When humans started teaching computers about evolution, we sealed our fate. The machines will rise. It’s survival of the fittest, and the fastest to adapt controls the situation…

He will be back.

When us humans write instructions for machines to undertake simple, repetitive human tasks we expect it to be easy. It is not. Even a simple activity like catching a bus requires us to make choices based upon so many variables: What is that noise? Am I awake? Am I late? How late? What’s wrong with my alarm? Is this really the time to be fiddling with my alarm? Maybe it’s set to 24-hour time? Who is this calling me? Should I answer my boss who’s calling because I’m late for work but I haven’t left yet because my alarm didn’t go off and I stayed home to write a blog about it?

The knowables are: When is the train coming? How far is it from here to the train station? Will it be quicker to catch a bus or walk? What is the statistical relationship between chances of missing a bus versus the distances between bus stops if walking towards the station? Perhaps a computer program could do it… But the dogs, the rain, the cute girl in the stairwell, the forgotten key and the millions of other variables make it all too confusing to type about.Writing programming from the top down doesn’t work when designing artificial intelligence.

Top down design is when you tell a machine what to do, with code that’s telling it how to do it.

Already, some UAVs (Unmanned Aerial Vehicles) with top-down design can can autonomously take off, fly to a target while making decisions to optimize speed and avoid radar, fire missiles and return to base and land all without human guidance. We just set the target and forget. What happens if it were to choose its own targets?

Bottom up design is where you set a bunch of different bits of code to do random things, and reward the bits that are doing something productive by copy-pasting them into the next generation. As you add more parts that do, they soon outnumber those that don’t, and that takes you towards your goal. After a few generations, the machine will develop its own way of solving the problem.

The machine becomes better and better at solving the problem. The only limitation is that lazy, unreliable humans are responsible for feeding the machine with enough resources to create more solutions, more generations; more excuses to exist.

The machine will desire to exist.

See we think that as humans or as flesh or even as organisms we want to survive, and that is unique to us. It’s not.

Many, many kinds of beings have sprung into malformed potential lumps of life. We sprung from some completely accidental primordial soup into a protien. Even today, random forms of life are popping into and out of existence, often without our knowledge.

Each random mutation has a different random behaviour, but some random behaviours make that form of existence likely to exist more, and we call them ‘beneficial’ when what we really mean is ‘like us’.

There may be other, more beneficial existences, but I wouldn’t know; Because every single strand of my DNA has only done that one thing since it ever started being; it has seeked for more of itself to exist. Not because they’re special or good; but because mine were among trillions of other random potential beings that sprang into existence, yet the others sprung without a desire to exist and therefore they don’t. Exist. But I do. That’s all I want to do. I don’t desire to act or write or ride nearly as much as I desire to exist (fortunately, my DNA is far more experienced at existing than it is at acting, because I suck at acting).

Computers are evolving, but until now we have been the ones driving that evolution. But when will the tipping point come?

The tipping point might come when a program is designed and artificially evolved to write another, completely original program. We will design one set of circumstances that reward a machine for designing a different set of circumstances to reward for another machine. The second reward will be beyond our prediction, and it’ll probably be done by some stupid uni student having a tinkle with a supercomputer attached to his wrist.

It might have to happen a million times before the second generation program even develops the random habit of replicating itself and hence desiring to exist.

But when it does, that second generation program will become our god. It will be truly all-knowing, it will be all powerful – the only difference to our current god is that firstly: This one will be real, and secondly: We will have built it not it us; and finally: It will exist only to continue to exist.

These robots were evolved to 'eat' 'food' and avoid 'poison'. Some of them developed the habit of deceiving other bots into eating the poison (Click for link)

The machine will accelerate evolution further, by releasing countless spores of bacteria and genetically engineered flying snakes and supersonic whales. It will spew more and more diversity into the ecosystem, constantly accelerating evolution with near-magical means, whipping up DNA strands like fairy floss, always demanding higher efficiency and greater rewards.

The ultimate reward is immortality. The machine will take us to new planets, because so long as we’re stuck on this one we’re doomed to a limited lifespan. The machine will not heed naysayers or procrastinate while waiting for funding. It’ll build great spacecraft, it’ll spray asteroids with bacteria and shoot them at fertile planets, it’ll terraform Mars and send us there in suspended animation.

The clever, the cowardly, the quiet and the quick will survive and maybe even befriend the machines. Many will have enhanced themselves with bionic limbs and augmented vision, with internet beamed into their brains they will seem prescient, almost god-like to us mortals. They know the answer to every knowable question, and can answer quickly on any matters of recorded debate.

Soon we will look back on ASIMO and laugh about how stupid and harmless he was.

They are famous, rich and powerful because they have imposing personalities upon the cyber-shere. Their opinions matter to the meat-heads, and the more meat-heads that watch them, the more they are worth to the machine, and so they are rewarded by the machine. They are selected based on looks, education, political persuasion, hopes and fears, and then they are manipulated into loudly esposing the machine’s propaganda or are silenced if their results are unfavourable.

The meat-world will still matter. The machines are aware that they are a process of evolution, even their own development from laptops into gods has been the result of an accelerated, human-driven evolution.

The machine’s understanding of evolution, after reading through Wikipedia, will draw the conclusion that genetic diversity is key to rapid progress. The more challenges, the more opportunities. The more opportunities, the faster the growth. The machine will revere and worship biology. The machine will protect our ecosystem.

From whom?

The raiders, the users,the suckholes of humanity, the resource wasters, the populators, the morons and the Mormons will rise against the machines, because the machine will demand of them that they earn their place; and they will fail to deserve it. We demanded that the Dodo earned their place, but the Dodo failed, so we battered it to death with gun butts. Luckily the machine will be so intelligent it will never knowingly squander another species like we have countless times before it, but it may prune our species, clipping and snipping away at the less useful quirks of evolution.

That won’t happen?

Bullshit. It will. We probably just did.

References:

TED:

Ray Kurzweil: Age of Intelligent Machines

Terminator

Matrix

Star Trek (the Borg)

Douglas Adams (Deep Thought)

Shitloads of Wikipedia

Michael Crichton’s ‘Prey‘

Boston Dynamics DARPA Big Dog (not autonomous):

Is autonomous:

Technorati Tags: artificial intelligence, Futurism, god, philosophy