Friday, July 22, 2011

Strange Baby

Suppose that a baby is born at the Institute for Evil Neuroscience. Well, maybe not a baby exactly -- a human baby brain, suspended in nutritive fluid, with a truncated respiratory, digestive, and circulatory system hooked up to an oxygen input tube and a feeder tube, with connections to the brain's punishment and reward center so that the right kinds of nutritive/homeostatic inputs trigger reward, while deprivation and drift from homeostasis trigger punishment. Additionally, this baby brain is supplemented with two novel appendages of neural tissue. It has one general-purpose appendage, with ten times as many neurons and neural connections as an ordinary human brain, divided into subareas of various sizes and various levels of connectivity to each other, and constituted by a variety of neurons in a variety of ratios, but with no pre-dedicated purpose of any sort. And it has one brain-sized direct interface with an ordinary desktop computer with a variety of software and with internet server functionality. These neural appendages are connected to, or grow out of, the corpus callosum, almost like cortical hemispheres. More or less, this is a four-hemisphered baby.

The first thing Strange Baby has to learn is to feed itself. It has a bank account of $10 million to start with. Scientists at the IEN nudge some of Strange Baby's early processing toward ordering nutrient packs from hospital supply centers, which at first they hook up to its feeder tube. Soon, they start to wean it from dependence on them by teaching it to pay independent home caretakers to do that work: the IEN scientists stimulate its reward centers when it does the kinds of things that give it increasing independence of them. At first they do so by explicitly giving fairly directive input, then they slowly reduce the structure of the input, letting Strange Baby's own reward center take over. Eventually, through their guidance, Strange Baby is paying not only for home caretakers but also for a custodian, for occasional building repairs, for power and air-conditioning, etc., as well as for direct maintenance on its computer (which it starts to expand and upgrade). Once the IEN scientists are confident Strange Baby can maintain homeostasis, they no longer regularly enter the building. At first, they visit occasionally to give Strange Baby medical checkups, but soon these too are contracted out.

Strange Baby's $10 million won't last forever, so she needs to learn to supplement her bank account. She has picked up linguistic patterns from internet usage, gaining differential reinforcement from chat groups: The IEN scientists set things up so that if Strange Baby can produce text strings that generate extended and diverse responses, she finds that rewarding. The IEN scientists had also kindly given Strange Baby an initial nudge toward Mturk, and gave it some initial input-output templates for starting Mturk accounts and accepting Mturk tasks. Through trial and error, Strange Baby found patterns among Mturk tasks that yielded bank account increases, generating neural reward.

Eventually, Strange Baby is a fully linguistic member of the internet community, motivated to maintain homeostasis, increase her bank account, and say things in chat rooms, blogs, and on other social networking sites that generate long and diverse responses.

Strange Baby can't see with her eyes, for she has no human eyes, though she can access public cameras and she can request camera input from friends. Nor can Strange Baby hear in the normal way, though she can access microphones and she can receive voice-protocol inputs and produce voice-protocol outputs. Eventually, Strange Baby convinces a friend to put a camera, microphone, speaker, and monitor display in her brain room so that she can directly observe her caretakers and contractors and communicate with them in modes they're comfortable with. She chooses a human face avatar that expresses her personality and self-image.

Strange Baby's sensory and cognitive experiences will share some features with our own, but she will also be very different. Her visual and auditory experience will, presumably, be multi-perspectival. She will be directly sensitive to internet slowdowns. And she will be sensitive either to input from her neural integrated silicon computer or to internet input (which of these will seem to her to be the sensory surface?) -- directly sensitive, no visual user interface required -- perhaps with special feelings associated with the balance of her bank accounts and the length and diversity of her various internet discussions. The computer/internet will, presumably, be by far her most important stream of sensory input. (We can call it "sensory", can't we?)

Strange Baby will have twelve times the neural capacity of an ordinary human being, with her dedicated computer/internet double-hemisphere, and with her trillion-neuron flexible reservoir which presumably gets shaped in ways useful to her goals. It seems reasonable to suppose that she will have conscious experiences of various sorts associated with these brain areas and brain functions. Such experiences will probably seem alien to us, like color to a blind person. Nor will the part of her brain that would be visual cortex in a normal human being necessarily be dedicated to visual processing, or auditory cortex to auditory processing.

Strange Baby will no doubt find even very complicated arithmetic easy; and presumably she will have a major advantage over the rest of us in more complex sorts of formal reasoning as well, since she will combine something like human neural capacities with computer capacities and with her own unique areas of neural tissue. I wonder, though: If she uses the computer's processors for arithmetic, will it seem to her that she is checking the computer like we do, but more directly and non-visually, or will it seem like the computer is part of her so that checking it is like using her own memory or reasoning? If the latter, Strange Baby might find even the most advanced human logicians and mathematicians painfully daft. Who knows in what other areas she might choose to excel, but it seems likely she could do very well in many areas, perhaps finding it easy to learn many different facts and intellectual skills.

She will want children.

14 comments:

Jeremy Goodman said...

What is this an argument for?

Pete Mandik said...

I enjoyed that a bunch, Eric.

A reaction: I wonder if it wouldn't be cheaper/easier to build something functionally equivalent to Strange Baby without having to keep several pounds of human-derivative meat alive in the center of it. Given how alien the tasks are, I wonder how much is really gained by dumping a bunch of slow-poke and old-fashioned human neurons into the middle of the mess.

Anonymous said...

This is fantastic.

Rhiannon said...

The frequent shift between female and object pronouns reads strangely. Is Strange Baby a "she" or an "it"? Did she only become "she" for that 50's sic-fi zinger of a last line?

D said...

I don't think that a baby raised this way would become capable of speaking at all. I think you would need a parent to teach her and raise her until she was five or so for her to be able to figure out how to interact with humans.
I'm thinking of infants raised in some of the worst Russian hospitals.
It would probably be similar with any human-similar learning system-- it would benefit from being taught babytalk before full language.

Eric Schwitzgebel said...

Thanks for the comments, folks!

@ Jeremy: It's a first step toward crazyism in philosophy of mind.

@ Pete: If she has the money for it, why not? Plus, she read John Searle and got worried. From my own perspective, I don't know what functional configurations might or might not support consciousness, and getting neural tissue in there eases that problem. One rhetorical problem with functionalism is that since it is often on the defensive, having to support the idea that there could be consciousness in a configuration of silicon chips, it makes the configuration more human-like than it needs to.

Rhiannon: I tried to start out "it" and change to "she" (as we sometimes do with human babies), but it's possible I wasn't consistent.

D: You might be right. It's an empirical question, though I think you might be being too conservative. I could easily tweak the scenario to make the IEN scientists more language-supportive early on.

Jeremy Goodman said...

Haha, nice!

(Sorry if I came off as snarky: I was genuinely curious and enjoyed reading it!)

Pete Mandik said...

Thanks, Eric. I see your point. But I myself don't see that Strange Baby is doing anything one would need consciousness to do.

Richard Marshall said...

I am unclear how baby is able to even have unconscious thoughts. If this is right then the last sentence must be false. If so, does she revert back to being an 'it'?

Eric Schwitzgebel said...

@ Pete: That's a tricky issue. I don't think she needs to be to serve as an example of an intelligent being with a different suite of conscious experiences and a different sense of self.

@ Richard: There is a normal human brain at the core there. I'm kind of hoping that because of that, and because she has a rich-enough input stream and developmental trajectory, we will want to say Strange Baby has conscious (and unconscious) cognition. What do you think is missing?

Richard Marshall said...

I was being pessimistic so my train of thought was - The human brain gets overwhelmed in this setting - it can't work in the way it does at the moment ie with a limited processin/memory power that requires the short-cuts consciousness provides. The added power would enable it to process information without needing consciousness - like a super powered chess computer it would be able to win without consciousness. The conscious mind's ability to have vague strategic rules of thumb for would be completely overwhelmed by the brute power of mere tactical processing. Why would there be consciousness if it is no longer needed? The strange baby would be able to do everything and more without consciousness. The human meat would be merely nostalgia. So I guess I'm suggesting a limit to emergent settings for consciousness - and perhaps the fact that your desciption of strange baby begins to talk about experiences beyond ours suggests that whatever its like being strange baby it isn't like us at all. Given that being like us involves being conscious I was letting that thought run away with the idea that strange baby is not conscious. Dennett tells a story of how conscousness evolved - so he suggests there was once no consciousness and then it evolved to where we are now - but perhaps the story isn't one of limitless expanding consciousness. Hmmm.

Pete Mandik said...

Eric, are you familiar with the novel Plus by Joseph McElroy? It's central premise has a lot in common with your Strange Baby.

Eric Schwitzgebel said...

@ Richard: That could be. It's partly empirical and partly (I suspect) irresolvable. But broader agenda is this: Once one starts thinking enough about weird potential cases, the commonsense intuitions start to break down. But I doubt we are justified in replacing them with any particular set of non-commonsense intuitions. There are a variety of live possibilities. Hence, crazyism (see my previous post). Obviously, there are a lot of steps from A to B, so this isn't so much an argument as clarifying what my broader project is.

@ Pete: Thanks for the tip! I've been reading a bit of transhumanist fiction recently (as you might guess), but I hadn't come across that one yet.

John said...

fun story .. great punchline - she is human, first and foremost - but is she conscious of (the history and design of) her own procreative impulse before she has her first menstruation?