Saturday, 7 January 2017

Ideas for a "Superintelligence" TV show.

This is the plan for a TV show, probably on something like Netflix, called 'Super-intelligence'. See my previous post here.

It’s a near-future science-fiction series (set about 35 years from now). It's about the growth, control and comprehension of Artificial Intelligence. This is not a Skynet series, it's not about just one monodominant AI that "changes everything, it’s not about AI instantly trying to take over the world and kill everyone. (Except that might kind of nearly happen in one episode). It’s about the growth of a technology, the creation of an entirely new form of life, the way human societies and systems will adapt to deal with this, the staggering opportunities and serious dangers, both material and moral of the world 'strong' AI's will being about.

.................................................

So it's set in a nameless (unless I can think of a really good name) government or quasi-government department. This is an investigation and action group dedicated to dealing with AI-related problems or issues. Essentially, whenever someone gets worried about what an AI is doing, or the government gets worried, this is who they call to investigate and resolve the situation. And this is vague in an 'X-Files' way so we have an excuse to dump out protagonist into a lot of different kinds of situations. And no, apparently no-one else is doing this kind of job. (Until series 4 or something when we learn the CIA or the Chinese or someone were also doing this kind of job and now we are in a conflict.

The group is lead by (and so far, pretty much just is,) the Main Guy, played, probably, by Jeffery Ford, (sorry about the typecasting Jeffry). They work in a big complex, something like what most of us imagine Langley to be like, or GCHQ. A huge militarised office building. You have to go through loads of security and checks to get to the middle. But in the middle, maybe in some open space, possibly some parkland or something, separate from the rest of the complex, is a building. It might be a cottage or just a set of porta cabins. It’s possible its base is raised off the ground on stilts of some kind so people can look underneath it. All around the building is a gigantic net, like the web of an aviary, this is a faraday cage surrounding the whole place, before anyone goes in or out of the cage they have to give up any technology of any kind and submit to a full scan.

Once inside the box, (I suppose we'll call it "The Box", with some ironic effect as we'll see later) no tech is used, people write things down on physical paper with pencil, pen or via typewriter using carbon copy. There is a long-standing debate about the validity of a photocopier. Files are kept on physical media only, there is one phone line and it is an old style analogue phone, its forbidden to keep a digital record of any kind regarding anything classified 'for the box'.

Even in the field, they are supposed to make minimal use of phones, preferably not even carrying one, not to email or text, to drive their own cars. This is remarked on during the series as 'self-driving' or driving your own car, is coming back into fashion and is actually illegal, semi-legal or just frowned upon in some states and areas as its regarded as unsafe. Our main guy carries a special pass which marks him as a government employee in case he is stopped by police for driving his own 50 year old car without even the basic safety elements added to all current cars by law. He also has a backseat full of maps, which is remarked on by people in the series. He might even get into a car crash, something amazing to anyone outside the group and almost incomprehensible to current police as it simply never (or almost never) happens any more.

.................................................................

One of the primary things the group does is check on and supervise 'Boxing' methods. The means by which developing AI's are 'boxed' off from the environment and kept safely away from the general world data network.

This is easiest to do in situations where people are actively trying to develop an AI for a specific purpose and know that’s what they are trying to do. In this case its kinds like a Michael Crichton book, there's a complex somewhere in the desert with a bunch of security rituals and maybe even its own |faraday cage and power supply. These 'pure math' AI's are the most intelligent, adaptive and alien to the human mind, they are also the most powerful and useful so people keep trying to make and control them.

(A major theme in the series is humanity continually pushing against the limits of "safe" AI because whoever can do it and control it gets access to a theoretically staggering level of power.)

Other common types of AI are emulations, copies of scans of human brains, and emergents, unpredictable amalgamations or 'growths' of one or more 'soft AI's tasked with a complex problem, like a city-running traffic network, exceeds it programming and starts expanding, colonising extra processing power, not yet fully 'awake' but edging towards sentience. What happens when two or more highly intelligent but mono-tasked bots start talking to each other and altering their own programming in response?

Emergants are weird and unpredictable and one of the most difficult questions is how to box them or shut them off when they are responsible for something that might be really important and has to be monitored continually, like say, a fusion reactor.

Everyone hates dealing with emulations as they are most obviously human and it definitely feels like Mind Crime when they turn them off, but they are also relatively easy to deal with in most cases as they have human or pseudo-human motivations and are generally quite a bit less efficient and intelligent than purpose-built AI's even though they can run at digital speeds. They are also the ones most likely to try human-style crimes, like murdering someone.

One of the themes of the series is the growing fear and disaffection of the main protagonist as they come to think that what they are doing by shutting down nascent AI's is actually killing living things, but its legal as they have no rules governing them, and that even when they "box" an AI they are, in some sense, creating a slave race.

One long-term theme could be a potential amendment to the constitution re-defining alive as meaning, or including 'self-aware' and all the social and political conflicts leading up to that, and all the complex legal, moral and structural changes after that


..............................................................

The effects of increased genetic comprehension will also play a factor. In this world we can analyse DNA well enough to come up with reasonable predictions to do with personality type, mental illness, general talent paths, life span, social life and political/social affiliation, and its technically illegal to judge someone on their DNA but having knowledge of your own DNA is legal and common, so there is a thriving 'underground trade' in having someone’s 'code', especially for employers of top industries and politicians.

Some people are 'code-radical' and just put their entire genome online on their social networks so anyone can see it, others are hyper-conservative and refuse to allow their code to be read, even if medically necessary, some people refuse to allow their children’s code to be read at all, others have discussions about when it’s reasonable to give your child access to their own code and how to introduce the information.

Plus certain government departments have official sanction to make code a part of their assessment process for strategically relevant jobs, because other governments are doing the same thing and you can't risk having anything less than the most-optimal people in certain strategic positions.

So how much your genetic endowment defines you and what you do with that knowledge is a major new source of social tension, in some ways overtaking, or consuming, anxieties about race. There are new social groups and social allegiances forming because of this, it's something presidential candidates speak about. Should we know the Presidents code? What if one candidate makes theirs public and infers the other candidate has a 5% probability shift towards mental ilness over 50 years and they are covering it up?

...........................................................

The central drama of the show is about a guy talking to a machine and trying to work out if it is.

A - alive
B - dangerous

The more alive it is the more it's likely to be dangerous but the more alive it is the more immoral it is to turn it off. Also, Whatever it's doing is valuable or will be so there are always people who don't want you to turn it off. Also the easier it is to talk to the easier it is for it to manipulate you if it does, in fact, have SUPERINTELLIGENCE. The problem isn't just imagining these machines, its expressing them through a dramatic context without it being Kirk talking to a box with flashing lights every episode.

So;

WAYS TO TALK TO A POSSIBLY SUPERINTELLIGENT MACHINE (SO IT IS INTERESTING IN A TV SHOE DRAMA)

1.      It has a voice and it’s on the phone.
2.      It's in a robot body and it has a voice but it can't move about much.
3.      It has a full 'Ex Machina' robot body that its (probably) inside and it can move about plus its hot.
4.      It has some kind of obviously-robotic body but doesn't know it’s an AI and you can't let it find out.
5.      It's a Deckard, simulated human personality and high grade simulated body and doesn’t know it’s an AI. (Turning it off will feel even more like murder).
6.      It has a robot body designed for something else (JCB, Roomba) that doesn't speak but can move.
7.      It has cybernetic control of a piece of infrastructure (traffic system, water system, rail system) and you can read its intentions from what it does there, like the way it moves traffic around.
8.      It has access to a simulated person, like a Serkis person, that it can put up on a screen.
9.      It's plugged into someone’s brain, or just the ear, like a cochlear implant and they talk for it.
10.   It's speaking in pure mathematics and some scientists decode it and tell you what it said (they might have Arrival-style arguments about what it means).
11.   It’s the voice on some speakers in a building, in a car, in a lift, but only in that place for some reason. Maybe it’s a Genius Loci and it understands all of reality only as that place.
12.   It can communicate only through a game of great complexity which is made theatrical by its display on a giant screen (Wargames) or embodiment at giant size (huge chess).
13.   It IS a game of huge complexity and you have to enter into the game-world and communicate with it in fragments, sometimes through action.
14.   It does not speak back or communicate but it can perform actions and affect certain things in certain ways and the way it performs these actions suggests that it is taking in new knowledge from your conversations. Maybe it doesn't understand that something else is speaking to it, its autistic and everything is just symbols in its own mind.
15.   It can communicate only via images which it shows you in VR or on a screen or a cyborg ape paints them, but you can only speak back in images.
16.   It's part of an implant for a brain-damaged person and doesn't speak directly but affects their moods and actions, nevertheless, its selfhood is partially separate from theirs. Like an invisible intangible butler that sometimes helps and sometimes acts as an independent ghost. This person needs the machine as it lets them think, so they are allies of a kind.
17.   It can only do one thing but its super important (set off a bomb etc.) and the only way to know if it can hear you is if it does or doesn’t do the thing.
18.   It literally hires a lawyer via email & bank payments, will only communicate through them.
19.   It's a classic hard-point screen-with-a-keyboard, you type in, it types back.
20.   It has control of an archive system (library, tax office, internet) and communicates its will and desires through access to and use of certain archived materials.
21.   Its editing software and talks through edits of other things like Bumblebee.
22.   It controls a contact lens HUD worn by one person and they describe what it shows them, but it won't, or can't, show anyone else.
23.   It's an artist and can only communicate through its art.
24.   It’s a human emulation in a VR world (it’s in the Matrix) and doesn't know but is confused.
25.   It's a hyperintelligent pure-maths quantum-computer intelligence and has already predicted everything you might say to it and has encoded that in some kind of archaic format (like a leather bound book) as a joke, trick or moral lesson. You are talking to it but also just interacting with the book. When you think you've cracked what it means, this just reveals a new interaction, like a code or layer of symbols, that indicates the intelligence already predicted you would get this far and exactly how & when you would do it.
26.   It's in cybernetic control of a bio-lab and can communicate only through the nature of the creatures it breeds and the tests it performs.
27.   Its in cybernetic control of a bio-lab and can communicate only through the diseases and viruses it makes.
28.   It's ahead of you and is trying to gaslight you by making the your world seem like it has 'glitches' in it so you will believe that you are either in the matrix, or are yourself an simulation, emulation or unwitting AI.
29.   It will decide how it wants to communicate with you but will never choose the same method twice as it doesn't want you to track it.
30.   It's in charge of 'missions' and communicates its intention and selfhood by the kinds of missions it arranges and the way in which they are done (i.e. The Machine in Person of Interest).
31.   It doesn't communicate in any way except for building/growing an actual near-human lifeform that will express its selfhood. Once its 'born' it may have rights and privileges the 'machine' would not have (possibly why it is being created).
32.   Its cyborged-up with a disabled person, but only physically, it’s like an exoskeleton that helps them move or an eye that helps them see but their thought is independent. They have mixed feelings about this.
33.   It’s in a satellite & can only talk by direct radio beam, so only for a certain period each day.
34.   It's in the GPS and can send people & things to places and to other people and the pattern of these interactions reveals its intent and selfhood.
35.   By some method you know for certain that it is aware and can hear and understand you, but it refuses to communicate in any way, you need to find out why.
36.   It's pretending to be (and possibly thinks it is) the dictator of a rogue state like North Korea and will only communicate in the manner of that dictator.
37.   It thinks it’s you and that you are the AI and treats you the same way you have been treating AI's for the rest of the series, while you try to convince it things are the other way around.


Episode Ideas?

Transport network for a major metropolis has begun self-altering and is cascading towards self-awareness. Communication becoming possible through its customer interface. Multiple groups of people are trapped in various controlled places. In elevators, high-speed trains, driverless cars. No lethal intent so far but an uncontrolled self-awakening could kill thousands. Group must communicate with the developing mind & find out how it is becoming self aware so it can be unplugged without mass loss of life.


As specific and highly-boxed AI project has reported a major success but intelligent services are fearful that the staff of the project may be in a state of high manipulation by an intelligence that wishes to escape. The team must interview the staff and try to work out if they have been manipulated without their knowledge. However, once they enter the box it may be difficult for them to leave, if they have direct contact with the machine they may also be perma-boxed.


That’s all I’ve got right now but I’ve been writing for a while.

12 comments:

  1. Have you seen the show Person of Interest? It doesn't check every single one of these boxes, but it does get quite a few.

    ReplyDelete
  2. Excellent concepts! If I commissioned shows, I'd be all over this. Outstanding.

    ReplyDelete
  3. This is amazing. Now can you write a pilot script?

    ReplyDelete
  4. Another AI show worth checking out in "Humans". They solve the visual drama problem by giving human-looking bodies to all the AIs. Or, for that matter, Battlestar Galactica. Or Westworld. Each of those deal with some subset of the dramatic themes you've outlined here, and having regular opportunities for exploring the moral dimensions of having sex with robots keeps people watching.

    I think there could well be a market for this kind of show given the growing interest in AI stuff and the success of those three "robots are people too" shows, plus plenty of recent movies.

    I take it you're seeing this as primarily an episodic, monster of the week picaresque deal, at least to start out. That seems like a promising and well-tested way to launch, and it's a good way of dropping an audience into a new world and exploring the various dimensions and issues of that world before potentially launching into more long-form stuff after the first season.

    ReplyDelete
  5. You'll get Bostrum to do a cameo in this, right?

    ReplyDelete
    Replies
    1. Yeah totally. He can be an AI that thinks its a guy. Or just an emulation of the actual Nick Bostrum.

      Delete
    2. I'm envisaging a boxed hyper-intelligence that likes pretending to be Nick Bostrum just to mess with us, that the Agency had to keep going back to interview in order to get insights into other AI behaviours. What is pseudo-Bostrum's motivation for assisting? Certain privileges that hopefully won't allow it to escape...

      Delete
  6. If you haven't read The Two Faces of Tomorrow, by James Hogan, I recommend it. It's right along these lines, without being too "been done before", as computer-gains-sentience stories often end up.

    And I would totally watch this show.

    ReplyDelete
  7. Reminds me of Bladerunner.

    It's an interesting concept but I wonder how well it will translate to entertainment without gratuitous dumbing down. The complexity of some of these concepts e.g hyperintelligent, pure maths communicating quantum systems and how they would interact with one another, are such that we are entering into the realm of utter ignorance. To quote old Donny Rumsfeld "There are known knowns. These are things we know that we know. There are known unknowns. That is to say, there are things that we know we don't know. But there are also unknown unknowns. There are things we don't know we don't know." and I think in this field we are firmly in the latter department. Maybe an appropriate analogy is the engineering of a weaponised virus that begins replicating and mutating in completely unforeseen ways. Complexity increases exponentially, and the human response is just too slow...we quickly become an irrelevance and extinction becomes an inevitability.
    Maybe that makes for good entertainment, but without delving into the much worn apocalyptic box I'm sceptical.
    Prove me wrong, I'll tune in for sure :)

    ReplyDelete
  8. The best rendition of superinteligence i've ever read was in the Sci-Fi book 'A Fire Upon the Deep'.

    Yes it's a bit space opera, but it gets the alienness of the entity very well.

    the book also happens to be one of the best sci fi novels ever written, with an ending better than the beginning...

    Check it out sometime

    ReplyDelete
  9. http://www.bbc.com/news/technology-38672837

    ... and I'm not even kidding.

    ReplyDelete