5
$\begingroup$

I have a scenario where, thousands of years ago, an alien AI is imprisoned in a human body in place of a traditional AI box to limit its mental capacities (it's being punished and not expected to do anything). It wants to escape so it can reclaim its former intelligence, which is being stored in a massive supercomputer built into the moon, but getting to it is a lot harder than just waiting for information technology to develop and beaming themselves up to it. Accessing its old body requires getting past a series of firewalls, passwords and other security measures they could only have solved with their former intelligence.

So instead, they decide to harness the power of the human mind during the age of transhumanism, instigating a plot to connect billions of human minds into a giant neural network that will have just enough processing power and intelligence to crack the security on their old body.

Would this be a sensible course of action for the alien AI to take if it was both malevolent and impatient, or would have singular computers long since surpassed the processing power of the entire human population by the time this future of transhumanism and neural networks becomes a reality? Basically, would they have any reason to use human brains as a replacement instead of machines besides just being a dick?

$\endgroup$
  • $\begingroup$ a human brain can't take the stress a computer can... unless however you force the brain to ''die'' and stop thinking but then it becomes just a normal computer. $\endgroup$ – user24999 Sep 10 '16 at 21:16
  • $\begingroup$ If you force the brain to ''die'' it would be even more similar to idea from Hyperion Cantos than it is now. $\endgroup$ – Mołot Sep 10 '16 at 21:20
  • $\begingroup$ While not a direct response to the OP's question, the ORION's ARM backstory implies the trillions of baseline human brains are somehow important to the "ecology of thought" of Super Turing AI devices. Perhaps the transcendent AI's use these brains as a sort of substrate for their thinking processes, and weird dreams are a result of some AI using your brain's "spare cycles" for some problem or other. $\endgroup$ – Thucydides Sep 11 '16 at 5:25
5
$\begingroup$

Using humans as processing nodes has pros and cons.

Firstly the brains prefabricated, and just walking around, ripe for exploitation. Individual brains have quite low energy requirements: about 20 Watts, and if you can get all the brains involved in cracking passwords, then they are not going to be foiling your dastardly plans.

However brains are specialised computers, much of their processing power is used to run the body, and engage in various behaviours. People today have a hypercomputer neural network sitting in their head, yet can't multiply 8x4 and not get 36 half the time. If you want to get the full power of the brains you need to get them to stop thinking and start processing. The memory system in the brain is also a pain to use. You can't just malloc.

And although brains only need 20W, they need it in the form of glucose and oxygenated hæmoglobin, and that's very inconvenient. It means you have to keep the body connected and fed.

It's your story, and if you want to keep humans on tap as a source of processing power you can wave your hands and make it work. Perhaps the AI just likes being a dick?

$\endgroup$
4
$\begingroup$

Going in the opposite direction from The Hyperion idea... Non-fiction book: Godel, Escher, Bach: A Single Golden Braid

In this Pulitzer-prize winning book, author noted that it is possible for ants to be non-sentient but for the ant colony as a whole to be sentient. The social interactions of the individuals could give rise to a higher level intelligence. Many pages follow discussing how this works.

Idea played in fiction in The Hitchhiker's Guide To The Galaxy. Earth was a computer so complex that life itself was part of its circuitry.

Combine these two threads and you get the idea that your AI doesn't have to control human minds. He just gets human society running as a computer... then loads his own program into it to run.

Two other fiction books in this domain: The Lady of Mazes by Karl Schroder The Golden Oecumene by John C. Wright

$\endgroup$
3
$\begingroup$

Have you read Hyperion Cantos? If not, make sure you will not, until you finish your book, to be able to honestly tell you didn't in case of plagiarism accusation.

That said, idea is not new. Problem is, human brain, especially adult one, is already programmed and hard-wired to do only specific tasks. It would require a significant breakthrough in neural engineering to allow reprogramming it. it is possible, people who lost parts of their brain responsible for some tasks were able to re-learn them and offload these functions to other brain parts, and some body modders even acquired magnetic sense that, after long enough time, was functioning like another, separate from touch, sense. At least in their opinion. So whilst we know re-wiring brain is possible, we do not know how to force it.

On the other hand, we are at the upper limit of what silicon-based computers can do, and graphene-based will only be faster by some %, probably two-digit %. It will require significant breakthrough in computer engineering to go much faster than, say, two to five times of what we have today. And quantum computers are nowhere near being ready for mass deployment.

For your idea to work, just leave them with graphene CPU that works with the same basic principles our modern ones does, and quantum computers still big, unwieldy and pretty low on q-bits, and allow them to make neural engineering a thing - and it would be believable. Some backstory about neural implants, about studying how this re-learning works, about seeing with a tongue (real thing!), experiments on synthetic neural tissue, and it'll set up your world in a proper state.

TL;DR It all depends in what science your world will have it's breakthrough earlier.

$\endgroup$
  • 1
    $\begingroup$ So basically the human race won't develop hyper-advanced computers that'd do the job better before they invent the kind of technology that'd allow a malevolent entity to brainjack people (it is also worth noting that this AI-disguised-as-human has been guiding technological development on this planet for the past couple hundred years in pursuit of this goal). $\endgroup$ – Z.Schroeder Sep 10 '16 at 21:12
  • $\begingroup$ @Z.Schroeder basically, yes. Just leave them with graphene CPU that works with the same basic principles our modern ones does, and quantum computers still big, unwieldy and pretty low on q-bits, and allow them to make neural engineering a thing - and it would be believable for me. I was studying physics, and I'm a bioengineering & computers enthusiast, if that matters. $\endgroup$ – Mołot Sep 10 '16 at 21:16
3
$\begingroup$

There are several reasons an AI would want to do this instead of using computers. For certain tasks, even supercomputers can't come close to human performance. I believe you should be able to shape your story such that this course of action is not only plausible, but ingenious. See this article for a summary and further references.

Convert humans from potential threats into allies

James K mentioned this. Humans, left to their own free will or coercion by the AI's enemies, could interfere with the plan. Plugged into our AI, they not only stop interfering, they start helping. Vast armies of humans that can accomplish complex tasks in perfect synchrony could be terribly helpful it the AI's enemies decide they need to physically invade to try to stop it.

Machines have been equipped with threat detection

At some point the AI's enemies surely observed the rising availability of computing silicon and decided to plant monitoring software to gather evidence of the AI trying to subvert it for its own ends. Using human brains instead of silicon sidesteps that security measure.

Security on the old body is designed to thwart machines, not human minds

This is the most compelling reason, for me. As pointed out in the article I linked to above, on some tasks, supercomputers still can't perform nearly as well as humans. The AI's enemies designed the security for the old body in a way that would thwart the machines but would not be as effective on a vast hive human mind.

For example, you might think about how some obscure piece of cultural knowledge might be required to get past the security on the old body. Or the task is something like a very complicated, multi dimensional CAPTCHA, difficult for one or even a thousand human minds but feasible to beat with a few orders of magnitude more. Or one of the security measures is monitoring the state of human society (for example, if world_peace == TRUE proceed), something the AI's enemies counted on never being true due to what they know about humans.

Great story idea though. Could be a real thriller.

$\endgroup$
  • $\begingroup$ Using humanity is probably the best option for this AI. Convince them that you are benevolent, that you were trapped by evil aliens, and that said evil aliens will come back and kill all humans unless they can get you to the moon and unlock your potential, which you can use to protect them. A few generations of social engineering and you can basically sit back and let the humans do the work for you. $\endgroup$ – IndigoFenix Sep 12 '16 at 6:53
2
$\begingroup$

In addition to the answers already given, it should be pointed out that human brains are highly developed (through several millions of years of evolution) to specialize in pattern recognition. For the Ancestors, determining if that shape in the tall grass was a rock, hiding deer or a lurking sabre toothed tiger was literally a matter of life or death. Determining the subtle differences between nutritive plants and poisonous ones was also very important, as well as being able to read the contours of the land to find water, shelter or even just finding your way around.

While this isn't going to help an AI very much with password cracking (harnessing millions of PC's and devices in a massive parallel computing cloud network would be far more efficient to engage in mathematical attacks on codes and passwords), being able to harness billions of advanced parallel processors specialized in pattern recognition will provide the AI with a very different way of "thinking" and doing things, which it's silicon competitors might not be able to do easily, if at all.

All is not sunshine in the biological cloud, however. Human and all biological brains work at a speed roughly 1,000,000X slower than electronic devices, this being the ratio between the speed of electrochemical signals passing through neutrons vs electronic signals passing through a computer. A silicon based AI might think geological ages have passed before the wetware AI finishes a thought. There is also the issue of latency between the various processor nodes in this wetware cloud. How is the AI communicating between all these nodes? Having the humans speak to each other or send text messages makes running the "program" that much slower and more difficult.

Finally, the very pattern recognition feature of human brains can often turn into a bug. Conspiracy theories are attempts to tease out meaningful patterns in the random noise generated by the millions of disparate facts and sensory inputs that people have access to. The green alien lizard people are not pawns of the Knights Templar, regardless of what the Illuminati told you. Human brains are also prone to other failure modes, many of which are probably not replicated in any silicon architecture. The AI will have to have some pretty impressive error checking routines to identify and isolate failing nodes due to physical or mental disorders.

So have fun with this idea. The key takeaway is any AI attempting to do this will end up thinking and operating much differently from its silicon counterparts, and the key limitations to overcome is getting all the nodes to work together and the very high latency of the "device" compared to its electronic counterparts.

$\endgroup$
  • $\begingroup$ about conspiracy, I like to point this PBS video, answer part of it Quantum Eraser Answer, funny part of it information is there, but we can't distinguish it from non-information. I find it is funny and symbolic for conspiracy stuff, in form of advocating them kinda. $\endgroup$ – MolbOrg Sep 11 '16 at 13:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.