The Danger of Artificial Intelligence: Humanity's Last Invention? | ENDEVR Documentary

[00:06] Intelligence equals power.

[00:09] Intelligence equals ability to influence

[00:11] the shape of the future.

[00:15] If you look at us in the world today,

[00:19] we are definitely shaping our own future

[00:25] and we're shaping the futures of other

[00:26] species around us, too.

[00:29] birds, fish, whales, even primates,

[00:33] they have no chance at being able to

[00:35] shape the world around them when humans

[00:37] are around

[00:39] because we are a very intelligent

[00:41] species

[00:43] and the most intelligent species that

[00:44] the world has seen to date.

[00:48] [Music]

[00:54] Artificial intelligence is the first and

[00:56] only technology we'll ever create that

[00:58] will bring an intelligent system into

[01:01] the world that can match or exceed

[01:03] humans in terms of intelligence. We'd

[01:05] basically be seeding the reigns to a

[01:08] more powerful species and we become a

[01:10] secondary species on Earth and that is

[01:13] just a completely unprecedented moment

[01:15] in human history.

[01:22] AI is basically just a type of system

[01:24] that is carrying out a thing that you

[01:25] would call an intelligent task.

[01:30] We have AI systems around us today.

[01:33] You have AI that's able to trade on the

[01:36] stock market in a way that's much better

[01:38] than humans can. We have AI that's

[01:41] increasingly being used in domains like

[01:43] healthcare or education.

[01:45] And the systems that we have today, you

[01:47] would call artificial narrow

[01:48] intelligence or ANI

[01:51] because they are only applicable or

[01:53] useful in a very narrow specific task.

[01:57] The current state of AI is that you've

[01:59] got these narrow systems which actually

[02:01] can exceed human performance in all

[02:03] these different areas.

[02:05] >> As with all computer operations, the key

[02:07] to designing intelligent machines lies

[02:10] in programming. So, I could have a

[02:12] program which can play chess and that's

[02:14] all it can do. You can't have a

[02:16] conversation with it, but it can play

[02:18] chess better than anybody.

[02:21] But in day-to-day life, we are

[02:23] navigating an incredibly complex

[02:25] universe. There's many, many different

[02:28] ways that we have to respond socially,

[02:30] emotionally, physically, whatever we're

[02:32] doing. We have to cope with all of these

[02:34] factors that life is throwing at us.

[02:37] A person can juggle while reciting

[02:40] Shakespeare. Communicate abstract ideas

[02:43] while surfing.

[02:45] We can use imagination to achieve

[02:47] incredible things.

[02:49] We can do math and English and we can

[02:52] sing and dance and we can ride bikes and

[02:54] we can dream dreams. And that is a very

[02:57] general form of intelligence. And it

[02:58] also means that we can learn new tasks

[03:00] that we haven't been competent at

[03:02] before. And we are the only

[03:04] instantiation of a general intelligence

[03:06] on earth at this point in time.

[03:10] And basically we are talking about

[03:13] creating an artificial version of the

[03:15] general intelligence that we are.

[03:17] Probably before the end of this century,

[03:19] we will be able to construct computers

[03:23] or artificial intelligences which can go

[03:27] out on their own and develop lines of

[03:29] thought irrespective of any programming

[03:32] and which may in principle be more

[03:34] intelligent than we are.

[03:37] The reason why AGI is potentially so

[03:41] amazing, but the reason why it's also

[03:42] potentially terrifying from an

[03:44] existential threat standpoint is because

[03:46] it is the only time at which we're going

[03:48] to create something that is more

[03:50] powerful than us.

[03:52] And from there onwards, you know, life

[03:54] as we know it is going to be shaped by

[03:56] the more intelligent species on the

[03:58] planet.

[04:02] One of the most common misconceptions is

[04:04] that you only have to worry about AI if

[04:06] it has a body.

[04:08] Robots, hinges and motors, that's old

[04:11] technology. It's the intelligence itself

[04:13] that gives the power.

[04:15] >> I can count on my fingers.

[04:17] >> 2 3 4.

[04:20] >> I know many very smart people who think

[04:22] that superhuman intelligence will never

[04:24] arrive because it's somehow impossible.

[04:28] And usually it's because they think that

[04:30] intelligence is something mysterious

[04:31] that can only exist in human minds.

[04:36] I view that as just carbon chauvinism.

[04:39] I'm quite convinced that intelligence is

[04:41] actually all about information

[04:42] processing and that it doesn't matter

[04:45] whether the information is processed by

[04:46] carbon atoms in neurons in brains or by

[04:49] silicon atoms in computers.

[04:53] If it can happen, will it?

[04:56] Yes, it will. Because we have an

[04:59] incredibly powerful incentive as a

[05:01] species to keep making new technology

[05:03] because of course it makes us money and

[05:05] we scientists are curious.

[05:09] So when will it happen?

[05:11] Most AI researchers guess that it'll

[05:13] happen within the matter of decades

[05:15] which is probably in within the lifetime

[05:17] of most of us. Anyone who talks about

[05:20] climate change or anyone who talks about

[05:22] their retirement savings should talk

[05:24] about this also because this one will

[05:26] have a much bigger effect than either of

[05:27] those two. And I'm not trying to

[05:29] diminish the importance of climate

[05:31] change or retirement savings.

[05:34] >> So when a lot of people working in the

[05:36] field say that artificial general

[05:38] intelligence could be coming down the

[05:40] track rather soon, I think it's

[05:42] something to listen to. Look at

[05:43] Unlocking the Atom. In the early 1930s,

[05:46] >> the uranium atom was split, releasing

[05:48] its catastrophic force.

[05:52] >> People weren't expecting that. 15 years

[05:53] later, there would be these mushroom

[05:55] clouds arising into the sky.

[06:01] If you created an artificial

[06:02] intelligence that in a sense had access

[06:04] to its own source code and could tweak

[06:06] it, it could be very speedily that you

[06:09] get an artificial intelligence far more

[06:12] intelligent than the most intelligent

[06:13] human alive. It just improves so

[06:16] quickly. It's this wall where it's very

[06:18] hard to anticipate beyond that. And at

[06:21] that point, it's calling the shots

[06:24] because after that, we're no longer in

[06:26] the driving seat of history at that

[06:28] point.

[06:31] I think the reason we underestimate

[06:33] existential risk is because our brains

[06:35] evolved to deal with problems that we

[06:37] had seen many times before. You know,

[06:40] we have a very hard time feeling afraid

[06:42] of some hypothetical thing in the

[06:44] future.

[06:48] >> Here is another UFO bulletin. Because of

[06:51] the ominous situation, the president has

[06:53] ordered the Strategic Air Command into

[06:55] action. If we got an email from outer

[06:57] space from superioral aliensization.org

[07:02] saying they're going to show up in 40

[07:03] years, we wouldn't just sit here and

[07:05] twiddle our thumbs. We would freak out

[07:07] and do everything we could to make sure

[07:09] this didn't become a disaster.

[07:13] >> Was a great big huge yellow glow out

[07:15] there.

[07:16] >> Looked like a big cigar shape with light

[07:19] lit from both ends.

[07:21] It's quite likely that we will get the

[07:23] superior intelligence arriving on this

[07:25] planet on the same time scale. It's just

[07:26] that we're going to build it, which

[07:28] means what we do now has a huge impact

[07:31] on whether it's going to be the best

[07:32] thing ever or the worst thing ever.

[07:36] If some organization or entity develops

[07:39] super intelligence, they will probably

[07:41] try to do it in secret because otherwise

[07:44] everybody else is going to try to stop

[07:45] them or or steal it from them because

[07:48] it's going to give them total power over

[07:50] the world.

[07:52] >> If you look at who's developing things

[07:54] that look closest to AGI today, for the

[07:57] most part, it is really big companies or

[07:59] really big private organizations.

[08:02] >> We were the original micro computer

[08:04] software company. We got started in

[08:05] 1975.

[08:08] >> We started Apple because we wanted the

[08:09] product ourselves.

[08:11] >> Probably the headline dynamic is

[08:13] competition. Really

[08:17] >> and you're seeing a lot of competitive

[08:18] dynamics crop up between private

[08:20] companies and nation states.

[08:22] >> What our goal is is to be one of the

[08:24] world's strongest brands. and

[08:26] competition is this kind of constant

[08:29] pressure forward

[08:31] where folks think that it's very

[08:33] important to become the first one to

[08:35] deploy AGI.

[08:38] At the moment we've got a handful of

[08:41] companies, you know, mainly based out of

[08:42] Silicon Valley that control all the

[08:45] world's information, all the world's

[08:47] social contexts,

[08:49] all the world's buying habits, all these

[08:51] different things.

[08:54] you need gazillions of dollars in order

[08:57] to build those server farms and it's

[08:59] impossible for anybody to compete with

[09:01] those companies. So we have created

[09:04] juggernauts.

[09:07] If you have a digital elite and you can

[09:09] build an artificial intelligence and you

[09:11] go okay we want to use this to

[09:13] manipulate these people's data to make

[09:15] them to change how they're voting. You

[09:17] have created private companies

[09:20] that have more control than governments.

[09:25] If you feel it sounds outlandish that

[09:28] there could be some sort of algorithm

[09:30] out there that's so powerful that it can

[09:32] go and do things that really screw

[09:34] people over, we already have that. It's

[09:36] called the corporation.

[09:40] There have been many examples of tobacco

[09:41] corporations, asbestous corporations,

[09:43] and so on who did things which in

[09:45] hindsight were really not good for

[09:47] humanity.

[09:49] Not because the people there were evil,

[09:51] but because the corporation itself had a

[09:53] sort of higher level of intelligence.

[09:57] It had its own goals to make money. And

[10:00] it then coerced all the people within

[10:01] the organization to say the right thing

[10:03] and do the right thing.

[10:04] >> Yes. First smoking that you're bound to

[10:06] like. You just can't beat a lucky

[10:07] strike.

[10:08] >> I suppose if I'm going to die, I might

[10:09] as well do it for my country.

[10:11] >> That's the spirit. This should be a

[10:13] wakeup call that if we're not careful,

[10:15] we can do that on a much more massive

[10:17] scale with artificial intelligence.

[10:21] Big tech industry need regulations just

[10:23] as much as the car industry needed to be

[10:25] forced to put seat belts in cars.

[10:33] Even though the car industry pushed a

[10:35] lot against seat belts,

[10:37] now they're perfectly fine with it.

[10:44] As a matter of fact, because there were

[10:46] seat belts and other safety features,

[10:48] people started to lose this fear of

[10:50] driving and they were able to sell much

[10:52] much more cars.

[10:57] We don't really understand as

[11:00] individuals

[11:01] how artificial intelligence is already

[11:04] shaping our lives.

[11:07] We don't really know who controls this

[11:09] stuff, what they know about us, what

[11:11] they're doing with the information.

[11:13] >> Every time I go on the internet, a

[11:15] profile is being built of of me and my

[11:18] consumer desires and other kinds of

[11:20] desires and so on.

[11:21] >> Now, we've got voice assistants in

[11:23] people's homes. And a voice assistant is

[11:26] basically the box which is listening in

[11:28] all the time, but you don't know it's

[11:30] listening.

[11:32] You would expect that if you're having a

[11:34] private conversation, you expect it to

[11:35] be private.

[11:37] We don't really see the value in what

[11:40] they're taking from us.

[11:43] It's like, oh, someone is stealing the

[11:45] carbon dioxide that I breathe out as I

[11:47] walk around. Well, you know, it's not of

[11:50] that much value to me anyway. I just

[11:51] produce it as I walk through the world.

[11:54] We need to see our data as a resource.

[11:58] Producing this thing almost as labor.

[12:00] That's something you produce by working

[12:04] and living your life and having the

[12:06] relationships that you do. And that's

[12:07] not the property of someone else.

[12:14] >> Data rebellion, the radical arm of the

[12:17] anti-data surveillance movement, has

[12:19] defied government bans with a day of

[12:22] global protest. The aim is to shut down

[12:25] dozens of cities around the world. The

[12:27] data trade is now a trillionoll

[12:30] industry. The group's message is that

[12:32] personal data is a resource owned by

[12:35] individuals and is not free for the

[12:37] taking with big tech companies and

[12:39] governments accused of breaching basic

[12:42] human rights by capturing personal data

[12:44] without consent or compensation. While

[12:47] cities around the world grind to a halt,

[12:49] some say the group's tactics are testing

[12:52] public tolerance for disruption. In a

[12:54] world that's become accustomed to

[12:55] surveillance capitalism,

[13:02] technology has really disrupted this

[13:06] balance between knowing and being known.

[13:10] The moment you invade the private space

[13:12] of a person by understanding their

[13:15] beliefs, their attitudes, how they vote,

[13:17] how they shop, gives you incredible

[13:20] power. And who we want to entrust that

[13:24] power to is is very important.

[13:27] Jordan, can you still help master?

[13:31] >> We're not far from the point where you

[13:32] could build an AI system that is the

[13:36] perfect Orwellian big brother that reads

[13:38] every text message and email and listens

[13:40] to every phone call and watches every

[13:42] surveillance camera and enables a

[13:44] permanent Orwellian dictatorship that

[13:46] can never be overthrown.

[13:49] Even in the West, we're actively

[13:51] building this thing now.

[13:54] If we ever get a leader who wants to

[13:55] switch it on, all they have to do is is

[13:58] flip the on switch.

[14:02] >> I was born in Romania and I lived for

[14:06] part of my life in a dictatorship.

[14:11] >> So for me, thinking about dystopia is

[14:14] something that I sort of easily relate

[14:16] to.

[14:17] [Applause]

[14:17] [Music]

[14:19] So, I have a strong history of what it

[14:22] means to live under surveillance because

[14:26] one of the realities of my childhood was

[14:29] that the government knew everything.

[14:35] And I think all of this could have been

[14:37] so much more amplified by technology.

[14:40] And that scares me

[14:44] because for us it was mostly, oh, the

[14:46] neighbor is going to inform the secret

[14:49] police,

[14:50] [Music]

[14:52] but now it can be your Alexa or your

[14:55] surveillance camera in the lobby.

[15:00] In the past, whenever dictatorships have

[15:03] been overthrown, it's always been

[15:05] because those who did the overthrowing

[15:08] had a certain amount of privacy and were

[15:10] able to organize.

[15:16] >> But if there are surveillance cameras

[15:17] everywhere and every single phone call

[15:19] and text message and communication you

[15:21] do is recorded and analyzed by an AI,

[15:24] how are you possibly going to organize

[15:26] some kind of overthrow of the powers of

[15:27] be?

[15:35] [Applause]

[15:41] [Music]

[15:44] If you're not worried about AI ever

[15:46] becoming so smart that it can take

[15:48] control over us humans, then I would

[15:50] like you to take a moment and just

[15:53] envision your least favorite political

[15:55] leader on this planet.

[15:58] And imagine that they come into control

[16:01] of this AI that's so powerful that they

[16:03] can dominate all of Earth with it. How

[16:06] do you feel about that?

[16:10] So the first thing we should worry about

[16:11] is not AI itself,

[16:15] but humans just using it to do horrible

[16:17] things to other humans on a hitherto

[16:20] unprecedented scale.

[16:22] people will somehow feel that robots

[16:25] will turn upon them. Now, this I don't

[16:28] think is very likely. However, in a very

[16:31] abstract sense, this is precisely what

[16:34] may happen.

[16:36] Some of my colleagues want to build

[16:38] super intelligence and just keep it in a

[16:40] box, disconnected from the internet,

[16:43] under human control. Now, aside from any

[16:45] moral qualms that you might have about

[16:47] enslaving a superior mind, you might

[16:49] worry about it outsmarting us and

[16:51] breaking out.

[16:54] If you were imprisoned by a bunch of

[16:56] fouryear-olds, do you really think you

[16:58] couldn't find a way of tricking them

[17:00] into somehow letting you escape?

[17:05] It's not necessarily a bad thing being

[17:08] surrounded by more intelligent entities,

[17:10] right? You and I have both already

[17:12] experienced that.

[17:14] They were mommy and daddy when we were

[17:16] tiny, right? And why did that work out

[17:18] well? Because their goals were aligned

[17:21] with ours, right?

[17:23] We don't kill ants in our living room

[17:25] because we're evil ant haters. It's just

[17:28] that the goals of the ants didn't align

[17:30] with our goals. We can be equally sure

[17:33] that if we fail to align the super

[17:35] intelligence's goal with ours, we're in

[17:37] the position of those ants.

[17:39] So I think the key to a happy

[17:41] coexistence with AI is not to get into

[17:43] an adversarial relationship, but to just

[17:45] make sure that their goals are aligned

[17:47] with ours.

[17:51] This challenge of aligning AI's goals

[17:53] with ours has two parts.

[17:56] First of how do you actually even

[17:58] explain to another intelligence what our

[18:00] goals are and how do you make sure it

[18:02] adopts those goals and and retains them

[18:04] as it gets ever smarter. And then

[18:06] there's a separate question of what goal

[18:07] should you put in?

[18:09] [Applause]

[18:12] >> Every time we talk about AI, we learn

[18:15] more about what it means to be human.

[18:18] >> Why do young children fly into rages in

[18:21] situations like this? But if we want to

[18:24] talk about aligning AI values with human

[18:26] values, we have to think deeply about

[18:29] what those human values are and learn

[18:32] that you can never expect homogeneous

[18:36] human values.

[18:37] [Music]

[18:46] Essentially

[18:52] [Music]

[18:56] the question is is AI culturally and

[19:00] ethically neutral and the answer to that

[19:01] is it's not

[19:04] anything that is man-made will be imbued

[19:07] with the values or at least the

[19:10] priorities of the people in society that

[19:13] made it. We do tend to react more

[19:16] favorably to people that look like us,

[19:18] that speak our language, that share our

[19:20] values. That's the presumption that our

[19:23] brains make. And then the byproduct of

[19:26] that is that where there is difference,

[19:27] there tends then to be a bias. People

[19:30] pretend that technology is a great

[19:33] equalizer and leveler. But I think we

[19:35] need to be quite careful about that to

[19:38] understand that technology has values

[19:40] attached to it.

[19:44] Rioting across the US today as police

[19:47] accidentally shoot an unarmed black

[19:49] woman in Chicago.

[19:50] >> We have a woman in her 80s shot in the

[19:53] head on the far south side.

[19:54] >> The yet unnamed 80-year-old was

[19:57] incorrectly identified by powerful AI

[19:59] predictive policing algorithms. The

[20:02] police were tasked with arresting an

[20:03] offender at a time and location

[20:05] generated by the algorithm. An

[20:08] investigation into the incident is

[20:10] underway. It's been going on for too

[20:11] long, man. I feel like this should stop.

[20:13] This whole racist stuff, white cop

[20:15] killing a black person.

[20:16] >> Predictive policing has been criticized

[20:19] for being racist, unfairly targeting

[20:21] groups, ethnicities, and locations.

[20:24] Despite police claiming a massive

[20:26] reduction of crime in the last decade,

[20:28] police can now track potential criminals

[20:30] using the internet, phone history, CCTV

[20:34] face recognition combined with new

[20:36] technology that examines the emotional

[20:38] state of potential perpetrators.

[20:42] But as protests erupt on the streets

[20:44] tonight, critics are warning predictive

[20:46] policing reduces the accountability of

[20:49] law enforcement with important life and

[20:52] death decisions. now automated by AI.

[20:59] >> How do we ensure that AI is compatible

[21:03] with us with the us that we want it to

[21:06] be compatible with? Because we are not

[21:07] the same. Like, do I want an AI that is

[21:13] serving me when I want to binge eat or

[21:16] do I want that AI to serve the me that,

[21:19] you know, wants to go on a diet and

[21:22] wants to be healthy?

[21:24] >> There. Yeah.

[21:25] >> So, people have a lot of problems

[21:28] defining what it is that we even want

[21:30] from AI and how we can make it

[21:32] compatible with us and which us do we

[21:37] mean. Humans

[21:41] often have pretty unstable preferences.

[21:43] So, we're quite uncertain about what we

[21:45] prefer. Our preferences change as we get

[21:47] older.

[21:50] Even at a societywide scale, our

[21:52] preferences about what seems right or

[21:54] wrong change. So, you can imagine that

[21:57] if we anchor on our preferences now,

[21:58] then future generations might think that

[22:00] it's fine to burn fossil fuels, for

[22:01] example. And that would be a huge

[22:03] problem.

[22:08] Many of my friends tell me that it's

[22:10] futile to even talk about this because

[22:12] people in different cultures and even

[22:14] from different political parties have

[22:15] such different goals. You're not going

[22:16] to agree on anything. But that's

[22:18] actually nonsense. Most humans agree

[22:20] that they would rather not have the

[22:21] species go extinct, for example. That's

[22:23] not so obvious to a machine whether

[22:25] that's a good thing or a bad thing. You

[22:26] have to teach it that.

[22:29] Not long ago, Andreas Lubitz told his

[22:32] civilian aircraft he was flying over the

[22:34] Alps to go down to 50 m altitude.

[22:38] And you know what the computer said?

[22:42] It said, "Okay." And everybody died.

[22:47] Nobody had even thought about teaching

[22:50] this AI that it should never fly in

[22:52] fixed objects ever. the vast majority of

[22:55] the sort of kindergarten ethics that all

[22:58] humans virtually agree. We still haven't

[23:00] even gotten around to putting into our

[23:01] machines. This is low tech. We have the

[23:04] technology to do this. It's not hard. At

[23:06] this stage of life, every activity is a

[23:09] lesson in itself. Every new experiment

[23:12] in play extends the horizon of their

[23:14] understanding.

[23:16] >> So, let's start by putting in these

[23:18] kindergarten ethics into any machine

[23:20] smart enough to understand this. And

[23:22] then as machines get smarter, we'll be

[23:24] in the right mindset to keep putting in

[23:26] ever more sophisticated goals.

[23:28] >> I aliance to the flag of the United

[23:34] States of America.

[23:35] >> I'd love to have a governing

[23:37] constitution that says, "Look, here are

[23:39] the bounds and here are the things that

[23:42] we care the most about preserving.

[23:44] Here's our expectations for how we will

[23:47] be making decisions that have this

[23:48] amount of magnitude and this amount of

[23:50] flowover effects into the rest of the

[23:52] world.

[23:54] People look to constitutional norms to

[23:57] answer the question, what do most people

[23:59] think is morally right?

[24:03] So, I think it would be a very good

[24:05] thing for us to crowdsource a

[24:08] constitution for humanity.

[24:11] There's something about the wisdom of

[24:12] crowds.

[24:14] I don't think it would be all that

[24:16] difficult um to run an expert

[24:19] facilitated crowdsourced

[24:21] constitution-making exercise that could

[24:23] give us some really first rate rules

[24:26] about what we should present to the

[24:29] super intelligent AI when it arrives.

[24:32] >> If you look to the Americans, the

[24:34] founding fathers, the constitution was a

[24:38] very influential document. It's had

[24:40] these vast rippling effects on history.

[24:43] You know, you can discuss, you can

[24:44] debate what they are and how good they

[24:46] are. But the idea that that constitution

[24:48] has had this propagating effect, you

[24:51] can't deny that.

[24:54] What I think would be a thing to, you

[24:56] know, put as the centerpiece of uh any

[24:59] kind of constitution for the future to

[25:01] pass down would just be that ability to

[25:04] correct our errors.

[25:08] That ability for intelligence to realize

[25:11] when it's gone wrong and then to make up

[25:13] for it.

[25:16] We know that humanity is committing

[25:17] moral atrocities right now, but there

[25:20] are also ones that we probably aren't

[25:21] even aware of. So what I would put as

[25:24] the centerpiece of any kind of

[25:26] constitution for the future would just

[25:29] be to stress that ability to correct our

[25:32] errors and realize where we're going

[25:34] wrong and make up for the moral

[25:36] tragedies of the past

[25:38] and also the future.

[25:43] [Music]

[25:47] I can't help laugh sometimes when people

[25:49] say, "Why would you possibly worry about

[25:51] a robot killing you?"

[25:56] And then there are all these companies

[25:58] online running advertisements for robots

[26:01] that kill you.

[26:04] >> Right now, people are being killed by

[26:06] drones using AI.

[26:09] One big thing that most governments are

[26:11] by design incentivized to do is to shore

[26:14] up both military power and economic

[26:16] power,

[26:18] which means that it's going to want to

[26:20] use something like AGI to shore up its

[26:22] military and economic power as well.

[26:25] God forbid we integrate AI into the

[26:27] nuclear command and control system, but

[26:28] that would be one example of how it

[26:30] could go drastically wrong.

[26:35] The US, Russia, China benefit from

[26:39] powerful weapons that are really

[26:41] expensive that only they can afford.

[26:44] If you want to kill 10 million people

[26:46] with guns, you basically need millions

[26:49] of people to carry those guns for you,

[26:51] right?

[26:52] >> But we have something much bigger.

[27:02] Your kids probably have one of these,

[27:04] right?

[27:06] >> Not quite.

[27:07] >> If we get an arms race in lethal

[27:09] autonomous weapons for our slaughter

[27:11] bots, then we're going to have these

[27:13] incredibly powerful weapons that weigh

[27:15] as much as a smartphone, cost as much as

[27:17] a smartphone.

[27:21] This would be much, much worse than

[27:23] guns. Because if you want to kill that

[27:25] many people automatically,

[27:27] a handful of people can just press a

[27:29] button. This is also like the ultimate

[27:31] obedient army in the future.

[27:35] [Music]

[27:41] Lowering the threshold of what it means

[27:43] to kill a person by just delegating all

[27:48] action and all responsibility to a

[27:51] machine is something that you know

[27:54] doesn't just lead to the death of a

[27:57] person but it leads to moral death.

[28:04] [Music]

[28:12] We have completely relinquished not just

[28:15] our responsibility but our actions and

[28:17] our decision making.

[28:20] It leads to this whole new level of

[28:22] dehumanization of our potential victims

[28:25] in the future. They're not even going to

[28:29] be considered by a human mind. They are

[28:32] going to simply be selected and targeted

[28:36] by something that won't have any

[28:38] understanding of what it means to be

[28:41] human and the importance of human life

[28:45] and the frailty of our experience.

[29:01] Machines will never ever be able to

[29:04] fully simulate humans because they will

[29:08] never understand human suffering.

[29:11] [Music]

[29:18] >> They will have some sort of alternative

[29:20] type of life and experience. They may

[29:22] experience things which are difficult.

[29:24] They may experience, you know, virtual

[29:26] pain or who knows what, but those

[29:29] machines will relate on their terms.

[29:33] If we're going to cooperate with

[29:35] intelligent machines in the future, then

[29:38] we need to understand the basis of human

[29:40] cooperation

[29:41] because human cooperation is the most

[29:43] powerful force in history

[29:46] [Music]

[29:48] and it has enabled us to do all kinds of

[29:51] incredible things which are actually the

[29:52] result of human cooperations.

[29:56] You know, to make medicines, put people

[29:58] on the moon. So it's not just one human

[30:00] brain, it's all these different human

[30:02] brains working together.

[30:04] >> In order to create a machine that you

[30:06] can really cooperate with. The more

[30:08] humanlike it is, the more intuitive and

[30:12] natural it is to communicate.

[30:17] So I was wanting to build a a brain

[30:19] model and put it in a in a character.

[30:23] Then I was thinking, what face should I

[30:25] use?

[30:28] You know, I was spending so much time

[30:29] with my real daughter that I thought,

[30:31] "Oh, that's perfect." Because

[30:33] metaphorically, she's representative of

[30:36] the level of learning that we're looking

[30:38] at.

[30:40] Most of our learning when we're young is

[30:43] social learning. We're learning from,

[30:45] you know, parents or caregivers about

[30:48] everything to do with the world.

[30:50] [Music]

[30:53] One of the goals of Baby X is really

[30:55] trying to build a truly autonomous

[30:57] model. So that touches on the notion of

[31:00] free will.

[31:01] >> What's this, baby? What's this? What's

[31:03] this?

[31:04] >> Good girl. Now, see if she knows what

[31:06] the word is. Okay, baby. Look over here.

[31:09] Okay. What's this? What's this?

[31:13] >> Good girl.

[31:14] >> The model is actually kind of deciding

[31:16] what it wants to do.

[31:18] >> Bobby, baby, over here. What's this? And

[31:20] it can be distracted by different things

[31:22] and may would not want to talk to you

[31:24] right now.

[31:25] >> Over here. Over here.

[31:27] >> And one of the most profound mysteries

[31:29] is the mystery of consciousness.

[31:32] As all these elements come together, can

[31:34] we get a sense of something that is

[31:37] actually aware of its environment? Where

[31:40] is the doggy? Doggy.

[31:45] Good.

[31:51] Humans have always been merged with

[31:52] technology.

[31:54] There are very good arguments that

[31:56] technology made us human.

[32:01] Almost everything we do every day is

[32:05] modulated through technology.

[32:09] Our phones are in a sense an extension

[32:11] of ourselves. And humanity is going to

[32:14] continue its thing of being intimately

[32:17] entwined with technology.

[32:19] [Music]

[32:21] That is going to become what we now

[32:23] would see as uncomfortable or strange or

[32:25] bizarre.

[32:28] Imagine

[32:32] that I was running 100 times faster and

[32:34] I was able to remember a lot more in the

[32:37] world, understand a lot more in the

[32:38] world, not just physically do more

[32:40] things but intellectually do more things

[32:43] and that is totally liberating in a way

[32:46] that is completely wild to the

[32:48] imagination as it is now.

[32:51] We will have human beings who will make

[32:54] more and more use of artificial organs

[32:58] of metal and plastic, artificial hearts,

[33:00] artificial kidneys, artificial lungs. In

[33:03] short, we may have a society in which

[33:05] human beings will drift away from the

[33:07] total organic toward the metal.

[33:11] Transhumanism is this idea that we can

[33:14] use technology not just to change and

[33:17] conquer

[33:18] the natural world and the environment,

[33:21] but also that we can use technology to

[33:23] change ourselves and to enhance

[33:25] ourselves, to prolong our lives, to make

[33:27] ourselves better people, to enhance our

[33:30] cognitive capacities and so on.

[33:34] >> At the center of transhumanism is a kind

[33:36] of anxiety I think which is the anxiety

[33:38] of obsolescence.

[33:40] [Music]

[33:44] the idea that we're going to sort of

[33:46] price ourselves out of the market

[33:47] intellectually by developing these

[33:49] machines that are going to be able to

[33:50] outthink us and outmaneuver us

[33:54] and we'll be basically irrelevant and

[33:55] maybe endangered.

[33:57] In order to avoid that, we need to merge

[34:00] with technology. We need to become the

[34:03] machine.

[34:07] We're not just here to do high level

[34:09] maths. I think there's a lot more going

[34:12] on within humans than solving problems

[34:15] and thinking rationally. And that seems

[34:19] to me to be a basic misunderstanding of,

[34:22] you know, what's valuable about being

[34:25] alive.

[34:28] Ultimately, transhumists tend to talk

[34:31] about the eradication of death through

[34:33] technology.

[34:35] And despite the veneer of like extreme

[34:38] futurism and sort of high-tech, the idea

[34:41] that we might be able to evade death is

[34:43] as old as humanity itself.

[34:49] [Music]

[34:52] [Laughter]

[34:55] So many of the transhumanists who I talk

[34:57] to are very invested in this idea of

[34:59] mind uploading of extracting the

[35:01] consciousness of an individual and

[35:03] putting it onto a software essentially.

[35:07] So some people might want to be robots

[35:10] >> moving uh I guess an average of about 8

[35:12] km an hour.

[35:13] >> Some people might want to upload

[35:14] themselves into like a Mars rover or

[35:16] whatever. Some people might want to live

[35:19] as like free floating consciousness in

[35:22] the cloud extending infinitely

[35:25] throughout time at the most extreme

[35:27] level where you just you don't die. But

[35:30] if you're a free floating consciousness

[35:33] extending infinitely through space,

[35:36] are you really anything at all? What are

[35:39] you? Are you a god or are you nothing?

[35:42] [Music]

[35:46] It's a branching in what evolution has

[35:49] been so far. To date, the pressures of

[35:52] evolution are natural selection. And

[35:54] this is the point in time when we will

[35:56] choose as humanity to branch off and

[35:58] choose a different selection pressure,

[36:00] which is our intelligence.

[36:02] That's why this moment is so important.

[36:04] I mean, if it does go in this direction,

[36:06] then we will be the final instantiation

[36:08] of humans. And that's a massive

[36:09] responsibility.

[36:11] [Music]

[36:16] Most people would look at the

[36:17] agricultural revolution and the

[36:18] industrial revolution and say, "Look,

[36:20] that was a total step change in our

[36:22] quality of life.

[36:26] [Music]

[36:28] We just didn't prepare enough for both

[36:31] of them."

[36:36] And so there were these costs that we

[36:38] paid.

[36:40] entrenched inequality is one example of

[36:42] a thing that the industrial revolution

[36:44] exacerbated in a really dramatic way.

[36:47] And so I'm not going to downplay how

[36:49] bumpy I think the transition is going to

[36:51] be. But if you look on the other side of

[36:53] that transition, we're effectively in a

[36:55] world where running a value creating

[36:57] economy doesn't need to be done by

[36:59] humans anymore.

[37:02] When you replace a worker by a machine,

[37:06] the income that used to go to that human

[37:08] who wasn't very rich instead went to the

[37:10] owner of the machine. That very

[37:12] naturally leads to a situation where the

[37:14] rich get richer and vice versa. The

[37:16] obvious solution to this, if you like

[37:18] equality, is to just make sure the

[37:19] government collects enough tax from this

[37:21] ever growing pie that it can make sure

[37:23] that everybody gets better off because

[37:25] there's almost no limit to how much

[37:27] wealth we can produce with the help of

[37:30] of machines.

[37:33] >> Work is a huge part of our identity.

[37:35] >> Mr. Brown's line.

[37:37] >> For a good part of human history, we've

[37:40] really valued people based on what they

[37:42] can produce.

[37:44] The more we develop technologies that

[37:46] replace us in in some of these work

[37:49] environments, the more we need to think

[37:51] about what is the purpose of the human

[37:55] life and of the human being.

[37:57] >> Jobs give us income. They give us

[38:00] purpose. They give us friends. If we can

[38:03] find a way of getting all those three

[38:05] without jobs, I don't think we're going

[38:07] to miss the jobs very much.

[38:10] The key thing is to really think hard

[38:12] though about what sort of society we

[38:14] want to create so that the people will

[38:17] still feel they have that sense of

[38:19] purpose. What sort of society would we

[38:21] really be excited and living in?

[38:24] >> And I think people don't spend nearly

[38:26] enough time thinking about positive

[38:29] visions for the future with technology.

[38:34] [Music]

[38:37] We may well be enlightened in ways which

[38:39] we don't even understand at the moment

[38:40] because people don't have that capacity.

[38:44] Imagine being able to chat with a super

[38:46] intelligent being about some of the

[38:48] mysteries of life.

[38:54] Imagine talking to a machine about well

[38:56] what was it like flying around the

[38:57] internet? What was it like when you were

[39:00] a satellite going around Jupiter?

[39:04] And we will get answers that no person

[39:07] has ever been able to give before.

[39:11] There are a lot of different things that

[39:14] AI could do for us. Imagine you could

[39:17] replicate incredibly intelligent humans.

[39:20] You could deploy them to solve any

[39:21] problem overnight while you're asleep.

[39:23] Like it's that fast. So for example,

[39:26] poverty and being able to redistribute

[39:28] resources in appropriate ways, climate

[39:30] change, climate technology,

[39:33] decarbonizing the economy, and then you

[39:35] can graduate to just harder technical

[39:38] tasks like for example, how do we get to

[39:40] different planets?

[39:43] [Music]

[39:55] The mission that I'm the most invested

[39:57] in is a mission to Venus to search for

[40:00] life. And that's really uh you know a

[40:03] full circle for me standing back um as a

[40:05] young child looking up at the stars on a

[40:07] really clear night in the bottom of the

[40:09] South Island with my father and the

[40:12] realization as he was pointing out to me

[40:13] those stars that that those stars have

[40:15] planets around them and those planets

[40:18] could have somebody just like me looking

[40:20] back and asking the same question.

[40:25] [Music]

[40:28] And that was what really, you know, got

[40:30] me got me interested in space from the

[40:32] very early beginnings is that yearning

[40:33] to understand, you know, are we the only

[40:35] life in the universe or or not.

[40:38] Always promised myself if I ever had the

[40:40] opportunity to go and do do something

[40:41] like that that I would do it.

[40:45] There's a few places in our solar system

[40:47] where it's been long hypothesized that

[40:49] there could be life. And in the clouds

[40:51] of Venus, there's a really interesting

[40:52] spot. It's about 50 km altitude where

[40:55] the environment is temperate enough for

[40:57] primitive forms of extreme life to

[41:00] survive.

[41:01] If I got one question in my entire life

[41:04] and somebody would give me the answer,

[41:05] my question would be, are we the only

[41:07] life in the universe?

[41:10] People have this very strong intuition

[41:12] ahead of evidence that if it looks like

[41:15] somewhere could have life, then it will

[41:17] have life.

[41:19] It's not based on evidence or scientific

[41:21] reasoning. It's based on wishful

[41:23] thinking.

[41:30] In the '60s, SETI, the search for

[41:32] extraterrestrial intelligence begins.

[41:38] >> Originally, people had very optimistic

[41:40] hopes that they were going to log into

[41:41] this uh, you know, populous thronging

[41:44] universe filled with conversation.

[41:48] They just heard silence back.

[41:53] There's lots of suns bathing uninhabited

[41:55] inorganic planets with loads of energy

[41:57] and nothing interesting is happening.

[42:01] The cosmos seems quite silent.

[42:05] So we can't take for granted anymore

[42:08] that the universe is filled with other

[42:09] intelligences.

[42:11] That's not to say that we are the wonder

[42:13] of the whole universe, but as far as we

[42:16] know, we are the only thing that can

[42:18] find the universe wonderful.

[42:21] Life is incredibly cosmically rare, and

[42:24] humans, it's the only thing that could

[42:26] spread it elsewhere.

[42:29] We have the potential to be this spark

[42:31] plug that could create this transition

[42:33] from a currently inorganic universe into

[42:35] one that's filled with life

[42:38] and potentially the richness of

[42:40] consciousness.

[42:43] [Music]

[42:47] [Laughter]

[42:48] [Music]

[42:53] We've realized not only that we're not

[42:55] insignificant,

[42:57] but that we were dramatically more

[42:58] significant than our ancestors ever

[43:01] dreamt of.

[43:04] I find this an incredibly empowering

[43:06] insight that science has given us.

[43:16] We know this planet can be a very nice

[43:19] place to live for hundreds of millions

[43:20] of years unless we screw it up.

[43:24] And if we want to live longer than that,

[43:26] there's absolutely no reason we can't

[43:28] use artificial intelligence to go to

[43:30] other solar systems and other galaxies.

[43:35] Life can flourish for billions of years

[43:38] throughout the cosmos.

[43:40] [Music]

[43:49] Either the future is going to be awesome

[43:51] with life flourishing for billions of

[43:53] years on Earth and probably beyond or

[43:57] we're going to go extinct probably

[43:58] within the century. I think it's never

[44:01] been more motivating to be a human to

[44:04] help contribute to making sure we make

[44:06] the right choice here.

[44:17] [Music]