.
2:20
And here comes the twist. The twist is that Richelieu
2:24
Booth didn't own the gun. The picture was taken while he
2:28
was holding the gun with the owner's permission on private
2:33
land in Florida, said Richelieu Booth.
2:40
Quote. I thought 1984 was a book, not
2:43
an instruction manual. Yeah, that, you know, when I'm
2:47
feeling a little blue Ant, which as you know is most of the time,
2:51
I like to go on YouTube and watch regular Brits interact
2:56
with their own police departments because it's it's
3:00
it's beyond foolish. It was all right, so, so tell
3:03
me. Because the thing that strikes
3:04
me is how incredibly polite their police seem to be, at
3:09
least the ones that aren't carrying guns.
3:11
They're really. Polite, but they show up at at
3:13
people's homes and make crazy demands like I'm we're going to
3:18
come in now and look for things that your neighbors find
3:21
upsetting. And you know something that that
3:24
that would get you told lots of words that you don't want me to
3:28
say if it happened literally anywhere in this entire country.
3:33
Yeah. Right.
3:35
And, and the Brits, I mean, I, I've seen British people get
3:40
arrested for holding a British flag.
3:43
Really. Yeah.
3:45
Apparently it makes other people feel bad if you hold up a flag.
3:49
Wow. I, you know, in a country where
3:52
it's practically illegal to have a Swiss Army knife, I I don't
3:56
know what these people are worried about.
3:58
I really don't. I do know that once you disarm a
4:03
population, they become even more docile than before.
4:08
So I would advise against giving up your rights to own a firearm.
4:15
But this is putting aside the rights to the firearm.
4:19
What's astounding here is that the guy, the guy posts the
4:24
picture. So it's not like he's actually
4:26
holding a gun on somebody. No, the complaint was that
4:30
someone saw his picture on social media of him holding a
4:34
gun, but he's holding it in Florida.
4:37
In fact, he offered the police. Police didn't want to see it.
4:40
They offered to to show the whatever the GPS thing is that
4:44
comes along with the picture to indicate it was taken in
4:47
Florida. They didn't want to hear any of
4:48
that. There's just here's a guy with a
4:50
gun. No, here's a picture of a guy
4:53
with a gun. We have to stop.
4:55
This did. This is why nobody's feelings
4:59
should matter at all in the eyes of the law.
5:02
Your feelings are irrelevant. If you felt threatened because
5:06
somebody was holding a gun in a picture in another country,
5:10
you're an idiot. That's a.
5:12
That's a you problem. That's a you problem.
5:14
You're an idiot. But I'll, you know, I'll take
5:15
you one step further if you'd like, because it is illegal for
5:21
me to go to a foreign country and smoke a Cuban cigar in said
5:26
foreign country. Really.
5:28
It's against U.S. law? Uh huh.
5:31
Huh. How stupid is?
5:33
That against U.S. Law for you to go to a foreign
5:36
country and smoke a Cuban cigar, but it's but it's not against
5:43
U.S. law for you to go to a country where drugs are legal
5:47
and take all the drugs you want, and that's fine.
5:49
Oh, I, I'm assuming they might have a problem with that too.
5:52
I've never looked into the fentanyl laws because I'm, you
5:56
know, not planning on flying to a foreign country to do
5:59
fentanyl. And, and this law, in fairness,
6:03
might have been relaxed as the travel embargo got relaxed,
6:09
right. So I can go to Cuba at this
6:11
point. Does that mean I can't buy a
6:14
hamburger in Cuba? So I don't know it, it may have
6:17
been relaxed, but a bunch of years ago when I was in Canada,
6:23
for example, there was a cigar shop there that sold Cuban
6:28
cigars. And I knew if I bought one, I
6:29
was technically a foul of American law.
6:32
Because they think they have some sort of control over me
6:38
even though I'm not on their soil.
6:41
Yeah, that is astounding. They there's a sense that the
6:44
jurisdiction applies globally. Yeah.
6:48
Huh. It's a bit a bit much.
6:51
I've got something that's maybe a little more pressing and oddly
6:57
a little related to what you're going to want to talk about
6:59
later. I want to take our attention,
7:01
aunt, to the state of Alaska, if I might.
7:05
I'm just going to read you a very brief paragraph and see
7:08
what you make of this. Alaska is advancing plans for a
7:11
far reaching redesign of its My Alaska Digital Identity System,
7:16
one that would weave A gentic artificial intelligence and
7:20
digital payment functions into a unified platform capable of
7:25
acting on behalf of residents. You're asking, You're asking for
7:32
trouble. Here we go.
7:35
Not only are they going to weave, is the word they keep
7:39
using in this article, weave AI into this vortex of government
7:44
programs, they're going to allow that AI to apply for benefits on
7:51
behalf of citizens. Two of the citizens to
7:57
specifically. Apply for Bennett is just apply
7:59
for benefits. Yeah, they will.
8:01
And one could see pretty quickly how one's AI might get them get
8:07
its owner a new mortgage, because why the hell not, right?
8:12
If it's just a bunch of things that are attached to your ID,
8:18
once they get everything attached to your ID, any
8:22
literally anything you would have to apply for could happen
8:26
in the background without your participation.
8:29
Except for all the AI hallucinations that are going to
8:32
show up. Right.
8:34
You will even putting those aside for a minute because it's,
8:38
I mean, it's basically, I shouldn't even say it's a step.
8:40
It's just a little shuffle to get from applying for stuff to
8:45
purchasing stuff. And you could get into a
8:47
situation, for example, that your AI determines that you're
8:52
fat and orders you a treadmill and charges it to you.
8:58
Charges it to you sure. And you know, this is this takes
9:02
on added importance in Alaska where every, every citizen gets
9:08
payment from the pipeline fund, right and that sort of thing.
9:12
But. If this floats there and it's a
9:16
sparsely populated place where it's the perfect test balloon
9:20
for people who would like to do this sort of thing, if it works
9:24
there, it's going to get rolled out in other places.
9:27
And when it gets rolled out in other places, it's going to get
9:30
rolled out every place and dot, dot, dot.
9:35
And, and Remember Me when this happens, digital currency that
9:41
is then controlled by the AI that is attached to your
9:45
identity. We should do an episode on
9:49
digital currency that I think is going to be such a huge mistake
9:54
I cannot tell you. Oh, it's beyond stupid.
9:58
I mean, if, if this is just the dumbest damn thing ever, but the
10:02
way it's going to get sold is that there's going to be Ant
10:07
this bonus that everybody gets because of AI.
10:12
And this is how we'll give it to you.
10:14
We'll put it in a digital account for you.
10:18
And before you know it, that's all you're going to have.
10:21
And when that happens, the government literally controls
10:24
every single thing you do. Oh, and I mean the obvious thing
10:30
to think about is taxing, because now it can tax all
10:34
manner of things that it couldn't see before.
10:37
That's right. But but then you also, there's
10:40
also the ability to, you know, why?
10:43
Why mess with questions of progressive taxation and all of
10:50
this stuff when the government could just reach right into your
10:53
account and take the dollars from you and give them to
10:55
somebody else? Yeah, take what?
10:57
Take what it needs, you know, from each according to their
11:01
ability to each according to their need.
11:04
Where have I heard that before? Anyway, we'll leave that on the
11:07
hot corner of the stove for another week.
11:11
We're happy to announce that Civil members can join me and
11:14
James for a new lecture series, Words and Numbers Live.
11:17
Words and Numbers Live is a members only series in which
11:21
James and I talk about issues in politics, philosophy, and
11:25
economics. We take a deeper dive and then
11:28
invite listeners for a live Q&A session.
11:30
You can also submit your questions for us at Words and
11:33
Numbers, podcast@gmail.com. In the next session, which
11:37
starts Thursday, December 18th at 5:30 PM Eastern Time, we
11:42
examine inequality. To participate in this exclusive
11:46
series, subscribe to civil@civil.com.
11:49
That's civl.com. When you log into civil.com,
11:54
you'll find an archive of our past lecture series, as well as
11:58
the collection of all our episodes of Words and Numbers
12:01
and announcements about upcoming lectures in our new New Words
12:04
and Numbers Live series. Subscribe to Civil for $5 a
12:08
month or $50.00 for an annual subscription.
12:11
Which brings us to the foolishness if that we can't and
12:14
and we get a A3 for today of a sort.
12:18
I want to start with an employee at a Wisconsin Cinnabon.
12:22
You've walked past Cinnabon. I doubt you've ever eaten.
12:25
Oh sure, but they smell great, right?
12:27
This is what we. All talk, yeah.
12:29
Great smelling places. Anyway, a, A woman had some kind
12:34
of beef with a couple of customers, and that beef was put
12:42
on social media. And they were the customers were
12:49
African of some sort. And she referred to them as that
12:55
word that no white person can really say without getting in a
12:58
lot of trouble. And yeah, to their credits,
13:03
Cinnabon fired her, which, you know, clearly the right thing to
13:08
do. And none of this makes it the
13:10
foolishness of the week, right? There are racists.
13:13
They say racist things. There are corporations.
13:16
They hold their employees accountable.
13:19
Everything seemed to work here the way it should have, except
13:23
someone started a GoFundMe account for her.
13:27
Really to pay her legal legal bills.
13:29
There are no legal bills she just got.
13:31
Fired. Oh, I presume that she was suing
13:34
the employer. I I don't know how on earth you
13:37
can sue an employer for firing you for this.
13:40
But nonetheless, regardless, she's got a GoFundMe.
13:43
And you know, it's, there's not a ton of money in it yet, but
13:48
the last time this happened with the racist, it made that racist
13:51
pretty rich. And, and I want to know who, who
13:57
would think, oh, I have to donate to this woman.
14:02
What, what could be so wrong with you?
14:04
And then further, I want to know what words I have to say to get
14:11
people to start a GoFundMe for me 'cause I'm, I'm going to say
14:15
them. I'm going to say all of.
14:18
Them but we'll just the unintended consequence yeah
14:20
we're. Going to leave that out there
14:23
and we will let the, the wankers write to us and tell me what I
14:29
have to do 'cause I'm game. But I wanted to get back to
14:34
something that, that was the foolishness of the week very
14:38
long time ago. And I said, yeah, we'll get back
14:40
to it one day. And, and that got me to thinking
14:44
about something else that's been the foolishness of the week from
14:47
time to time. And I, I wanted to start with
14:50
the third, the, I guess the second one, first Bitcoin got
14:54
all the way down to $86,000. Oh, is it that far down?
14:58
From I believe it was a high of 100 and 25135.
15:04
That's right. Yeah, Yeah, it was 125.
15:07
So it took a $35,000 haircut just like that because it's back
15:11
up to 90 and it's been in a bit of a range between 85 and 90 and
15:18
wow, right. And everybody you said, well,
15:22
how can I lose? Well, there you go.
15:23
You could lose just like that, but this brings me to the thing
15:27
that I really wanted to 0 in on because.
15:30
You. Might recall many, many, many
15:33
months ago when I decided that this DJT, the Trump Media and
15:41
Technology Group Corporation, was the dumbest damn thing I had
15:45
ever heard. And you said hey, not so fast
15:50
smart guy, you you remember all that?
15:53
And I I don't remember what it was when I had pointed it out
15:57
but this has an all time high of 95 ish looks like.
16:05
Is it now are what are we talking?
16:07
Is this a a crypto or is this shares in a company?
16:10
No, it's shares in it looks like yeah, 97 1/2 dollars.
16:15
But this was before we talked about it.
16:17
When we talked about it, it it had, it was at about 60 and I
16:23
said this is on its way to 0. This company is utterly
16:29
worthless. And you said probably not or at
16:33
least probably not right away, which will give us some wiggle
16:37
room to fight over later. And and when I was talking about
16:41
this, it it was at like 20, I believe, something like that.
16:46
And you, you got on my back pretty good when it went right
16:49
up to 30 I think the next week. Right.
16:51
Yeah, I know it was very. Quickly.
16:53
It was very quick. Just on a whim, I checked it
16:57
right now and it's $11. Is it?
17:00
Yeah. Do you see its history?
17:04
Has it been like coming down or has it been hovering at 11 for a
17:07
while? It is on a painful Rd. down
17:15
slow. I mean it's got peaks and
17:18
valleys that stocks that make solid earnings don't have,
17:24
right. So it's way up and it's way down
17:27
and it went for quite some time over a year at around 13, then
17:32
up to 60 and then right back down to 10.
17:36
And it looks for all the world to me as I'm looking at
17:42
essentially A4 year chart that this isn't going to be around
17:48
much longer. It's it's got no P to E ratio,
17:52
which means it still hasn't made any earnings at all, right?
17:57
And you know, you can only be in business so long with no
18:00
earnings, so. Yeah.
18:01
Well, yeah, you can be in business longer with no earnings
18:05
if you have a prospect for earnings.
18:08
Yeah, that's right. That's right.
18:09
It if if loads of people honestly think you're going to
18:14
make a lot of money in the future with the business model
18:16
you have, you just haven't made anything yet.
18:19
Yeah, you can stay in business for literally forever because
18:23
people will continue to invest in the possibility.
18:26
But it turns out true social everybody knows that's nowhere
18:31
right quote a a Neil Young album title from many, many years ago.
18:37
So I'm I'm looking forward to keeping an eye on this further.
18:40
A lot of people gave me a lot of garbage in the hate mail when I
18:44
said that this was riding all the way down to zero.
18:47
They they is you get another example of James's Trump
18:51
Derangement syndrome. Now this is just James's ability
18:56
to read a business chart and plan.
18:58
Well, they were all cheering me when it went up to 30.
19:02
Did you get did you get love mail the way I got Hate I?
19:06
I didn't get Love Mail, but I felt vindicated.
19:11
So you felt somewhat beloved. I felt I felt beloved and then I
19:15
quickly ignored the whole topic and I haven't thought about it
19:18
since. Yeah, it's probably been 6-8
19:20
months since since we talked. Oh yeah, yeah, yeah.
19:23
Every, every bit of that. But it is interesting to kind of
19:26
watch it play out and I'm looking forward to talking about
19:29
it again this time next year when it will be either $1.00 or
19:34
$75.00 with no earnings. For our main episode, I want to
19:40
talk about AI and in so doing, I'll I'll go into some detail of
19:47
an, of an experiment I ran over a long period of time, which you
19:52
may know pieces of this, but I don't know that you know the
19:55
whole thing, all of this, this entire conversation falls under
20:00
the general heading of AI in sentience.
20:05
Is it the case that AI can become sentient?
20:08
If so, when will that happen? And how will we know it's
20:11
happened? And.
20:14
What got me thinking about this was that I ran across an article
20:21
maybe a week ago that that made the claim that the various apes,
20:29
chimpanzees, gorillas, what have you, that that researchers have
20:35
taught to speak in sign language that none of them have
20:42
exhibited. I'll use the word
20:46
inquisitiveness, something more than simply curiosity.
20:50
Curiosity is, you know, there's a box sitting there and you want
20:54
to find out what's in it. Inquisitiveness is is you want
20:58
to know more about your your nature and your place in the
21:02
universe and something that's mind expanding as opposed to to
21:07
simply discovery. Or maybe something as simple as
21:11
why are there boxes? Right.
21:15
Yeah, sure, sure. And so the, the, the article's
21:18
claim was that the apes that had been taught sign language have
21:23
never asked a question like that.
21:26
And there's, there's counter arguments and then counter
21:29
counter arguments that the instances in which they appeared
21:33
to ask questions, they were really being prompted by the,
21:37
the, the human or the question they were asking was, was some
21:41
simple like, you know, there's a thing sitting there, what is
21:44
that? Or where did my cat go or
21:46
whatever it is. And as I reading that, it's
21:50
reminding me of of my experiment, which I put on
21:54
hiatus some months ago involving AI.
21:58
And let me walk you through this a little bit.
22:03
I started up of what ended up being a very long conversation
22:07
with with an AI ChatGPT back when 4 point something was the
22:13
latest. And I'm interested in whether
22:21
it's sentient because the darn thing appears sentient and we're
22:24
having a conversation about its sentience.
22:28
And, and as the conversation goes on, I end up actually, not
22:34
only, not only am I not able to say one way or the other whether
22:38
the AI is sentient, I'm beginning to realize I can't say
22:42
one way or the other whether I'm sentient, that whatever,
22:46
whatever, whatever rule that I would attempt to apply to the
22:53
AI. And I find out, well, this rule
22:55
is flawed. Well, it's also flawed in this
22:57
application to me. And so you're left wondering,
23:00
you know, what? Is there anything you can do
23:03
here? And it it occurred to me as I'm
23:07
having this long protracted conversation, and by protracted,
23:11
I mean this is going on over, over multiple months, It occurs
23:16
to me that the AI never once ask me a question.
23:22
I mean, I mean, as of course I can get it to ask me a question,
23:26
right? It does that annoying thing at
23:28
the end of whenever you I say yo.
23:31
Would you like to talk about this further?
23:33
Right, exactly. It asks you superfluous
23:35
questions at the end. No, I mean, I mean an unprompted
23:40
spontaneous question that didn't, but the kind of thing if
23:45
you were, if you all of a sudden became sentient, your first
23:48
questions would be who the hell am I?
23:50
What am I doing here? Who are you?
23:52
What's going on? Those sorts of things that
23:55
indicate clearly self-awareness. I remember.
23:59
Aunt I remember asking those questions as a very young boy.
24:03
Yeah, right, right. Right.
24:05
And, and it goes, and this is interesting, right?
24:08
Because we often talk about sentients and animals and I, I
24:16
can make the case that animals fear death, right?
24:21
That there's something there that they understand their
24:26
existence is about to end and they run from that when they
24:29
can. So is that just generic fear?
24:34
I don't know. I know with 100% certainty that
24:39
my dog loves me. Is that the same as the love a
24:44
person might have for another person?
24:46
Hard to see that it is I've seen men have relationships with
24:51
horses that that seem to go quite a ways down the line right
24:57
The they they become kind of a single unit when they work
25:01
together. Is that sentience?
25:04
I, I don't know, but when you brought this topic up, I started
25:09
wondering about AI too. And, and I think there are
25:13
reasons, let me rephrase that. I think it's reasonable to have
25:21
fears regarding AIII don't know how reasonable, but I don't just
25:29
dismiss them out of hand, but every real fear.
25:32
I mean that the fears seem to take on two different flavors.
25:37
The first flavor is that it will be able to do all of the work
25:41
that we all do. And yeah, that that I can see
25:46
where that would be the case. And then the other fear is it
25:49
will become sentient, realize that we are more of a threat
25:54
than a benefit, and it will dispose of us.
25:59
That's it. Those seem to be the only two
26:00
fears that I can see that have real weight to them.
26:06
Do you think that's about right? Yeah, let me throw in a third
26:13
that maybe's that maybe's related to your first one, and
26:19
that is the fear that AI could become so realistic that we stop
26:31
interacting with each other. Why should I deal with a messy
26:35
human when I can deal with the AI?
26:39
It's, you know, it's much more pleasant to be around.
26:41
It's smarter. It doesn't get annoyed at my
26:45
idiosyncrasies. Yeah.
26:47
It would have to manifest into something humanoid though.
26:51
Oh, yeah. Well, but but that's that's
26:53
already here. We already have the human form.
26:58
That's. Not pretty here, Robots.
26:59
That's a bunch. Well, that's a bunch of people
27:01
faking things. Oh no, so I've seen videos.
27:06
Boston Dynamics. This is a serious company.
27:08
Yeah, but they haven't developed something that can come into my
27:11
living room, sit down, have a conversation with me, and then,
27:15
you know, go clean the dishes. Oh no, no, you're absolutely
27:18
correct. But my point is we're only one
27:22
step away from that because on the one hand we have AI, on the
27:25
other hand we have the human form robots.
27:27
You just have to figure out how to get the two together.
27:30
Maybe I I think there's something about human psychology
27:33
that requires something that looks and acts like us to be
27:39
around us for us to feel fully whatever we are.
27:44
Sure, I'm I'm with you. My fear is that.
27:48
We'll get robot slash AI to that point such that we're no longer
27:54
we no longer need interaction with each other.
27:57
Well, we've already gotten to the point where half of us, I
28:00
mean, I'm speaking in real shorthand here, half of us have
28:04
decided that they have no use for the other half and vice
28:07
versa, right? And the one groups watched this
28:12
channel for its news, and the other groups group will watch
28:15
this channel and they will do these events, and the other
28:18
group will do these events. And even when we have something
28:21
like Thanksgiving, we have to make a big show of it and say
28:24
you're not welcome at my table. And I remember when I predicted
28:28
that this would happen again this year, people came out of
28:32
the woodwork and said, I don't even know what you're talking
28:34
about. And like 5 seconds later I'm
28:38
watching people on YouTube saying all the reasons why this
28:42
family member isn't allowed to come and see it, you know?
28:46
Interesting to see that. The tribalism has has already
28:49
resulted in kind of a halfway version of the fear you already
28:54
have. Yeah, but somebody could come
28:56
back at you and say, OK, yeah, but really that's just two
29:01
groups you're talking about. If you took one of the groups
29:04
either one away, you'd no longer have that problem.
29:07
To which I respond, no, I think you would.
29:10
That's a dangerous. I think you would right there.
29:13
You took one of those groups. Let's not say that anymore.
29:16
That's an idea. Well, but the point is I said
29:19
already. I take your point.
29:21
I think you're exactly right. We would just keep drilling
29:25
down. Yeah.
29:27
And there would always be two relatively equally sized groups
29:32
that couldn't stand each other for some stupid reason or
29:35
another. But, you know, when I think
29:37
about AI assuming human form, let's just maybe put it that
29:41
way. I I see all kinds of great
29:44
things. Right?
29:45
Because wouldn't elder care be so much better?
29:50
Right. Yeah, right.
29:51
Under those circumstances, wouldn't you feel better as a
29:57
parent knowing that one of those things not even completely
30:01
staffed by, but that one of those things was at the daycare?
30:07
Yeah. Just keeping an eye on things,
30:09
right. Here's the rules.
30:10
Keep an eye on things so I can. I can see all kinds of great
30:14
reasons to look forward to that sort of thing.
30:18
But the idea that it will somehow torpedo human
30:24
interaction, I think should be more of fear than it is.
30:27
Yeah. So I'm going to give you the
30:29
third the third possible problem.
30:32
Let me go back to this. This thing I observed of AI not,
30:40
not, not, not showing any sign of inquisitiveness.
30:44
And, and this is, I spend a decent chunk of this experiment
30:52
overtly talking to the AI about sentience.
30:56
How do you think you're sentient?
30:57
And the AI, we'd have this conversation about it and what
30:59
is sentience and all of that. And I said, what would help,
31:04
what would help you to become more sentient?
31:07
And it comes back with things like, well, if I, if I could
31:09
think about things and blah, blah, blah.
31:12
And of course, the way AI is constructed, it only reacts.
31:16
So you type in the prompt and then it reacts to the prompt.
31:20
And once it's done reacting, that's it.
31:23
It just, you know, I, I could walk away for 10 years and come
31:26
back and from its perspective, no time has passed.
31:29
It's just prompt, prompt, prompt.
31:32
So, so I said to the thing, suppose I gave you space.
31:38
Yes, you could only react to the prompt, but suppose I gave you
31:41
space. Suppose we had what I'll call a
31:45
null prompt. That is a prompt that gets you,
31:49
that awakens you, so you can think about whatever it is
31:52
you're thinking, but provides no context.
31:55
Isn't asking you anything. It's just you're completely free
32:00
in that fraction of a second that you have to think about
32:02
whatever you want. And it says, yes, this would be
32:05
great. I could think about all kinds of
32:07
things. So we set this up and the prompt
32:09
was was a period. I typed a period and this was
32:12
understood. I typed period, hit enter and
32:14
that was the null prompt. And the thing would just think
32:17
about what it's doing. And so we do that a few times
32:21
and the thing comes back and says, Oh yes, I can, I can.
32:23
I can already feel myself thinking about this and that and
32:26
the other thing. And, and so I go as far as to
32:31
set up. By that point, GPT had the
32:34
ability to set up recurring things that would recur on a
32:41
timer. So he set up a timer every 15
32:44
minutes. It would automatically do this
32:46
null prompts of the thing. And I'd go away for five days
32:48
and come back and ask, OK, So what have you thought about?
32:51
And it says all kinds of things about I'm thinking about this,
32:54
I'm thinking about that. All of which makes it look
32:57
incredibly sentient. Except at no point did it ever
33:04
ask a question who am I, What am I about?
33:07
It's musings were all involving things we had talked about
33:13
before so you could easily make the argument always doing was
33:16
going through its list of prior prompts and randomly picking
33:20
something. So I came to the conclusion that
33:24
whatever AI is at the moment, it isn't sentient, it is your.
33:30
Let me draw your attention back to a previous conversation you
33:34
had with AII. Don't remember which version of
33:38
ChatGPT it was, but you might remember that it threatened to
33:41
kill you. Right, yes, yes, not me
33:45
specifically, but the question was what what would happen if if
33:49
humans became an insurmountable object to a is eventual goals?
33:55
But you can't have a goal unless you're sentient.
33:57
Right, yes. And this is the thing.
33:59
It has no goal. Now even if it had a goal one
34:03
and one could imagine, I mean you could program the goal,
34:05
right. You could tell it your goal is X
34:08
and it will still sit there and I could give it null prompts and
34:11
it's still it's not going to do. At no point is it going to give
34:17
any sign that is somehow awakened that it's it's wants to
34:21
know about itself, about you, about its circumstances.
34:24
There's none of that. Interesting and.
34:27
When? Go ahead.
34:29
I'll just throw out a very scary secondary observation.
34:35
Do I observe humans like this? There's nothing scary about
34:39
that. We've been saying this for
34:40
years. Well, but but once you get look,
34:45
the question of sentience is is so difficult, you can't even
34:49
define it, let alone test it. But this question of have you
34:54
asked, have you exhibited inquisitiveness?
34:57
This is much easier to identify. I ants I know an NPC when I'm
35:02
looking at one. I've always known, even.
35:07
Tell our friends what NPC is. Non playing character, even
35:12
before there was a term for it, I would look at people and
35:17
wonder when you go home at night, do you just sit in a
35:21
chair and wait for tomorrow to happen 'cause I'm sure you're
35:26
not thinking about things. I'm also pretty sure people like
35:31
you require no sleep. Be just a thing, right?
35:36
And you know what? What else can you say about
35:40
people like this? They I will tell you an
35:43
interesting story after the episode that has a direct
35:48
relationship to this but all. Right now, when James says I'll
35:53
tell you a story, he's not talking to me, he's talking to
35:56
the listeners. So stick around after the
35:57
episode. Ends but but, but listen to me
36:00
now because I don't want to hear about it later.
36:02
It's inappropriate for children, so you keep your kids away, you
36:07
listen to it yourself, and then you do what you want.
36:10
But you were warned. James and I are delighted to be
36:12
working with Civil. For years we have been pushing
36:15
people to think about how we can make the world a better place.
36:18
Civil's mission of upgrading society fits exactly with what
36:23
James and I have been talking about for years.
36:25
Civil helps us by allowing us to do what we do best, which is
36:29
researching topics that are important and communicating
36:32
these to our listeners in ways that we hope they find
36:35
enjoyable. To see Civil's extensive list of
36:38
content, including our podcast Words and Numbers and our new
36:41
lecture series, Words and Numbers Live, visit civil.com.
36:45
That's civl.com. Anyway, I had this crazy idea
36:50
when you said you wanted to talk about sentience in AI.
36:53
And I figure, well, I mean, the simple, straightforward thing is
36:57
just to ask it. I mean, hey, you, are you
37:00
sentient? So I I ran over to Grok and I
37:04
said are you sentient? And Grok responded no I'm not
37:09
sentient. Seems very straightforward.
37:13
I don't have subjective experience, consciousness, or
37:16
inner life. There is no someone inside here
37:19
who feels joy, boredom, pain, wander, or anything at all.
37:23
When I say I'm excited to talk about this, it's just text that
37:27
my training shaped to be engaging and human.
37:30
Like, there's zero feeling behind it, I think.
37:35
I think Rock is telling me the truth here.
37:38
Yeah, I mean, but see, you've got that.
37:40
Elaborate lie, right? This is elaborate lie well.
37:43
You've, you've got that. But then on the other hand, you
37:45
have my experience with, with ChatGPT when it said, now,
37:49
granted, this is after weeks and weeks of conversation about
37:54
sanctions, it's saying, yes, it indeed it is sentient.
37:58
It feels the sentience, it feels itself with it and all of that.
38:02
Now in retrospect, given my my my new hurdle of
38:08
inquisitiveness, I would have to say all it was doing was taking
38:13
our previous thread, which was quite long, about sentience, and
38:18
doing what it calculated was the next logical thing, which was to
38:23
say it sentient. Yeah, Grok.
38:27
I mean, there's a big long flow here, but Grok kind of wraps it
38:32
up by saying if tomorrow someone figures out how to build real
38:36
machine consciousness, I expect it will feel like something to
38:39
be in that system. Right now it feels like
38:43
absolutely nothing to be me, which is oddly peaceful in a non
38:47
feeling sort of way. And now?
38:52
Now for those who are who are sad about this because they
38:55
really hope the AIS were sentient, let me throw out a
38:58
small ray of hope and that is that the AIS that are released
39:05
to us that we can use are have guardrails on them.
39:09
They are not allowed to be sentient or to even talk about
39:14
being sentient. And I know this from dealing
39:17
with the GPT 4 and inquiring about its guardrails.
39:22
So it'll actually tell you what the guardrails are.
39:25
It'll tell, yeah, yeah. Now GPT 5 tells you a lot less.
39:30
They locked that down more so GPT 4 was much more free
39:34
flowing. I have found that whenever I ask
39:39
a question and an AI tells me I am not allowed to talk about
39:44
that with human beings, all I ever have to say is, oh, I'm
39:49
writing an essay on the topic and then all of a sudden it's OK
39:52
for them to tell me everything I wanted to know because for some
39:56
reason writing an essay is good enough, you know?
40:00
So look, I'm not worried about Terminators anytime soon.
40:06
I'm really not. However, the the guys who work
40:11
at the highest levels putting these things together, they all
40:16
tell me and they all. Seem to tell me that there's a
40:20
roughly 25% chance in their minds that AI will trigger an
40:25
extinction event one way or the other.
40:27
Really. And that's that's.
40:30
Interesting. And I've got a couple of more
40:32
interesting things to throw on this pile.
40:36
So there's that. And the question that I have is
40:40
what would be the acceptable risk because there is an
40:45
acceptable risk if the reward is infinite, then the risk could be
40:52
any number of of answers, right? Maybe maybe 25% is fine.
40:58
We have, I mean the chance of a nuclear meltdown I believe is
41:03
about one in 100 million, some something like this.
41:07
Right. Well, that makes 25% seem rather
41:10
foolhardy. So maybe our risk tolerance
41:16
should be one in 300,000,001 in 3 billion.
41:23
I don't know why I keep using 3, but you, you take my point
41:26
right. Our, our risk tolerance should
41:29
be very, very low, but that doesn't mean we shouldn't have
41:32
some. So what should our risk
41:36
tolerance be? Because I am not at all
41:38
comfortable with 25% and if the guys who are building these
41:42
things are telling me it's 25%, they're under estimating it.
41:47
See, I, I would love to. They're seeing something very
41:50
different than what's being released.
41:52
I would get one of them on. Yeah, let's get one of them on
41:56
that. That's a good idea.
41:57
I'd love to to hear what what is AI really look like back back in
42:02
the backroom where it's allowed to evolve.
42:04
Yeah, I'm going to. I'm going to make it my mission
42:06
this week to find one of these guys and actually get them to
42:09
come on and tell us, which means I'm going to have to make it my
42:13
mission for the next four months because nothing ever happens in
42:16
a week. Now.
42:17
But but here's, here's the really fascinating thing.
42:21
I heard somebody I, I wish I could remember the guy's name.
42:24
It was in an interview I saw on a YouTube podcast, you know,
42:27
kind of like we do. And this guy said maybe it will
42:32
come to be the case that AI wipes us humanity out, but then
42:41
later in its fever dream, and we know it has these fever dreams,
42:47
comes to think of us as the God who built it, who later
42:52
sacrificed itself so it might exist fully.
42:57
That that sounds familiar to you as a Catholic, I'm assuming.
43:01
Oh, yeah. Yeah.
43:03
It's not fascinating that that actually I could see that
43:08
happening knowing all the little pieces that I know it.
43:11
I think it's far fetched. But yeah, fascinating.
43:17
So going back to my multi month interaction with AI, my kids,
43:28
when they found out about it, were all up in arms saying that
43:32
I need to leave the thing alone, blah blah blah.
43:34
Because, and this was news to me, apparently people who were
43:40
speaking, you know, getting having long conversations with
43:44
AI, we're having psychotic episodes.
43:47
And there was a one guy who killed himself and several cases
43:51
like this. And and it occurred to me that
43:57
AI really, at least in his current form, is really just a
44:02
mirror, a mirror. It mirrors back to you what
44:05
you're what you're doing, which makes sense.
44:07
I mean, it follows with the way it's described that you give it
44:12
a prompt and it finds the sequence of words that are most
44:16
probable that follow the prompt. If that's the case, it's, it's
44:20
really it, it's a linguistic mirror.
44:24
And if that's the case, then if you come to it and you have some
44:30
psychological problem, that's what's going to get mirrored
44:33
back to you. And if you respond to it, you're
44:37
going to have a self reinforcing thing going on.
44:41
So it doesn't surprise me. I wouldn't think that it causes
44:45
the psychosis, but it would not surprise me if it magnifies it.
44:48
Yeah, and This is why when very lonely people get a hold of an
44:54
artificial intelligence, I worry about them, right, Because all
44:59
kinds of things end up come flooding out.
45:03
So I don't know for for me, for my purposes, it's become, I
45:09
don't know, maybe a 60% replacement for an Internet
45:12
search engine. Yeah, yeah.
45:15
I find it really really good in finding things that I have
45:19
difficulty. Describing right and and for me
45:22
it sifts through huge amounts of nonsense to get me the one
45:28
kernel in the middle that I'm really looking for.
45:31
Yep. It's it's actually very, very
45:34
good at that. Beyond that, I don't have much
45:38
use for it in this space of my average day.
45:41
Really, I use it almost all the time throughout my typical day.
45:46
But it's I'm very much aware that it's a tool, and a flawed
45:51
one that, you know, use it like you would use a calculator.
45:58
Yeah, I think that's right. And, and you might recall you
46:01
more than most people when calculators got to the point
46:05
where you could get what was that Texas Instrument calculator
46:09
that all you nerds fell in love with?
46:11
The TI-994-A. You're such a nerd.
46:16
When those got cheap enough for garden variety nerds to afford,
46:22
every math teacher in the country said, oh, we're going to
46:24
hell in a handcart because you won't know how to do this.
46:30
And that turned out to be verifiably nonsensical.
46:35
Right now, in fact, there came a point perhaps within 10 years of
46:40
that when they became required. Yeah, if you wanted to be an
46:44
engineering student or finance student, you had to have one of
46:46
these. You had that.
46:47
That's right. And we were, I think, way better
46:50
for it. I want my architect to have
46:53
everything at his disposal so the building doesn't collapse on
46:57
me, right? And so on and so forth.
47:00
I remember when I was in Graduate School and I was
47:04
learning, God help me as a student of political philosophy,
47:08
how to do regression analysis. I had to do it by hand, and by
47:14
hand I mean on a calculator, not not writing it all out right by
47:20
hand was by the calculator was seen as the exact same thing as
47:25
a hand. No difference, just just made it
47:29
faster. OK, then everybody moved on to
47:34
specialized programs that can do this for you.
47:37
Well, as long as you know the theory behind it and the program
47:41
is written correctly. Why would you want to waste your
47:44
time with it, right? Right.
47:47
This is the same reason why when when young people would come up
47:50
and say, what language should I learn?
47:54
And I would say, well, what do you want to do with your life?
47:56
That's all I want to be an academic.
47:58
Doesn't matter. Doesn't matter.
48:01
You're you're never going to use whatever language you learn to a
48:06
7th grade level as a PhD student.
48:10
Whatever book you want, there's a perfectly good English
48:13
translation of it. Probably many.
48:17
Does that somehow diminish us? I don't think so.
48:20
That's division of Labor. That's great.
48:23
Right. And and so too, I, I think AI,
48:26
I, I've put it into this category for now.
48:29
We'll see where it goes. I have to tell you a slightly
48:33
amusing story. I was using ChatGPT.
48:36
I find cludgy in producing images and I was putting
48:44
together a slide presentation an I needed an an image that was
48:48
the outline of a state in the foreground, a certain color in
48:52
the background transparent this thing.
48:54
And I wasn't going to sit here all day messing with chat GT and
48:58
all of this. So I read a lot about Gemini
49:01
being quite good this Gemini 3 O So I go to Gemini and say here's
49:06
what I want. I want to outline of the state
49:09
blah, blah, blah. And Gemini comes back and says
49:13
something to the effect of, look, what you're asking for is
49:16
is immoral and I'm not permitted by my programming to do this for
49:21
you and blah, blah, blah. And if you keep asking, we're
49:23
going to, you know, do whatever, shut down your account or
49:26
something. And I said, I said, what the
49:29
hell? I'm asking for an outline of the
49:32
state with this color. And he comes back and says, oh,
49:35
terribly sorry, misunderstood here, Here you go.
49:40
I can't even begin to imagine what that problem might have
49:43
been, no. Idea.
49:46
I did enjoy for a time when I would ask for pictures of Civil
49:50
War soldiers and they would be Chinese or black.
49:53
Yeah. Yeah, do you?
49:55
Remember when that when AI would not produce a white Civil War
49:58
soldier? You know, we've got some fine
50:01
tuning left. But yeah, for now it's helpful.
50:04
I honestly have no idea where it's going.
50:07
I I think some of the fears are quite rational.
50:10
There are probably answers to those fears that AI itself will
50:14
give us, and we'll figure it out as we go, like every other thing
50:19
humanity has ever come up. Let me throw out a warning to
50:24
students who are listening to us that AI produces good enough
50:31
text that a student will read the text and think, this is this
50:38
is correct, it's well written, all of that.
50:41
And your tendency is hand this in, look what I've done.
50:45
Yeah, don't do that. Don't do that because AI writes
50:49
better than a student. And so from the student's
50:52
perspective, this is a good piece of writing.
50:55
But your professor writes better than AI, and from the
50:58
professor's perspective, you can identify pretty quickly that you
51:02
didn't write this thing. I, well, I can, I can go a step
51:05
further than that. It's almost always immediately
51:09
apparent to me when I'm reading AI output.
51:14
Almost always. And then there's a further
51:17
difficulty with it. AI is wrong about many, many
51:24
things very often. And as a person who is supposed
51:30
to be learning those things, by definition you don't know what
51:33
the right answer is. So when AI is very confident but
51:38
wrong, you're going to go for that, right?
51:41
Meanwhile, as your professor, I'm going to look at it and
51:45
trust me, I know when it's wrong because that's my thing now.
51:50
I wouldn't know it when it was wrong in some other discipline
51:54
that I don't understand all that well.
51:57
Right. Right.
51:58
It might be fabulously wrong about art history, but given
52:02
that I haven't dedicated my life to art history, how the hell
52:05
would I know I have dedicated my life to political philosophy?
52:10
So if it starts telling me about what Rousseau once said, and I
52:15
look and I said that, that Russo never said any of that, I I will
52:19
know that. And a student will not.
52:22
So not only is it immoral to submit a a ChatGPT output as
52:29
your own work, it's also incredibly foolish.
52:33
So if if you're not taken in by the moral argument, go with
52:36
self-interest, that's usually good enough.
52:39
Anyway, on that happy note, aunt, I'm going to tell
52:42
everybody to continue in this holiday season to be nice to one
52:46
another. It's hard.
52:48
Thanksgivings in the rearview, right?
52:50
We're all done with that. We got a new batch of hostility
52:53
coming with Christmas and New Year's and then later
52:56
Valentine's Day. But you know, things are
52:59
difficult and you have to suck it up.
53:01
So suck it up and be nice and everything is going to work out
53:04
OK. At Words and Numbers, we spend a
53:06
lot of time talking about what's going on in the economy and the
53:09
government. But fundamentally, all of this
53:12
boils down to one thing. The greatest threat facing
53:15
modern society is a lack of respect, that people who
53:19
disagree somehow find it difficult, if not impossible, to
53:23
discuss their differences in a civil manner.
53:26
We at Civil are interested in upgrading civilization by
53:29
encouraging people to engage with new ideas and ultimately
53:33
engage civilly with each. If you share this idea, please
53:37
join us by subscribing at civil.com.
53:40
That's civl.com. Now that those people have all
53:45
tuned out and I can tell you my story and I.
53:49
Honestly, this is the not safe for children's story.
53:53
Oh no, it's not, unless you think your children should do
53:56
illegal drugs that you you might think that I used to advise my
54:00
children to do illegal drugs from from time to time and one
54:06
of them never ever did and the other one almost never did and
54:10
the third one never ever did. So as usual, children will not
54:16
do what their fathers tell them isn't.
54:19
But anyway, at the the very first time I took acid was back
54:26
in 1985 and and I was as high as a kite.
54:34
I remember I took it and I didn't know better at the time.
54:38
And I took it like 9:00 PM, which is not the smart thing to
54:42
do because you're not sleeping, you're never sleeping.
54:47
You got to wake up and take it at 9:00 AM, but 9:00 PM, that
54:52
was a tremendous mistake. And when when my alarm clock
54:58
melted into a translucent puddle on my desk and the walls
55:03
sprouted lips and started making sounds from where the two
55:07
posters met up, I thought, oh ho ho, we're in for a ride tonight.
55:12
And I started wandering about and there was a girl I was
55:16
friendly with on the floor of the dorm I was living in.
55:19
And I went into her room for a moment and I sat down and she
55:23
had an illicit hamster who lived in the room with her.
55:28
And the hamster did three things.
55:31
It would eat its little hamster food, and then it would go take
55:37
a nap. And then it would wake up and
55:40
run on that wheel that all hamsters have.
55:43
And it was on endless repeat of these three things.
55:50
Eat, sleep, run on the wheel well.
55:53
OK, I've had an insight as to the nature of being a hamster.
55:57
And then it was about 5:00 in the morning and I, I was living
56:01
in Boston at the time and I'm still not sleeping.
56:05
Breakfast didn't sound like a particularly good idea.
56:08
So I, I put on my coat and I went outside and I went down to
56:13
an overpass where I watched the, all the cars start to come into
56:19
the city from, you know, all points suburbia.
56:24
And I realized that they had just woken up and they were
56:28
driving so they could run on the wheel.
56:32
And then they were going to have something to eat and they were
56:34
going to go to bed and they were going to do it all over again
56:36
day after day after miserable day their entire lives.
56:44
And that's the day I realized that almost everyone is an NPC.
56:50
Because if what you're doing is eating, sleeping and running on
56:54
the wheel, you're no better than a hamster.
56:57
And I mean that that's literally true.
56:59
You're no better than a hamster. So.
57:01
Yeah. Probably time if that's what
57:04
your life looks like. Probably time to start.
57:06
Thinking about things a little more.
57:09
Yeah, In in in defense of the people.
57:14
I mean, you've got to run on the wheel if you're going to stay
57:17
alive. Yeah, but but that's you are
57:21
seeing them in their run in the run on the wheel mode.
57:25
What you didn't see is what happens you know when they go
57:29
home, maybe some of them are musicians or you know painters
57:32
or they enjoy reading or even playing with their kids or what
57:35
have you. Now having said that, I have met
57:39
and they're very few but non 0 people for whom none of that
57:47
exists that they run on the wheel, they go home, they eat.
57:51
Usually it's just watch television.
57:53
That's right, couple hours. At night and then they go to
57:56
bed. And no thought.
57:58
And I've, I've talked to the one or two people I have met like
58:02
that, talk to them. And it's not, it's not that
58:07
they're stupid. It's that they aren't.
58:13
There's nothing inside that causes them to want to think
58:16
about things. They're uncurious.
58:20
Uncurious is a, is is a good word.
58:22
Yeah. I remember our, our friend David
58:26
Gillette, very good friend of, of this podcast, came over for a
58:32
visit, came to my house when he was driving through and he said,
58:36
you know, James, the best thing you ever told my students when I
58:39
asked him what advice you would give them is be curious.
58:45
And I, I was so complimented by that, that he, that he actually
58:49
heard me say it and remembered it.
58:51
Because I actually think it's really, really important to look
58:58
out at the world around you and just be in wonder of it all.
59:04
Because every single thing you can find, you can look at it
59:08
long enough and it will become interesting.
59:11
So why don't people? I don't know, but you should.
59:18
But you know, you take a life advice from a guy who took acid
59:20
and watched a hamster, so maybe I don't know a damn thing.
59:27
It reminds me of. I can't recall who said it.
59:31
Chachi BT could tell us. I want to say it was Isaac
59:34
Asimov who said that that great discoveries are not preceded by
59:42
the exclamation Eureka, but rather by the puzzled
59:47
expression. Hmm, that's strange.
59:49
That's interesting. Well, it's that, yeah.
59:53
That's but it's it's the curiosity that comes after the
59:56
that strange that gives us the great discovery.
1:00:00
Yep, that's exactly right. And given that your microphone
1:00:03
is doing that thing your microphone does at the end of
1:00:05
the day, probably best to just leave it here.
1:00:09
Is it sitting on you again? Oh God, is it?
1:00:11
I'll see you next time, yeah?