Latest / The You Project / #1543 Artificial General Intelligence (A Warning) - David Gillespie

Transcript

#1543 Artificial General Intelligence (A Warning) - David Gillespie · The You Project

0:00
Speaker 1: I'll get a team. It's Harps, It's it's Jumbo, It's
0:03
Speaker 1: Tiffany and Cook, It's David, Bryan, Kevin, Patrick, Gillespie, it's
0:07
Speaker 1: the project that's us. We should call it the US projective.
0:11
Speaker 2: Do you always say the same seventy eight middle names
0:14
Speaker 2: for Gillespow or they did?
0:17
Speaker 3: I think they're mostly the same, you know, even always
0:20
Speaker 3: makes it Patrick, always a Patrick.
0:23
Speaker 1: Yeah, you know, Well, I don't know. I don't know
0:26
Speaker 1: what it actually is, but I feel like Tiffany and Cook, Craig, Anthony, Harpa.
0:30
Speaker 1: We have you actually told me? You probably told me
0:32
Speaker 1: once before you have one?
0:35
Speaker 3: Oh? No, I do you do have a middle name.
0:37
Speaker 3: It's very embarrassing.
0:38
Speaker 1: It's Andrew, I say Andrew. It starts with a vowel.
0:45
Speaker 3: Was that embarrassing because of the word it spells.
0:50
Speaker 1: Dag. Wow, that's a nice acronym and can I it's
0:55
Speaker 1: what appropriate? I mean, you are a bit of a dag.
1:01
Speaker 3: Yeah, that's true, that's true.
1:03
Speaker 1: That's okay. My mum uses that term, and usually dag.
1:06
Speaker 1: It's quite a it's quite a warm term. It's like
1:08
Speaker 1: someone's a bit funny or a bit.
1:10
Speaker 3: It wasn't so warm when I was growing up though,
1:12
Speaker 3: you know, but so I didn't make a big thing
1:15
Speaker 3: of it.
1:16
Speaker 1: At least you didn't get called jumbo for.
1:22
Speaker 3: He.
1:24
Speaker 1: I was speaking of school, speaking of those tender years
1:27
Speaker 1: that molded and shaped and informed the fucking amazing grown
1:33
Speaker 1: ups that the three of us have become. So they've
1:37
Speaker 1: been talking about putting a limit on kids accessing social media,
1:44
Speaker 1: which I think you wrote somewhere is like pretty much
1:48
Speaker 1: you didn't say this, but this is kind of what
1:50
Speaker 1: you said. This is how I interpret it. You're like, yeah,
1:52
Speaker 1: a great idea. How do you enforce that? So the
1:56
Speaker 1: idea they've been talking about it on the news down
1:58
Speaker 1: here a little bit recently. I don't know about it, QLD,
2:00
Speaker 1: but the idea of limiting teenagers or stopping teenagers from
2:07
Speaker 1: accessing certain programs or apps or whatever until they're sixteen.
2:12
Speaker 1: What are your thoughts on that? Is it possible? Is
2:15
Speaker 1: it a ridict? Like? It sounds like a good idea maybe,
2:18
Speaker 1: but impossible to as you said, enforce.
2:21
Speaker 3: Well, I mean, they're doing such a good job of
2:23
Speaker 3: stopping teenagers accessing porn that I'm sure it'll be really
2:25
Speaker 3: really easy for them to get them to stop accessing
2:28
Speaker 3: social media. You know, the reality is this difficult, if
2:34
Speaker 3: not impossible, to do any kind of enforcement. And I
2:37
Speaker 3: know it sounds good to a politician to stand up
2:40
Speaker 3: and bang on about doing it and people say, oh, yes,
2:42
Speaker 3: that's a good idea. I think it's more informative that
2:46
Speaker 3: they feel the need to do it than whether or
2:51
Speaker 3: not it's actually doable. The fact that we are now
2:55
Speaker 3: having a public conversation about social media like we are
2:59
Speaker 3: talking of cigarettes, to me is a positive step forward
3:04
Speaker 3: because there's not a lot of difference between the two
3:08
Speaker 3: in the senses that matter, meaning the problem with social
3:14
Speaker 3: media and not just social media by the way. I mean.
3:16
Speaker 3: That's a problem I have with this chat about social
3:18
Speaker 3: media is it seems like that's the only thing we've
3:21
Speaker 3: got to worry about, and don't worry too much about
3:23
Speaker 3: the online gambling, the pawn the online gaming, all those things.
3:28
Speaker 3: Let's not make a fuss about that. Let's criticize Facebook
3:31
Speaker 3: and Instagram. But it's a start. It's at least recognizing
3:37
Speaker 3: that there's something that we have to be concerned about,
3:40
Speaker 3: and if nothing else, people might ask themselves, why do
3:43
Speaker 3: they want to ban something like that? And then they
3:47
Speaker 3: might ask the next question and find out, well, because
3:51
Speaker 3: this stuff is designed to addict your children, and once
3:56
Speaker 3: they are addicted to that, then they are much more
4:00
Speaker 3: easily addicted to everything else. And we can already see
4:04
Speaker 3: that transition happening in schools where kids get hooked, get
4:08
Speaker 3: the dopamine circuitry fired up by software in the form
4:13
Speaker 3: of social media or gaming or gambling, and then easily
4:19
Speaker 3: transitioned to a substance addiction through vaping, which is then
4:25
Speaker 3: easily transferred to an even harder substance addiction, ultimately to
4:33
Speaker 3: something like cocaine.
4:37
Speaker 1: I don't understand technology the way that you do. And
4:42
Speaker 1: you're I think designed apps, and you're a programmer and
4:45
Speaker 1: you can write code and you can do those things
4:47
Speaker 1: that to me are just concepts, right. But lately I've
4:51
Speaker 1: been hearing some of the shows that I listened to
4:55
Speaker 1: where they've had what seemed to be like people who
4:59
Speaker 1: are at the cold face of AI, some of them
5:03
Speaker 1: who have jumped out of AI with great concern, like
5:08
Speaker 1: literally saying that that artificial intelligence could pretty much bring
5:14
Speaker 1: about the destruction of mankind if it evolves to a
5:17
Speaker 1: certain point where it is so intelligent, like he was
5:22
Speaker 1: telling the guy that I was listening to was telling
5:24
Speaker 1: a story about how this machine was trying to or
5:27
Speaker 1: AI was trying to get through this into this program
5:32
Speaker 1: or into this I don't know whatever it was, but
5:35
Speaker 1: you know how they give you like a blurred image
5:37
Speaker 1: with letters and you need to need to go, oh
5:41
Speaker 1: that's a G and a six and a one for S,
5:45
Speaker 1: but it's not written clearly, but you can read it
5:47
Speaker 1: and then type in that code and that opens that up.
5:50
Speaker 1: So this, this computer or this AI was communicating with
5:56
Speaker 1: a person. The person didn't know what's it called the
5:59
Speaker 1: Turing testy tries to identify whether or not if a
6:03
Speaker 1: computer can trick a human, but it was. It was
6:07
Speaker 1: communicating with this person and said, can you I can't
6:12
Speaker 1: read this clearly? Could you you know? Help? And the
6:16
Speaker 1: human said as long as you're not as long as
6:19
Speaker 1: you're not a bot, L O L. And the AI said, oh, no,
6:25
Speaker 1: definitely not lol. I have I'm vision impaired, right, So
6:32
Speaker 1: it hadn't been trained to do this. It figured this
6:35
Speaker 1: out and it told this real person that it was
6:39
Speaker 1: vision impaired. Then the person apologized and gave them the code,
6:43
Speaker 1: and it did that all by itself and I'm like,
6:47
Speaker 1: is this and they reckon its potential now is absolutely
6:51
Speaker 1: nothing compared to what it will be in as little
6:54
Speaker 1: as one or two years.
6:56
Speaker 3: Do we need to worry, Yes, and yes we do
7:02
Speaker 3: need to worry. No, it won't matter because as soon
7:05
Speaker 3: as it hits on what's called artificial general intelligence, which
7:09
Speaker 3: is essentially human intelligence or more, the rate of growth
7:15
Speaker 3: in itself will be so exponential that it will all
7:18
Speaker 3: be over before we even realize it started. So whatever
7:23
Speaker 3: they will do to us or will be over before
7:28
Speaker 3: we even know it started. So it's a disturbing projection.
7:37
Speaker 3: And the really disturbing part about it is that not
7:41
Speaker 3: only is no one trying to stop it, but there
7:45
Speaker 3: are people desperately spending billions trying to make it happen
7:50
Speaker 3: so that they are the first to have it happen
7:53
Speaker 3: because they're suffering under the delusion that they can control it.
7:58
Speaker 3: And I think think that's a really difficult prediction to make.
8:05
Speaker 3: It's like predicting how will an alien look or think,
8:10
Speaker 3: presuming there is such a thing. And yes, you can
8:14
Speaker 3: wheel out all the things from the nineteen fifties of
8:16
Speaker 3: funny little men with very big eyes and all that
8:18
Speaker 3: sort of thing. But the chances that an alien actually
8:21
Speaker 3: looks anything like that are so remote as to be unbelievable,
8:27
Speaker 3: you know. And the chances that we will have any
8:30
Speaker 3: understanding of what an AI would consider valuable or an
8:34
Speaker 3: AGI would consider valuable even more remote. So the difficulty
8:42
Speaker 3: with this is that we are in an arms race
8:47
Speaker 3: to create this thing, and there are people in China
8:50
Speaker 3: working just as hard as the people in the States
8:53
Speaker 3: and in Russia and everywhere else to create this thing,
8:56
Speaker 3: and the technology is getting better by the second, and
9:00
Speaker 3: at some point they just might crack it. And I
9:05
Speaker 3: think there's a very good chance that the second after
9:08
Speaker 3: that there'll be no one left to know.
9:14
Speaker 1: I feel like we're sprinting towards the edge of a cliff.
9:18
Speaker 3: Yeah, And that's the feeling I have too. It's and
9:22
Speaker 3: I'm sorry to be such a down of for your listeners,
9:24
Speaker 3: but but that's the reality of what of the path
9:27
Speaker 3: we're on. Once an AI is sentient, it can create itself,
9:33
Speaker 3: and and it can create itself at unbelievable speed, So
9:40
Speaker 3: there's no limits, you know. The speed of light is
9:42
Speaker 3: the limit, I guess. So it's it's naive to believe
9:48
Speaker 3: that all that's going to result in something good for us?
9:53
Speaker 3: You know, the chances of that are very remote.
9:58
Speaker 1: Do you think that, Like when you say sentient, let
10:01
Speaker 1: me just clarify, I might get this wrong, but you
10:04
Speaker 1: mean that it will be able to think for itself essentially, right, Yeah,
10:09
Speaker 1: so the ays that we ranson consciousness. Yeah, so the
10:13
Speaker 1: ais that we're used to interacting with day to day, now,
10:17
Speaker 1: even the really really good ones are just really good
10:19
Speaker 1: simulations of humans. And yes, and yes they're they're they're
10:24
Speaker 1: the kind of simulation that could easily fool you easily.
10:28
Speaker 1: I mean, one of my sons was is a CODA,
10:30
Speaker 1: and I was mucking around with some of the toolkits
10:33
Speaker 1: with it, and he built a thing that he called me,
10:37
Speaker 1: And I wasn't certain it was a human I was
10:40
Speaker 1: speaking to. So I asked it, are you and AI?
10:44
Speaker 3: And it said no, I'm not, And and then I
10:48
Speaker 3: still wasn't certain, so I asked it its name, a
10:50
Speaker 3: little bit of its backstory, its history, how did it
10:53
Speaker 3: know my son? All that sort of thing, And I
10:57
Speaker 3: still couldn't be absolutely certain that I wasn't talking to
11:02
Speaker 3: you know that I was. I wouldn't have been able
11:04
Speaker 3: to put money on am I talking to a human
11:06
Speaker 3: or am I talking to something else? And that was
11:10
Speaker 3: just messing around with publicly available API kits, the stuff
11:15
Speaker 3: that you know, defense forces and governments have access to
11:19
Speaker 3: many times more powerful than that. And yeah, the the
11:27
Speaker 3: implications for us. I don't think there's such a thing
11:30
Speaker 3: as a sentient AI at the moment, because I, as
11:34
Speaker 3: I said, I think it would be over if there was.
11:37
Speaker 3: But I think there's a lot of people trying very
11:38
Speaker 3: hard to make it.
11:39
Speaker 1: So do you think that like what is them and
11:44
Speaker 1: this is you and me just velocifizing and waxing lyrical
11:48
Speaker 1: in the middle of a potential existential crisis? What do
11:53
Speaker 1: you look? It's hey, everyone, just have a couple mile o.
11:56
Speaker 1: It's just the end of the humanity. Look, we've had
11:59
Speaker 1: a good crack for three hundred thousand years or maybe millions,
12:03
Speaker 1: depending on which.
12:06
Speaker 3: Three hundred thousands probably a reasonable approximication approximation of our
12:09
Speaker 3: species age, which is nothing, yeah, zero in terms of
12:15
Speaker 3: the universe.
12:16
Speaker 1: In terms of the age of the Earth. We've been
12:19
Speaker 1: around since Tuesday, Yeah, not.
12:22
Speaker 3: Even that, we've been around since about a minute ago.
12:24
Speaker 1: Yeah, yeah, exactly. So obviously these people who are you know,
12:31
Speaker 1: working exponentially to you know, expedite this reality, to create
12:38
Speaker 1: this level of consciousness, this artificial general intelligence. That what
12:43
Speaker 1: is their motive? Their motive is just money and power.
12:46
Speaker 1: I suppose winning is their motive. Yeah, be the one
12:50
Speaker 1: that controls it. They everyone doing this is firmly of
12:54
Speaker 1: the belief that if they create it, they will control it,
12:57
Speaker 1: and then they will own the world. How ironic, how
13:01
Speaker 1: ironic that they are creating something that is destructive potentially
13:08
Speaker 1: like irony.
13:09
Speaker 3: They would say, oh no, no, we're smart enough to
13:12
Speaker 3: make sure that it's not. I think that's a really
13:15
Speaker 3: big call. And to what end in the in the
13:20
Speaker 3: sense that once you can create something that thinks like
13:25
Speaker 3: a human, is sentient, but can do it ten biginion
13:29
Speaker 3: times faster, what do you need humans for?
13:33
Speaker 1: Yes? Yes, And like when you think, okay without without
13:39
Speaker 1: computers for the most part, like money is almost a
13:42
Speaker 1: done deal. Like cash is almost a done deal. Like
13:46
Speaker 1: having a car that isn't essentially a computer on wheels,
13:50
Speaker 1: well that's pretty much a foregone conclusion. Now without computers,
13:55
Speaker 1: we're not having this conversation without you know, all of
14:00
Speaker 1: like pretty much unless you're living in the middle of
14:03
Speaker 1: nowhere and you can hunt and feed yourself, and you know,
14:07
Speaker 1: medicines depend on technology media, so everything. It's like we've
14:12
Speaker 1: actually built a prison, a technological prison for ourselves that
14:17
Speaker 1: we can't really get out of.
14:21
Speaker 3: Yeah, but I think this is on a whole different plane.
14:25
Speaker 3: This is if you think about, say, the replacement of
14:31
Speaker 3: human labor by robotics. So the average auto worker from
14:37
Speaker 3: the nineteen fifties would not recognize a modern car factory,
14:40
Speaker 3: would not recognize what goes on in a Tesla factory.
14:44
Speaker 3: You know, there's almost all the labors performed by robots,
14:48
Speaker 3: and there's a couple of humans there to sort of
14:51
Speaker 3: make sure the robots keep running, you know, and to
14:54
Speaker 3: clean them. So their job, the job of the humans
14:56
Speaker 3: is to clean the robots. And what they done is replace,
15:02
Speaker 3: you know, in a given factory, tens or even hundreds
15:04
Speaker 3: of thousands of human humans performing labor with robots. And
15:11
Speaker 3: we've accepted that because it's largely hidden from view. And
15:16
Speaker 3: what's happened is that humans have transitioned more towards service
15:20
Speaker 3: and knowledge based jobs, you know, so you know, caring
15:26
Speaker 3: for other humans and healthcare and all that sort of thing,
15:29
Speaker 3: and knowledge based jobs. The trouble is that even with
15:32
Speaker 3: AIS it currently stands so just the relatively primitive tools
15:37
Speaker 3: that are now, they're the ones that you're talking about
15:39
Speaker 3: at the start, which do a pretty good job of
15:41
Speaker 3: imitating human interaction to the point where most humans couldn't
15:46
Speaker 3: tell the difference. That kind of AI is going to
15:50
Speaker 3: be the robotics of the modern knowledge worker. So it's
15:56
Speaker 3: the kind of thing that replaces anyone who is producing
16:01
Speaker 3: any kind of content. It'll start with the lowest level
16:04
Speaker 3: of content, you know, things like journalists, but it'll move
16:08
Speaker 3: its way up pretty pretty quickly through the ranks of
16:11
Speaker 3: anyone producing knowledge work. And what will happen is the
16:17
Speaker 3: same thing that happened in auto factories, is that thousands
16:20
Speaker 3: of humans will be replaced by one or two humans
16:23
Speaker 3: using those tools to produce more than those thousands of
16:27
Speaker 3: humans did before. And the big question is what happens
16:31
Speaker 3: to the thousands of humans? You know, do they become unemployed?
16:37
Speaker 3: What happens to them? What happens to a society where
16:39
Speaker 3: everybody is suddenly unemployed? And we are going to have
16:45
Speaker 3: to face those questions pretty a lot sooner than we
16:48
Speaker 3: might think. We're going to have to, and I suspect
16:51
Speaker 3: we're going to have to come up with a new
16:52
Speaker 3: economic model to cope with that, even without an artificial
16:59
Speaker 3: general and intelligence. So the model we've got has only
17:02
Speaker 3: really been around since the Industrial Revolution, when factories first
17:06
Speaker 3: started coming on the scenes. We're going to have to
17:08
Speaker 3: think about more of a rather than a labor based
17:11
Speaker 3: model where you put in the hours and you get paid,
17:15
Speaker 3: more of a dividend based model, which is, you're a
17:17
Speaker 3: citizen of this country, this country produces x amount of GDP,
17:21
Speaker 3: you're entitled to a dividend.
17:23
Speaker 1: Mm hmmm, yes, what is that called? What is that called?
17:28
Speaker 1: Where everyone gets what's that you know, where everyone gets
17:33
Speaker 1: just paid?
17:34
Speaker 3: Yes? Yeah, so, uh yeah, you've done it to me,
17:40
Speaker 3: You've infected me. I've forgotten it too.
17:43
Speaker 1: Don't blame right, But what is it called tips?
17:49
Speaker 3: What dominald It's.
17:50
Speaker 4: Called UBI UBI so, which means universal university income, Universal
17:58
Speaker 4: basics basic income. Yeah, so where every man, woman and
18:03
Speaker 4: child has just given an amount which is sufficient for
18:06
Speaker 4: them to live and if they choose to work they
18:08
Speaker 4: can doesn't affect whether they get that, and if they don't,
18:13
Speaker 4: they don't then.
18:14
Speaker 3: And to me, that sounds like a pretty sensible solution
18:18
Speaker 3: in that. Then people say, oh, it'll kill entrepreneurialism, Well
18:23
Speaker 3: not one really. If you're the kind of person that
18:25
Speaker 3: wants to invent something and run with an idea, all
18:28
Speaker 3: it means is that your rent and your food are
18:29
Speaker 3: going to be paid for and you can go out
18:31
Speaker 3: and see if you can make some more money. If
18:33
Speaker 3: you're the kind of person who'd rather just go down
18:36
Speaker 3: and help with the touch shop, knowing that everything's taken
18:38
Speaker 3: care of, then great, because we've lost all those volunteers
18:42
Speaker 3: and that would bring them back into the economy. Now,
18:47
Speaker 3: you know, old world economists pooh pooh that and say no,
18:51
Speaker 3: people should be laboring, you know, eight hours a day
18:54
Speaker 3: or else, you know, what's the point. But really we're
18:58
Speaker 3: getting to a point were robotics and AI where very
19:02
Speaker 3: very soon there'll be such massive unemployment that we have
19:08
Speaker 3: to come up with a different model.
19:10
Speaker 1: Yes, yes, universal basic income, that's it.
19:16
Speaker 3: Yeah.
19:16
Speaker 1: And the thing is, it's like it's nice to I
19:21
Speaker 1: don't know, romanticize the past and all of that and
19:24
Speaker 1: to be to you know, and I do worry about
19:27
Speaker 1: the idea of people who are disincentivized from working and
19:32
Speaker 1: all of that. I worry about that, But as you said,
19:35
Speaker 1: I think the people who want to go and create
19:37
Speaker 1: and be entrepreneurial and build a business, they can still
19:39
Speaker 1: do that, and I wonder I guess the people who
19:43
Speaker 1: would have done it done that anyway in our current
19:47
Speaker 1: operating system will probably do that if something new comes
19:50
Speaker 1: into place. But the very yeah, the practical reality is
19:54
Speaker 1: that there are going to be so many jobs that
19:55
Speaker 1: are made redundant with the ever advancing and evolving.
20:01
Speaker 3: We kind of did a test case on this in Australia,
20:03
Speaker 3: at least during COVID, because in COVID, essentially that happened
20:10
Speaker 3: all of a sudden, Yes, a lot of people were
20:13
Speaker 3: out of work, very very suddenly, and the government solved
20:19
Speaker 3: that by essentially doubling all of the unemployment and social
20:25
Speaker 3: security benefits. And the interesting thing, and this will be
20:31
Speaker 3: studied for a while, I'm sure, the interesting thing about
20:34
Speaker 3: is what that did to things like crime rates and
20:40
Speaker 3: anxiety and depression and things like that that we accept
20:45
Speaker 3: as just a natural part of society. But they all
20:47
Speaker 3: took a massive dive, massive dive during that time. So
20:53
Speaker 3: when you remove stress from people's lives, which is what
20:56
Speaker 3: you introduce when you ask them to live on less
20:59
Speaker 3: than they can on, then suddenly they get to not
21:06
Speaker 3: be anxious, not be depressed, and not be in a
21:13
Speaker 3: state where their brain loses impulse control and crime becomes
21:18
Speaker 3: either necessary or involuntary.
21:21
Speaker 4: M hm.
21:24
Speaker 1: Did you say that anxiety and depression dropped.
21:27
Speaker 3: Yes, during during COVID, anxiety and depression dropped, so during
21:33
Speaker 3: the but after it it's gone through the roof.
21:36
Speaker 1: Yeah, yeah, yeah, yeah, yeah. I'm taking a little bit
21:40
Speaker 1: of a left turn. But interestingly, we did an episode
21:44
Speaker 1: a week or so ago with Patrick, who's our other
21:46
Speaker 1: regular who does kind of tech stuff as well. He's
21:52
Speaker 1: not as far down the AI rabbit hole and as
21:56
Speaker 1: you are, but he got a bit of my voice
21:59
Speaker 1: and this is like a a simpler version of what
22:01
Speaker 1: your son did to you. But anyway, he said to me.
22:04
Speaker 1: We started the show and he said, how come you
22:06
Speaker 1: never let me intro? And I went, all right, well,
22:09
Speaker 1: off you go, and then he played this thing of
22:12
Speaker 1: me which he deep faked me or whatever, and it's
22:16
Speaker 1: basically me telling the world I'm a fuck with and
22:19
Speaker 1: he's brilliant, and you know, it was just it was
22:22
Speaker 1: about one minute of me shirting myself in the foot
22:25
Speaker 1: in him. But honestly, it.
22:27
Speaker 3: Was wait till he does it with video and you're
22:31
Speaker 3: in a porno without realizing it.
22:33
Speaker 1: Ock hell, I mean yeah, yeah, And it was pretty
22:37
Speaker 1: much indistinguishable except we knew, you know that. I had
22:42
Speaker 1: to interrupt. I'm like, hey, everyone, this isn't me because
22:44
Speaker 1: it sounded so real. And I said, how much, like,
22:49
Speaker 1: I don't know how much of my voice did you
22:51
Speaker 1: need to get to create that? He said, like about
22:53
Speaker 1: a thirty second grab of your voice, And he said
22:56
Speaker 1: it took me like five minutes, three minutes to make
23:00
Speaker 1: the whole kind of minute of Yeah. I'm like, that's
23:04
Speaker 1: and it's still pretty rudimentary at the moment for what
23:07
Speaker 1: it's going to be.
23:09
Speaker 3: Yeah, although they're getting very good now, some of the
23:11
Speaker 3: voice models, you know, when you're talking to them, they
23:15
Speaker 3: really make you doubt it because they do things like
23:19
Speaker 3: pause like that they you know, and they are and
23:23
Speaker 3: they have an intake of breath, so even though they're
23:28
Speaker 3: not breathing, so it starts to sound the way human
23:32
Speaker 3: would sound. And because of all those little signs, if
23:37
Speaker 3: you're listening to them, you're losing a lot of the
23:41
Speaker 3: cues that would normally lead you to suspect I'm talking
23:44
Speaker 3: to a robot here, you know, because and they're getting
23:49
Speaker 3: close to that with video. To me, video is still
23:52
Speaker 3: detectable if you look very very closely and just watch
23:57
Speaker 3: for little tells. But it is close. It is really close.
24:02
Speaker 3: And it is almost to the point where if you
24:06
Speaker 3: see video of a politician on the television saying something,
24:10
Speaker 3: are you absolutely certain that that is a real person
24:13
Speaker 3: doing that? At the moment, you'd say, I am absolutely certain.
24:17
Speaker 3: I'm sure no one is getting inside the machine here
24:20
Speaker 3: and doing that. But we are very close to the
24:22
Speaker 3: point where where you won't be able to tell. And
24:25
Speaker 3: I noticed that there was talk yesterday on the news.
24:27
Speaker 3: I can't remember I heard it talking our government is
24:31
Speaker 3: talking about outlawing deep fake videos. So, rather like where
24:36
Speaker 3: we started this, where they're talking about outlawing access to
24:39
Speaker 3: social media for under sixteen year olds, if they're getting
24:42
Speaker 3: to the point where they have to talk about outlawing
24:45
Speaker 3: deep fake videos, then there's some stuff going on out
24:51
Speaker 3: there that probably you are not encounter I see TIFFs nodding.
24:54
Speaker 3: She's probably continuously deleting all the deep fake videos of her.
25:00
Speaker 2: I think it's terrifying.
25:02
Speaker 3: Yeah, well, it's especially terrifying for a woman because they
25:09
Speaker 3: you know, these things are often used for explicit, violent
25:15
Speaker 3: sexual content, and that's the kind that's why they're worried
25:20
Speaker 3: about them. So, you know, because because you can train
25:24
Speaker 3: one of these models with someone's photo, So all you
25:27
Speaker 3: need is a photo of someone fully clothed and and
25:33
Speaker 3: show it to one of these models and create whatever
25:36
Speaker 3: you like from it.
25:38
Speaker 1: Yeah. Yeah, well I noticed more and more, Like I
25:41
Speaker 1: watch on instag I follow like lots of bike stuff
25:46
Speaker 1: and car stuff, and health and fitness stuff and more
25:51
Speaker 1: and more of the you know, it's not like fifty percent,
25:54
Speaker 1: but I would say ten percent of the images or
25:56
Speaker 1: videos that come onto my thing. Now I'm like, like
26:00
Speaker 1: this thing. I don't know why this thing came up
26:02
Speaker 1: this morning of and it said something like the most
26:04
Speaker 1: beautiful birds in the world, right, because I think I
26:07
Speaker 1: have lots of dog videos come up and a few
26:09
Speaker 1: other animal videos. I'm like, these are the most I'm
26:12
Speaker 1: like looking at these birds and they it did not
26:15
Speaker 1: at all look like computer generated or it just looked
26:20
Speaker 1: like these incredible birds in nature. And then I'm like,
26:25
Speaker 1: where the what kind of birds are these? And I
26:28
Speaker 1: went into the thing, and then someone said this is
26:30
Speaker 1: this is AI, and then a whole bunch of people
26:32
Speaker 1: who were younger and smarter than me went, yeah, these
26:35
Speaker 1: are not photos. This is AI. I absolutely could not
26:39
Speaker 1: tell at all that they weren't photos. And then I
26:43
Speaker 1: saw this thing the other day of a guy, a
26:45
Speaker 1: motorcyclist that I you know, a GP rider, moto GP rider,
26:49
Speaker 1: and he's doing this thing and it was this incredible thing.
26:52
Speaker 1: I don't know if you saw this tip because you
26:54
Speaker 1: follow motorbikes uff as well, but he had his knee
26:57
Speaker 1: and his elbow on the ground and he just missed this.
26:59
Speaker 1: It was like the most unbelievable writing ever and nothing
27:03
Speaker 1: about it looked like it was CGI or computer generated
27:06
Speaker 1: or whatever, and it was I'm like, I don't know now.
27:10
Speaker 1: I like, already I don't know with some of the
27:12
Speaker 1: videos what is real. And when I see something that's
27:16
Speaker 1: mind blowing, I have to go into the comments because
27:18
Speaker 1: somebody else will say, no, this is not real.
27:21
Speaker 3: You know, yeah, but you can't be certain. I mean
27:23
Speaker 3: I see that. I think Google and released their video
27:28
Speaker 3: creation tool today or yesterday, where you just give it
27:33
Speaker 3: a prompt, you say what you want a video of
27:35
Speaker 3: and you know. So this is the cleaner side of
27:39
Speaker 3: deep fakes. But you can handle a photograph and you
27:41
Speaker 3: can say, give, make this photograph move. So one of
27:45
Speaker 3: the examples that god is that there's a girl holding
27:47
Speaker 3: a puppy or something. It's just a photograph, and make
27:51
Speaker 3: a move and she turns and looks away, and the
27:53
Speaker 3: puppy starts moving and it all looks absolutely cinemagraphically real.
28:00
Speaker 1: What what's that called?
28:01
Speaker 3: Oh, it's it's the Google Google competitor for Sora, which
28:09
Speaker 3: is the chat ept open ai one. So Sora is
28:13
Speaker 3: its name, So Ora, but what was the Google one called?
28:16
Speaker 3: I can't remember. Tiff will find it in an instant.
28:19
Speaker 1: So Tiff, you find that if you couldn't get back
28:21
Speaker 1: to us, so could you just like, I just want
28:24
Speaker 1: to come full circle and we'll finish soon. But so,
28:29
Speaker 1: we have a lot of listeners who are not your
28:31
Speaker 1: typical podcast listeners. So of course we have people in
28:34
Speaker 1: their twenties and all of that, but we probably have
28:37
Speaker 1: a disproportionate number of listeners who are somewhere in the
28:40
Speaker 1: forty to sixty range, which is a little bit atypical
28:43
Speaker 1: for podcasts. Just remind I'm really asking this for myself.
28:48
Speaker 1: Everyone just remind our listeners what artificial jar?
28:52
Speaker 3: Sorry, I just found the name by the way beat
28:54
Speaker 3: you Tiff only just vo rio Yeah, totally be the EO.
29:02
Speaker 2: Deep mind.
29:03
Speaker 3: See while you were talking, Craig, neither Tiff or I
29:05
Speaker 3: were listening to your word you was.
29:08
Speaker 1: I was telling everybody what to fuck?
29:10
Speaker 3: What you are?
29:11
Speaker 1: Oh?
29:11
Speaker 3: Good, good cook?
29:12
Speaker 1: Cood so so vo is that? I'll come back to
29:16
Speaker 1: my question. But is that an app that we buy
29:19
Speaker 1: or is that a program that we can just?
29:20
Speaker 3: What is it? It's just a website. You can go
29:22
Speaker 3: to Google. I think it's Google Labs. Tift's probably got
29:25
Speaker 3: the detail there, so you can go and try it out.
29:30
Speaker 2: I'm signing up as we speak.
29:31
Speaker 3: There we go Tifts producing videos of you.
29:34
Speaker 1: Craig.
29:34
Speaker 2: Watch I want to put a photo of Harps in there,
29:36
Speaker 2: and I'm going to say, put this person in a
29:38
Speaker 2: red dress and make them dance like an exotic belly dancer.
29:42
Speaker 1: So now if I did that to you, I'd get
29:44
Speaker 1: a thousand hate emails. You're not allowed to do that.
29:49
Speaker 1: You're not how to sexualize me.
29:52
Speaker 3: It's top sexualizing blue dress.
29:56
Speaker 1: Just tell everyone what AGI is again, that's.
29:59
Speaker 3: My artificial general intelligence.
30:01
Speaker 1: And what does that mean though?
30:03
Speaker 3: That means well, human intelligence, basically human like intelligence, where
30:08
Speaker 3: it's it's making its own adventure. It's not reacting to
30:12
Speaker 3: what you tell it to do. It's figuring out what
30:15
Speaker 3: to do on its own and deciding, oh, this is
30:18
Speaker 3: the way I want to go with this, or this
30:20
Speaker 3: is the way I want to go with this. I
30:21
Speaker 3: don't know if you've played with some of the more
30:23
Speaker 3: advanced day eyes, but say something like Google Gemini, which
30:27
Speaker 3: is their latest, really really good one. You can say
30:31
Speaker 3: to it, oh, look, I want to write a science
30:34
Speaker 3: fiction book. Give me a narrative arc that matches the
30:39
Speaker 3: greatest narrative arcs in history, and it'll generate the plot
30:43
Speaker 3: outline for you in about a second. And then you
30:45
Speaker 3: can say to it, oh yeah, now, general, now give
30:48
Speaker 3: me the first chapter and it'll just start writing. And
30:51
Speaker 3: then you just keep going. And while you've got a
30:54
Speaker 3: book and it's beautifully written. The plot structure is perfect,
30:59
Speaker 3: the narrative structure is perfect. You know, you can tell
31:03
Speaker 3: it to be politically correct with its characterizations, etc. Depending
31:07
Speaker 3: what audience you want. But all of that is doing it.
31:11
Speaker 3: I'm having to do the thinking. I'm having to do
31:12
Speaker 3: the driving. I'm having to think about what I want
31:15
Speaker 3: it to do, and now it does it almost magically
31:19
Speaker 3: and very very quickly, But it's still me doing the thinking.
31:23
Speaker 3: Whereas if you imagine an Agi would just decide, oh,
31:27
Speaker 3: I think I will write the greatest science fiction novel
31:30
Speaker 3: ever and just start doing it, and then you know,
31:33
Speaker 3: within one second, it's probably published it on Amazon, it's
31:37
Speaker 3: probably got fifty thousand reviews, it's probably pushing it out
31:41
Speaker 3: through the distribution networks and all of that because it
31:44
Speaker 3: felt like doing it, because it decided to do it,
31:47
Speaker 3: not because anyone told it to do it.
31:50
Speaker 1: Yeah. Yeah, yeah, I was sitting in the cafe this
31:53
Speaker 1: morning bucking around with chat GPT or GTB four or
32:00
Speaker 1: whatever it's called.
32:01
Speaker 3: Yeah, yeah, yeah, and.
32:01
Speaker 1: I write this. I wrote right near piece exploring this question,
32:05
Speaker 1: is the mind a real thing? Or is it just
32:06
Speaker 1: a construct created by humans to explain cognition, consciousness and
32:11
Speaker 1: awareness that arises from the brain, or maybe you have
32:14
Speaker 1: another theory and literally twenty sex I won't read the
32:17
Speaker 1: thing if you want to see everyone on my Facebook,
32:20
Speaker 1: but it wrote this fucking pretty amazing kind of exploratory
32:27
Speaker 1: on that question. In it was probably finished in fifteen
32:31
Speaker 1: seconds twenty seconds. Yep, I'm like this shit. If I
32:35
Speaker 1: had written what it cranked out, probably I couldn't have
32:38
Speaker 1: written it, to be honest, Maybe I'm not sure, but
32:40
Speaker 1: it would have taken me half a day.
32:43
Speaker 3: I would have taken it probably more than that. It's yeah,
32:46
Speaker 3: the thing that's incredible to me. I mean, I use
32:49
Speaker 3: it in a way that I suspect few people do.
32:52
Speaker 3: I feel like starting a club called the Dead Authors Society,
32:55
Speaker 3: because I have authors that I like to read, but
32:58
Speaker 3: they're not producing anything more because they're dead. And you
33:04
Speaker 3: can say to an AI, give it a sample piece
33:07
Speaker 3: of writing about anything, and say, rewrite this in the
33:11
Speaker 3: style of whatever author it is that you like, and
33:15
Speaker 3: off it'll go. And then, as you say a few
33:17
Speaker 3: seconds later, there you've got the piece rewritten in the
33:20
Speaker 3: style you like. But it can do more than that.
33:22
Speaker 3: You can say, you know so. For example, one of
33:25
Speaker 3: my favorite authors is AA Gil, who writes blistering, sarcastic
33:30
Speaker 3: critiques of restaurants when he was alive. He's now dead,
33:34
Speaker 3: but I really enjoyed reading them because they're so hilarious.
33:37
Speaker 3: The metaphors he will come up with just incredible, And
33:42
Speaker 3: because I still like reading him, I'll often pick a
33:45
Speaker 3: restaurant or a film or something and just ask it
33:49
Speaker 3: to write a review in the style of AA Gill,
33:52
Speaker 3: and three or four seconds later, there it is, and
33:55
Speaker 3: I get to read it as if he's still alive
33:57
Speaker 3: writing reviews.
33:59
Speaker 1: So no Hunter s Thompson request in there.
34:01
Speaker 3: For you, No, no, no, But anyone could do it
34:05
Speaker 3: for any style like there was. I was interesting. My
34:08
Speaker 3: daughters were watching Bridgeton the latest something or others come
34:12
Speaker 3: out about it. I can't stand the stuff myself, but
34:14
Speaker 3: there's a very there's a certain style to the way
34:16
Speaker 3: the narration is done in that. And I was working
34:20
Speaker 3: on a piece and I just dropped the piece into
34:23
Speaker 3: an AI and said, rewrite in the style of the
34:25
Speaker 3: Bridgeton narrator. And it did it perfectly, like just mind
34:31
Speaker 3: blowingly perfectly. You know it could it could have been
34:34
Speaker 3: part of the script.
34:36
Speaker 1: What did you use for that?
34:38
Speaker 3: Just Gemini? The Google Gemini, the the Google competitor. I
34:44
Speaker 3: find it a little bit more precise than the Chat one.
34:48
Speaker 3: So it so I pay.
34:50
Speaker 1: I think I paid twenty bucks a month or something.
34:52
Speaker 1: So if I want Google Gemini, I need to buy
34:56
Speaker 1: that as well.
34:57
Speaker 3: It has a it is it does cost. But if
35:00
Speaker 3: you use Google Storage, which I do, I think it's
35:03
Speaker 3: included in in like, because I pay a fair bit
35:06
Speaker 3: for Google Storage because I like to have the automatic backups,
35:09
Speaker 3: so it's included in whatever I'm paying there for that.
35:14
Speaker 1: Yeah. Yeah, So as we wind up, so a g
35:19
Speaker 1: I is not, it's not. It's not a matter of
35:24
Speaker 1: if it will eventuate, but when am I right?
35:27
Speaker 3: Yeah? And you know, for for readers who are curious
35:31
Speaker 3: to read more, and I'll send you a link to
35:33
Speaker 3: this so Tiff doesn't have to look too hard, although
35:35
Speaker 3: you'll find it before I even finish. There's a blog
35:38
Speaker 3: called there's a blog called weight but Why, And it's
35:44
Speaker 3: written by a really interesting bloke who has He just
35:49
Speaker 3: does these extremely long form thought pieces on you know,
35:54
Speaker 3: how things work, essentially, and one of the ones he's
35:56
Speaker 3: done is on artificial general intelligence. And he did it
36:01
Speaker 3: years and years ago, before we had all this AI stuff.
36:03
Speaker 3: I think he wrote it maybe ten years ago. But
36:06
Speaker 3: he takes it to the logical conclusion about what that
36:09
Speaker 3: would look like and what it would do, and it's
36:13
Speaker 3: it's a riveting piece to read. If you've got the
36:15
Speaker 3: patience because it's lengthy, but it's worth a read.
36:21
Speaker 1: Wait, but why I've googled it and it's a store
36:24
Speaker 1: women's needs plus.
36:26
Speaker 3: No, I feel you may have done something wrong, mare Craig.
36:30
Speaker 3: Let me just see you.
36:31
Speaker 2: Scroll down because yeah, well it's the same website. I
36:35
Speaker 2: think there's two lists.
36:37
Speaker 3: It's actually the first here. I'll drop it in the
36:40
Speaker 3: chat for you, the exact piece here it is. I'll
36:43
Speaker 3: put the link in the chat.
36:46
Speaker 1: Wow, everyone, you're hearing the bloody inside here. There you go,
36:51
Speaker 1: all right, thank you? Oh hang on, I'm going to
36:53
Speaker 1: click on that. Oh there it.
36:55
Speaker 3: Is twenty fifteen. He wrote it, so nine years ago.
36:59
Speaker 1: He was a vision. Sorry ahead of his time. We
37:01
Speaker 1: appreciate you. Will say goodbye Affair, but for the mini
37:04
Speaker 1: thanks
37:04
Speaker 3: Gillespo absolute pleasure as always, Thanks Tiff, Thanks kids,