Artificial Morality
blog.lareviewofbooks.org
blog.lareviewofbooks.org
A parallel observation: In the "old days" of the internet, the shining new capability of connecting computers, sharing documents, web pages, software, messages, opinions, and, of course, pornography, the feeling of the day was one of betterment. The network would allow for up-turning out-dated businesses and processes of the past, and we will all be better for it. Information wants to be free!
Yeah, sure.
Information wants to be free? Well, if you consider my click-stream, search history, and probably every keystroke I enter, then yes, that information has certainly been freed. I can even search this treasure trove of information, if I'm just willing to share every piece of personal information about me, from my contact list, to my personal emails, to my phone's location, my calendar and my purchase history.
I no more trust the AI "morality" pundits than I trust the (well meaning, but mis-guided, IMO) visionaries of the internet's early days (and, yes, you can count me among those ranks).
This created and bootstrapped the whole surveillance capitalism and "adtech" industry. Instead of paying for information we have people paying to manipulate us via free media built to be addictive.
I'm sure some of this would have happened had the Internet possessed a payment mechanism and people willing to pay for stuff, but it would not have become the dominant model for everything nor as pervasive.
Note that any billing system is also necessarily a surveillance system - someone has to keep the billing information and know who their customers are. The fight is ongoing as to whether cryptocurrency might change this, but you have to accept that any system which allows very large scale money laundering is not going to be tolerated, which puts an upper limit on how successful that project is going to be.
And a billing system doesn't save you from adverts or selling customer information; US cable TV shows that.
Could they have built an Internet where there was a system to claim refunds, with a "judge" who would decide? Could there have been a rating system ("90 out of 154 visitors said the information on this page was worth the money"), and how could that have been abused?
The internet would be much smaller and much more centralized.
Back in the 90s and early 2000s people really took privacy seriously. The whole PC culture considered spyware anathema. Anyone around back then can remember the outrage when any piece of software was caught sending any private data anywhere without explicit user authorization.
Then the mobile revolution happened, and mobile basically re-educated and re-conditioned the market to accept surveillance capitalism as the de-facto business model. That was then "back-ported" onto the PC and the web.
Why was mobile built this way? Because the web learned that surveillance and ads is the only working business model. Mobile was built as a surveillance/ad platform first and an app platform second because the web showed that this was economically the way forward.
If users were accustomed to paying for things mobile might have been built like a mini-PC for your pocket with a touch UI but carrying forward the old-school PC culture of "I own my stuff" and "don't spy on me."
If I were given a billion bucks to build an iOS/iPhone competitor that respected user privacy and such, the first thing I'd do would be to actually prohibit free apps in the default App Store. All apps must be paid. I would pair this with really advanced privacy and security controls and a good UI for those and a lot of dark patterns applied in reverse to discourage apps from asking for elevated permissions. Push the market back to "you are the customer" from "you are the product."
I see convergent evolution towards ad networks/data brokers, monthly payment walled gardens, and crowdfunding.
It would rather be inevitable. In the news media there were paywall-only news sites in the early days of the Net and this model has almost disappeared (even sites with payed content show ads to their paying members!) as free sites with ads get a lot more views and are more profitable than the ones having a fraction of paying members.
> The practitioners of AI are not up-front about the genuine allure of their enterprise, which is all about the old-school Steve-Jobsian charisma of denting the universe while becoming insanely great.
But that... doesn't seem true? Most modern AI practitioners, especially the prominent ones, profess that their work is going to change the world forever and that's why they want to do it. So I'm not sure what the author is really proposing that AI practitioners should do. Work on less important problems? Play it by ear instead of trying to develop moral principles preemptively?
If he wants us to work for our moral betterment ('Nobody does AI for our moral betterment') he's asking us become like old-fashioned saints or modern-day activists. But they are not problem-solvers per se. They act like they already have the solutions to our problems.
The irony is that, whatever the intentions of its creators, AGI probably will lead to our moral betterment by giving us a clearer understanding of what it means to be human.
Well said. Pessimism is a reasonable reaction to this sweeping line, but I offer a quick summary of what I hope is non-naive optimism:
1) Problem: When we think technology is neutral, we miss how it creates niches outside our intended design and social/political constraints, since the environment of economics thwarts all. We have agency as individuals, yes, but the environment rewards and select for any uses of the technology which yield advantage. “If you incentivize something, it will happen.”
2) Solution?: Nobody fucking knows. But “Game ~B” is what people are calling a nascent research effort to identify what attributes a more cohesive civilization must have to allow better collective “sense-making” and “choice-making” so we may become capable of directing technology instead of leaving it to the emergent dynamics of hapless competition. What's also interesting are the considerations for employing antifragile, transitionary social structures that can thrive in our current entrenched “Game A” dynamics to plant said attributes. Gotta change the underlying rules instead of designing atop those which are broken.
3) Future Ethics: I've also seen some excitement about a so-called non-relativistic ethical framework within Game~B, called The Immanent Metaphysics. It claims to be commensurable with — but not derivable from — science, rigorously formalizing intuitive ethical principles, to at least create a consistent grounding for making effective ethical choices.[1] The hope is to better inform (but not determine) our use of technology after we step into a more coherent phase of civilization. Tech can NOT be put back in the bag, so we must develop our wisdom to carefully wield the godlike powers they give, which we are currently making a mess of.
[1] only a few people currently understand the dense framework, but I'm currently working through it and may be able to offer a summary of what I know so far here. neat stuff
^ This is a link to a relevant article containing an important counterpoint: it argues based on a few assumptions that we should let computers write our moral codes and not the other way around.