I think it’s way too early to even raise the question of ethics. Let us acheive AGI first, see what form(s) it takes, and then we can at that point raise specific questions about ethics.
Ethical issues are all about how the AGI is used, not the AGI itself. Many seem to miss this fine distinction. For instance, the current “machine learning” tech should not be used in a court of law to determine someone’s guilt or innocence or likelihood to commit a crime, etc. Gradient descent approaches are hardly “magic”, but may seem that way to the uninitiated.
I think a relevant nuance here is that when the AGI itself has agency, then it becomes one of the players in its own use. Going back to the example of a hammer, if the hammer had agency, it could by itself decide to hit someone upside the head.
Some definitions, from Wikipedia, so that we’re all on the same page.
Morality is the differentiation of intentions, decisions and actions between those that are distinguished as proper (right) and those that are improper (wrong).[1] Morality can be a body of standards or principles derived from a code of conduct from a particular philosophy, religion or culture, or it can derive from a standard that a person believes should be universal.[2] Morality may also be specifically synonymous with “goodness” or “rightness”.
Immorality is the active opposition to morality (i.e. opposition to that which is good or right), while amorality is variously defined as an unawareness of, indifference toward, or disbelief in any particular set of moral standards or principles.[5][6][7]
No… laws mean nothing by themselves… First you have to educate and convince (not forcing them by creating a law) people, then it has to become a custom, then it have to become part of the culture and then and only then you make laws…
I agree that definitions are likely the source of confusion. For example, I’ve always defined “ethics” as moral values in actions. Using the example of a schoolyard bully, this is how I have always used these terms:
You see a hefty kid, strong for his age, on the schoolyard. He sees himself as the king of the hill. That perception is a relative value that he holds.
Now you see the kid start strong-arming another kid to steal his lunch money. In this case, he is placing the relative value above his moral values (i.e. his sense of right and wrong). You see the bullying happening, and you immediately recognize that it is wrong. This means you are moral.
You overcome your fear of uncomfortable situations and go tell the teacher, or maybe you step in to defend the other kid at risk of harm to yourself. This means you are ethical (moral values in action).
I feel like I have a wrong definition of ethics, though, because Wikipedia and various posts here and on other sites seem to refer to it in terms of studies and frameworks (versus action). This leads me to having a confusing interpretation of a sentence like “Ethics includes (but is not only) metaethics, normative ethics, and applied ethics”, which I end up interpreting as something nonsensical like “Ethics includes the study of its field, the study of itself, and itself”.
Anyway, I’ll avoid injecting myself into the conversation until I am better aligned on the definitions (I’m sure I am just causing more confusion)
If that were the case, few if any laws would exist ever throughout human history.
They mean nothing only if not enforced. In terms of regulation, they usually are, and it usually has an effect. I’ve already provided multiple personal observations of where it does so. At the moment, you’re simply providing assertions and ignoring my examples of where regulatory frameworks provide positive outcomes.
Food producers are more “afraid” of Groceries-inspectors than the gov-inspector.
Do you think the house builders will build houses to kill ppl ?
The gov is always ineffective by design … there is no LOSS motive.
You can take the libertarian view and wish for things to be a certain way. That’s your prerogative.
But for the sake of having communal conversation with others who don’t share your specific wish for how things should be as we discuss reality, you are NOT allowed to simply dismiss the fact that regulations, however flawed they can be, DO work, or that laws should only come into existence after the entire culture has already internalized them. That is in fact not how it works in reality.
Usually there’s a perceived societal harm, then a reactionary response, with potential for repeal/modification later down the road based on feedback and experience.
I feel you’re conflating government contracting (a la. “Cost plus” contracting) with government regulatory bodies (“Food and Drug Administration”/“National Highway and Traffic Safety Administration”/“Environmental Protection Agency”… and other U.S. regulatory agencies), which have transformed our lives. They don’t have to care about a profit/loss motive, because their value is specifically providing safety and protection. They are not, and should not be considered profit centers, although I could make an argument that their existence has a net positive on the economy and society.
For the topic of “Ethics of AI Research”, while people are spending time debating the ethics, corporations are moving forward in a mostly unregulated manner producing systems that already that produce societal harm. Where the cat’s already out of the bag, and the effects are already seen, we’re at the point where we need to be looking at the “reactionary” mechanism, even if it stifles things for a little bit.
I can give you many counter examples … of bad consequences of laws and gov’s.
Let say there is a law not to release AI algorithm under the penalty of 20y in prison, whats will stop me from releasing it if I really-really wanted to !!
“They mean nothing only if not enforced.”
you prove my point … they have to be “enforced” … the word “force” gave it up
I think we should obviously regulate AI in our own countries, but if some countries will use it for oppression, developing AI is wrong right now. It’ll make my life better regardless, but it’s not worth it. Most of us live pretty comfy lives already (at least compared to some people living under authoritarian governments), and the existing problems won’t be solved by making the economy bigger.
That said, if someone’s going to develop AI anyway, best try to introduce it in a way which will minimize the problems it’ll create. But no one’s going to develop general AI soon, though maybe some form of general intelligence limited to perceptual inferences.
People generally follow the law. Yeah it can have bad consequences (war on drugs messed things up a lot for example) but I doubt regulating AI would do much harm.
What I’m saying is you can pass whatever regulation you want to pass unless you convince ppl not to do bad things all your laws are just piece of paper.
It is as simple as that !
I recall a quote (I think by Steve Deace), “We are not a nation of laws, and we never will be. We are a nation of political will, and we always will be.”
BTW, I should include my definition of politics: Politics is collective ethics. Where ethics asks the question “what should I do”, politics asks the question “what should we do”.
Wow 134 replies … it is time to stop building imaginary ethical castles and go build this freaking damn dangerous gangsta AI … it is not going to build itself
The moral framework which I use is that: AGI are domesticated animals. The underlying assumption is that the AGI will be built by reverse engineering and duplicating an animals brain. All of the ethical and legal implications of dealing with domesticated animals should apply to AGI.
Guidelines and Laws - useful for big organizations like google and universities.
Mostly useless for bad state actors.
The state of the art is bringing powerful AI tools to a desktop near you. Now you have to deal with people that fly drones where they don’t belong and shine green lasers into pilots’ eyes. Either you have to make the laws so draconian that nobody wants to risk the penalty or you have to accept that somebody is going to make “bad” AI programs.
At that point, you have to accept that the restrictions will have some utility, but it will be impossible to prevent the likehood that someone will be making a “Skynet like” evil AI, either by accident or by design.
As I said above, pushing for regulation of those larger organizations listed above may give you a nice warm feeling but the net effect is still mostly Virtue Signaling.
If internet has taught us anything, this is likely to be one of the first DIY AI projects to make the rounds, and in this arena, the purveyors have always pushed the edges of moral and legal issues:
On the plus side, DIY atom bombs are currently out of reach for your average hobbyist.
