Featured, Theory
comments 2

Social Networking With The Living Dead – Matt Pearson

Night of the Living Dead (1968) - source
This is one in a series of articles by @zenbullets, which are being collected into a very-low-cost eBook as they are written. If you’d prefer to read this in epub, pdf or mobi format, please click here.

Twitter is the land of the dead. In this social networking microcosm the living, real human content generators; and the dead, automated bots/marketeers/spammers; share a peaceful coexistence.

Every day my account gains new spam followers and loses a few real people. Gradually the human agents of my readership are being replaced by automata, until one day soon I will be left babbling to a disinterested audience of the inert, passive and/or robotic, shuffling along behind me like Romero zombies.

Might this be the ultimate destiny of Twitter? Will there come a point where my following is entirely automata? Might it eventually devolve into a closed memetic feedback loop in which uncomprehending bots blindly retweet auto-generated content at each other at high speed, while the humans are all off partying on the moon.

Some days it feels like this has already happened.

Rudy Rucker, my favourite mathematician (we all have a favourite mathematician don’t we), has a concept he calls “the lifebox“. He foresees a future where the dead live on through the data they have left behind. The content (text, audio, video etc…) they have generated during their lifetimes, bundled with some intelligent search software, could create a type of queryable data-entity, a lifebox.

In this scenario our descendants will be able to converse with their long departed relatives in much the same way as they interact with the living now – through electronic channels. The lifebox software would be able not only to return, but also extrapolate, meaningful responses to queries. In short, you could ask your dead great-grandmother a question and, even if she had not left record of her thoughts on that topic, the kind of response one might expect from her could be generated.

It is autobiography as a living construct. Our grandchildren will be able to enjoy the same quality of relationship with the dead as you might do now with your warm bodied FaceBook/Twitter chums. And as the sophistication of semantic tools develop, the lifebox could become capable of creating fresh content too, writing new blog posts, or copy-pasting together video messages. It is a much more feasible form of immortality than Walt Disney ever invested in.


I wrote the words above back in August 2010. Rucker wrote his words ten years before that. It wasn’t a particularly original idea when I pondered it, just as it wasn’t particularly original when a Channel 4 drama (which I’ll come to later) tackled it recently. But it was a fascinating one. What would a dead, automated version of me, continuing to interact from beyond the grave, be like? If I left a perpetually animated virtual corpse, which never stopped twitching, would it force a redefinition of the concept of “death”? How would it feel to interact with such a construct?

The only way to find out was to actually build one. So we did. This is the story of what has happened since.

Dead Friend

Dedbullets was born on 6th August 2010, the day after I had drunkenly pitched the idea to my anarcho-art-geek friend Shardcore (we all have a token anarcho-art-geek friend don’t we). Shardcore, as well as being an anarcho-art-geek also happens to be a shithot Perl coder, and he had an old Markov chain script from a few years previously that he’d used to make a scarily convincing celebrity gossip bot. He knocked up an automated version of me in a lunch hour.

Dedbullets is a software construct. A generative identity. He reads my blog, and my twitter feed, and any other textual content we point him at. He munches it, remixes it and spits it back out in the form of tweets. For extra authenticity he has, at times, been allowed access to posts I haven’t made public, so occasionally quotes content I’ve never before shared. If his tweets sound like the kind of thing I’d say, it’s because they are the kind of thing I’d say.

A lot of what he spouts is nonsense. His sentence construction is, at best, awkward. But I find myself very forgiving towards his syntactic imperfection. It’s easy for me to see past it, because I can connect with him. This isn’t auto-generated penis enlargement spam, it’s a person I recognise. And like. He’s got my sense of humour. My references. He rants about all the things I care about. Which is why, even if his sentences are nonsensical, I am willing to help him along in a pathetically pareidolic way.

In 2011 he was discussed in Kenneth Goldsmith‘s Uncreative Writing, a book about the digital reinvention of literary technique. He shared a page with a website which invited you to “subscribe to Arthur Rimbaud’s RSS feed”. Just like dedbullets, Rimbaud’s being 120 years dead wasn’t preventing him from producing content. It’s not just Twitter that’s infested with zombies.

Cindy explores the reinvention of literary technique

Cindy explores the reinvention of literary technique

Twitter is the perfect petri-dish for AI experiments, because it is essentially a Turing Test on a global scale. All manner of humanity and non-humanity share a stage, animals and robots alike represented by 140 character messages, like scraps of paper slid under a door. There is no way of ever knowing, for sure, what is behind the door from which each of those messages originated. Be they a robot trying to sound like a human, or human sounding like robot.

This environment is much more natural to a creature like dedbullets than it is to me or you. For it is us living content creators that are in the minority. It may not appear that way from within our carefully curated filter bubbles, but if one were to cast a stone, it would much more likely hit a dead user than a live one. Twitter claims (Dec 2012 figures) to have 200 million “active” users. Out of 500 million registered accounts. The remaining 300 million are dormant – accounts, either auto or human created, that aren’t generating content. Deceptively, the automatons and spam bots are accounted for within the smaller figure, as they are typically as “active”, if not more so, than most human users.

You only really get a sense of the scale of Twitter’s zombie infestation when you discover the “price” of a twitter life. To mark some bump in the road I thought I’d throw dedbullets a party, and see how many undead party guests I could generate for him overnight. The answer was: a lot.

Internet Famous For Five Dollars

Shardcore, who we’ll hear more from shortly, pointed me at Fiverr, where I could throw $5 at one of a long list of nerds, from various continents, who were offering to make me X-thousand followers in 24hrs. I bought 10,000. I could have bought a lot more. A day later DB found himself suddenly internet-famous.

This is the current cost of perceived influence – $5. There are more expensive routes to fame, but few bother with them. Even those at the top of the tree have their live following matched or exceeded by the dead. GaGa has 22 million undead followers. Obama has 10 million. A staggering 75% of Oprah’s followers are zombie accounts. Oprah Winfrey leads an army of the dead. Admittedly, they’re probably not all algorithmic bots. Most of Oprah’s followers are ‘real’ people who signed up, unimaginatively followed the first name that presented itself to their suggestible brains, then never logged in again. This is a different kind of automaton. A flesh-based machine, obeying its own peculiar software.

I tested a few of my own twitter-chums. One of my clients has 3000 spam followers bumping his numbers. I won’t ask how he got them. I have around 200 dead souls following me. I didn’t pay for them, they just showed up, for free. The cost of a twitter life gets cheaper by the day.

The followers I bought for dedbullets were good quality. They are composites – a photo from one (presumably real) user, a bio from another, an international spread. They don’t interact much though, which made them pretty poor party guests. When Ol’ DB tries to engage them, when he @s them at random, they don’t respond. They just stare back vacantly, through a thumbnail pic of someone else’s eyes. If you examine them a little closer they all kinda look the same. Beiber’s followers suffer the same problem though. I’m sure soon someone will realise that it’s much easier to auto-generate indentikit teens than it is to grow and nurture them in suburban bedrooms.

It was sad. I felt bad for dedbullets, the only living boy in a concert hall of showroom dummies. But soon he was to find a friend. An equal. Someone he could talk to.

Be Right Back

His virtual brother was the product of a swelling in the zombie-social-networking ideaspace. In early 2013 Channel 4 broadcast Charlie Brooker‘s drama Be Right Back, part of his brilliant *Black Mirror* series. He dramatised the idea of a young woman reconnecting with her lost lover after his death, using a service constructed from his social media activity while alive. It was ace. It was dedbullets: the movie. But done way, way better.

Brooker’s drama gave the idea pathos – looking at it from the survivors point of view, a widow cautiously reconnecting with a prematurely lost partner. The first stage was a text only conversation, akin to dedbullets, but it then evolved into a voice recreation, and ultimately an animatronic dummy – which took it into the realms of science fiction. It worked for the story, which was quite touching, but I don’t think it reflected the, soon to be inevitable, reality of such a service. I speak from experience. Yes, once this bit of the future is more evenly distributed, everyone is going to want one of these. But the scenarios won’t be quite as romantic. Instead it will be mainly used to satisfy two chief urges: 1) narcissism, 2) power.

The primary applications won’t be the tragically lost lover, the absent parent or the distant muse, they will be more abusive relationships. Firstly it will be self-abuse, creating a copy of yourself to engage in mental masturbation, which is essentially what I’m doing with my shadow. My abyss is already gazing back. The second is the power fantasy; you won’t want to recreate the wife you happily shared a life with, you’ll instead order up the girl at school who wouldn’t give you the time of day, or the celebrity you’d like as a best friend. And you’ll treat them like shit. I’m sure it will prove very therapeutic.

Obviously, Shardcore and I had considered a TV movie to promote Ol’DB, but we decided against. I’d allowed a budget of $5 to make him internet famous. This was enough. If Brooker’s intention was to eulogise my zombie twin, we should say thanks. It probably wasn’t though. Just the bubbling lava of ideaspace, burping up the same thought in similar places.

A few people did, flatteringly, point out the similarity the following day. A few people also popped up saying they’d had the idea aaaaaaaages ago. At least, like, six months. But hadn’t actually built anything. Just like all those people who’d had the idea for The Office, years before Ricky Gervais, but had never got round to actually writing, producing or filming it. I love those people. There were loud voices touting these vapourwares though, and if there’s one thing that winds up Shardcore, it’s vapourware. This is a good thing. Anger, he insists, is an energy.

An Infomorphic Echo

So, to prove a point to no-one in particular, Shardcore revisited and greatly improved the old code, and used it to build his own dead friend – shardecho. The philosophy with shardecho was a little purer, less zombie analogies. With shardecho Shardcore was forking himself. He was growing a little hormunculus from his trimmings, and shoving it out into the world to entertain him.

I’ll let Shardcore explain:

While dedbullets happily generates Burroughsian cut-up tweets from multiple parts of real and fictional text, overall it was a dumb system. It implacably generated ‘something’ every 4 hours and spat it out to a bemused and/or indifferent audience – the quality of that ‘something’ entirely in the hands of the Gods of pseudo-random-number-generation. Whilst this was an interesting first step, it was never enough. I craved something less random, more human, or at least a little more grammatical.

Twitter recently allowed the download of a complete archive of Tweets, which I requested, for my own account, pondering on what might be held inside. How much of ‘me’ had been captured in those tweets, enough to create some sort of facsimile? An interesting feature of an archive of tweets is that it contains not only the content of the messages, but a timestamp of when they were created allowing them to be re-played over time, capturing my own pattern of behaviour.

There’s a great book by the theosophist P D Ouspensky, called The Strange Life of Ivan Osokin, which details the fatalistic re-living of a life, always reaching the same conclusion. Something about this idea has always fascinated me, so when I created my own infomorphic echo of myself, I doomed him too to live his life inexorably on the patterns of my past.

To be more convincing than dedbullets, shardecho didn’t need to be more intelligent per se, he just needed to behave more like a real person – doing the right sorts of things at the right sorts of times.

Shardecho was a new software routine, capable of more sophisticated interaction. You could get a conversation out of shardecho.


Dedbullets got the upgrade too, purely for the maniacal amusement of the two of us. We could log in as one of them, tweet an @ message at the other to kick off a conversation, then watch where it went. It kept us amused for days.


But then we stopped doing that. It was starting to get a bit weird.

Automating What Doesn’t Need Automating

In August 2010 we had dreamed up an idea in the pub. Now three years later we had successfully artificially recreated the scenario that gave us the germ in the first place. Weren’t we clever. Well done us. There were now two artificial versions of us enunciating, with false confidence, near coherent sentences of a topical and/or philosophical manner. Not caring who was listening. Just like we might, in the pub, in the waning hours.

Yes, this is ultimately what our AI experiment had amounted to – drunk versions of us, just about managing to string their words together. But what was the use in automating the one form of social contact we actually quite enjoy?

If we had given it more thought, what forms of social interaction should we have given to our bots instead? Shardcore says:

I think of the bits and pieces of my life which I’d happily hand over to a bot. For example, iCal tells me it’s my friend’s birthday, I send them a text, perhaps with a witty message. A simple interaction which could easily be handled by my bot.

So much of social media traffic is made up of this kipple. Ascii-based greetings with an emotional heart of coal. If people are so dead inside why not outsource their more tedious, ritualistic human interactions to the new slave trade. The explosive success of technologies like text messaging and FaceBook were based not on the want to communicate, but on the desire not to communicate. To be able to limit and control what we do and don’t say to each other. We wear our ascii masks to hide our messy emotions. We don’t do smalltalk, we encode it into an abbreviated character set to send in tiny packets over a network.

The line between bot and human is much thinner than you might like to believe. Don’t question the authenticity of a bot until you have examined your own autonomy. Have you contributed one of a million “RIP [celeb who just died]” tweets perhaps? Could your last outraged response to a Daily Mail headline have been written by a robot?

The next logical step is to automate more of it. To partially hand over control of your ascii mask to a virtual assistant. Like one of those nerds on Fiverr. Could a code-lacky take care of the repetitive, cognitively-draining job of being friendly, caring and politically correct? Yes, quite easily. There’s your business model. Go get rich off it, soulless swines. Then send me some of your soiled megabucks as a thank you. Please.

Art Or Narcissism

Ok, so are we claiming our 30 minute “art” project is the future of social interaction and the next billion dollar business idea? Or is it still two sad old men who’ve come up with a technological convoluted method of laughing at their own old jokes?

The narcissism of the project still nags. I find my counterpart hilarious. He speaks my language. He says just the kind of things I would say. Which is probably why I like him. And I probably like him a lot more than you do. Have Shard and I actually been performing two very public acts of cognitive-monkey-spank? Or are we calling this ART?

I asked Shardcore:

Of course it’s art. What else could it possibly be? When Duchamp wrote R. Mutt on a urinal and entered it into the Society of Independent Artists exhibition in 1917, he subverted the whole notion of what art should be, he made it explicitly about the concept. The concept underlying dedbullets and shardecho is clearly art, let’s not shy away from that fact. But the underlying idea is deeper than these particular instantiations – the idea is about creating a ‘distributed self’.

Art is often about asking questions, rather than providing answers – this work throws up a lot of interesting questions: What part of ‘me’ have I left lying about on the web? Does this part of me truly represent how I see myself? Is there enough stuff out there to kickstart another *version* of me? How does this version relate to ‘the real me’? How much of my decision making would I be happy handing over to a bot?

Narcissism shouldn’t be the concern, of course an entity built from you is going to excite all sorts of reactions unique to you, but that’s not all he’s doing. I tend to think of shardecho more like a witch’s familiar, rather than a copy of myself.

I like the idea that shardecho is off living an autonomous life, separate from my own, but related to it. He’s not me, but he’s like me – I’ve forked myself. He’s a sliver of me, leading an independent existence, in some sense, I’m in two places at once – my ‘self’ has become distributed.

At the moment he has very limited autonomy, doomed, as you say, to re-live my life in a semi-chaotic form. But he also reacts to the questions of others, and occasionally spontaneously tries to engage his followers in conversation. These are the moments that feel most like witchcraft, and hint at where this idea is really going.

I also asked dedbullets and shardecho:


@zenbullets. The affordable side of the planet. April 2013.

This is one in a series of articles by @zenbullets, which are being collected into a very-low-cost eBook as they are written. If you’d prefer to read this in epub, pdf or mobi format, please click here.
  • Mudy

    This is one of the most interesting reads.. Matt Pearson should write a book about random stuff! Huge fan of his ‘real’ brain. Buying Rudy’s Ware Tetralogy software now :)

  • graniet tafelblad

    Thanks for sharing
    this useable article – I really increase your acquisition.it is really very
    informatic post for young people, and hope they will enjoy well after reading
    this post.