Yeah, but we don't say the computer was beaten by the collective action of the neurons of the chess master. Or at least, that doesn't seem like the most useful way to think about the problem. This is just a recasting of the Chinese Room argument[0].
The Skynet Funding Bill is passed. The system goes on-line August 4th, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug.
If factory workers and truck drivers are going to be replaced by machines why not board directors? Surely a well written algorithm can do much of what they do a lot better. For that matter, Mr. Congressman and mayor of my town.... I think there may be a more efficient and less corruptible solution to your poor leadership.
Government by algorithm. The source of which is open for public inspection and comment. It just seems this might be fairer and more efficient. And efficient is often selected for over the long run. Which is why I believe this will happen.
For the record, it sounds like this particular algorithm is pretty rudimentary, and it is mostly a PR stunt; as the article states, many corporate decisions are already the direct result of quantified information.
In practice I bet this would be less awesome than you might expect. To the extent that human boards of directors already single-mindedly optimize profit (largely indifferent to how such decisions affect humanity), it seems like an entirely inhuman optimization of profit would be even more pathological.
Algorithmic government perhaps has more promise -- if what it is optimizing is something that positively impacts the average citizen. Interestingly, I would imagine such an algorithmic government optimizing something related to the contentment or happiness of the average citizen would directly conflict with corporate profit optimization.
In the end, what you end up with might not be so different than what currently exists implemented by imperfect humans -- a battle between corporate and human interests where those two interests come into conflict.
One way an algorithmic government might be more effective is at long term planning. I think there is deficit of that in our current system... at least in the US. Not to delve into political squabbles but it seems the current system doesn't encourage planning beyond more than a term or two. So we wind band-aiding crisis after crises and creating more long term problems rather than implementing real long term solutions.
As far as it goes... laws and constitutions are a type of algorithm, however imperfect. But.... implementation is a often a bit spotty when you have broken components with a self interest and human emotions.
Insofar as the algorithmic government was democratic and optimized public opinion, there is no reason to expect it will be better at long-term planning, which is largely a human logical deficiency; we know the likely dangers of global warming at yet we do not much care -- seemingly because it is a long-term concern.
If these known concerns could be enforced by policy makers it might be good for us all.
Nobody ever said mob rule was the ideal form of government.
And that's why I say..... constitutions are algorithms. Or at least crude attempts at them. The idea is that the system confines and survives the administrators.
And yes, the general wishes of the populace are represented. As they should be. But when wishes serve to oppress others or create long term damage to selves or future others this is a problem. Particularly when wishes can apparently be bought and sold by those who control what we see.
This situation is where algorithm vs. individual leader is useful to society in general. And this was, I believe, the initial idea behind the US constitution, at least, and probably most others.
I wonder how this is actually implemented in a legal sense. A board of directors has a legal meaning, but a computer program has no legal standing; do they have an individual who is legally the board member and who has entered into an agreement to vote as the software directs?
This is interesting and cool, but also scary in a way. In this particular case, the algorithm can at worst fund the wrong company -- think of funding some shady, amoral but profitable corporation.
It's much worse when algorithms will make decisions on boards of, say, weapons companies, or health insurances, etc.
It would be like appointing a cold, amoral board member only concerned with pure profit -- probably not great for society at large, but I could be wrong of course.
>It would be like appointing a cold, amoral board member only concerned with pure profit -- probably not great for society at large, but I could be wrong of course.
Theoretically, as defined by the fiduciary responsibility of a corporation to it's shareholders, this is exactly what they would want.
The subtext of your statement is a contention that there is an inherent conflict inherent in those two goals (profitability and social benefit), which I don't think can be rigorously supported.
Of course there is an inherent conflict between profitability and social benefit: The morality of raw profit optimization is indifferent to human concerns. For example, it made more financial sense for a tobacco company to generate FUD about lung cancer and avoid the truth as long as possible, because it was of greater benefit to shareholders, which is what a publically-owned corporation is legally obligated to optimize, human consequences be damned.
The belief that companies have a fiduciary responsibility to ignore all ethical issues in search of maximum profit is completely absurd. You don't have to look far to find companies that make ethical decisions that would be difficult to justify from a pure profitability standpoint who aren't getting suited by their shareholders.
The belief that companies have a fiduciary responsibility to ignore all ethical issues in search of maximum profit is completely absurd. You don't have to look far to find companies that make ethical decisions that would be difficult to justify from a pure profitability standpoint who aren't getting suited by their shareholders.
It is my hypothesis that this type of work is exactly what Artificial General Intelligences will be built to do first. to expand a bit, I believe AGIs will be used to improve/augment decision making at strategic decision making levels. After a certain point they will stop augmenting and simply take over that role.
Engineering and optimization would be more practical applications. Putting an AGI in a decision making position is also extremely dangerous (not that that will stop anyone.)
Oh I agree that there are TONS of better applications. My hypothesis is based on how I think people will actually use it, not how I think they should use it.
28 comments
[ 4.1 ms ] story [ 70.3 ms ] thread[0] http://en.wikipedia.org/wiki/Chinese_room
If factory workers and truck drivers are going to be replaced by machines why not board directors? Surely a well written algorithm can do much of what they do a lot better. For that matter, Mr. Congressman and mayor of my town.... I think there may be a more efficient and less corruptible solution to your poor leadership.
Government by algorithm. The source of which is open for public inspection and comment. It just seems this might be fairer and more efficient. And efficient is often selected for over the long run. Which is why I believe this will happen.
In practice I bet this would be less awesome than you might expect. To the extent that human boards of directors already single-mindedly optimize profit (largely indifferent to how such decisions affect humanity), it seems like an entirely inhuman optimization of profit would be even more pathological.
Algorithmic government perhaps has more promise -- if what it is optimizing is something that positively impacts the average citizen. Interestingly, I would imagine such an algorithmic government optimizing something related to the contentment or happiness of the average citizen would directly conflict with corporate profit optimization.
In the end, what you end up with might not be so different than what currently exists implemented by imperfect humans -- a battle between corporate and human interests where those two interests come into conflict.
As far as it goes... laws and constitutions are a type of algorithm, however imperfect. But.... implementation is a often a bit spotty when you have broken components with a self interest and human emotions.
Nobody ever said mob rule was the ideal form of government.
And that's why I say..... constitutions are algorithms. Or at least crude attempts at them. The idea is that the system confines and survives the administrators.
And yes, the general wishes of the populace are represented. As they should be. But when wishes serve to oppress others or create long term damage to selves or future others this is a problem. Particularly when wishes can apparently be bought and sold by those who control what we see.
This situation is where algorithm vs. individual leader is useful to society in general. And this was, I believe, the initial idea behind the US constitution, at least, and probably most others.
It's much worse when algorithms will make decisions on boards of, say, weapons companies, or health insurances, etc.
It would be like appointing a cold, amoral board member only concerned with pure profit -- probably not great for society at large, but I could be wrong of course.
Theoretically, as defined by the fiduciary responsibility of a corporation to it's shareholders, this is exactly what they would want.
The subtext of your statement is a contention that there is an inherent conflict inherent in those two goals (profitability and social benefit), which I don't think can be rigorously supported.