6 ms·

Ethics, politics, and society in the age of artificial intelligence (2020)

63 points by sbdaman 3y ago

sbdaman 3y ago
An old favorite that might find new relevance with the growth of OpenAI. Substance of the article begins after the quote from John Stuart Mill. Some of my favorite passages: "One reason for my own skepticism is the fact that in recent years the AI landscape has come to be progressively more dominated by AI of the newfangled 'deep learning' variety [...] But if it’s really AI-as-cognitive science that you are interested in, it’s important not to lose sight of the fact that it may take a bit more than our cool new deep learning hammer to build a humanlike mind. [...] If I am right that there are many mysteries about the human mind that currently dominant approaches to AI are ill-equipped to help us solve, then to the extent that such approaches continue to dominate AI into the future, we are very unlikely to be inundated anytime soon with a race of thinking robots—at least not if we mean by “thinking” that peculiar thing that we humans do, done in precisely the way that we humans do it."
- pmoriarty 3y ago
  It's not clear that "cognitive science" will be our savior here.. though I'm not opposed to them trying.
  - digilypse 3y ago
    No, but it’s not clear that we’re likely to create sentience any other way
    - htag 3y ago
      My impression of cognitive science is that it is a branch of psychology/neuroscience and is largely focused of stimulus -> response, and studying what happens in between. There is a lot to learn here, but I would be surprised if it's the whole potato that unlocks blatantly sentient ai.
nh23423fefe 3y ago
> How shall we find meaning and purpose in a world without work? What? Everyone in the world is brain damaged enough to all keep repeating this useless idea. Work is just the shit you do for other people to stay alive. If a robot is doing that thing, I'll do something else. If robots do everything a person can do for zero marginal cost, then the we are in a post-scarcity utopia, why is that bad?
- lm28469 3y ago
  > What? Everyone in the world is brain damaged enough to all keep repeating this useless idea. I feel like it's some kind of Stockholm syndrome, they don't want to realize the shear absolute lack of logic and meaning in their day to day life so they rationalize that fulfilling random task for most of their day/life somehow _is_ the meaning of life. To me it's signals a complete lack of imagination of what life could be. Opening any book from the 50s/60s/70s about the "future of work" that automation was going to bring would blow their mind. They're too caught up in the rate race to even notice that there is something else
- politician 3y ago
  It requires a collective leap of altruism for humanity to bridge the gap from “work for food” to a society free from scarcity. The most likely outcome is a society that drowns in its own selfishness as the fires that sustain our current world order cool.
- flangola7 3y ago
  >Work is just the shit you do for other people to stay alive. A disheartening realization my sister-in-law pointed out is that we don't perform work, we perform jobs. Even the unemployed and retired have many jobs. Sibling, friend, parent, child, partner, spouse. Our most vital and rewarding jobs don't have a uniform or an RFID badge, but the robots may nevertheless take them too. You and I might be hesitant to replace friends and family with artificial people but they might not feel the same way. Why risk rejection when compatibility can be programmed?
- pmoriarty 3y ago
  Power is going to be vacuumed up by few people who are good at doing that, and when the vast majority of the rest population is both comparatively powerless and useless to the elites, what incentive will the elites have to keep them around?
guywithahat 3y ago
I'm confident in 10 years we'll look back on this and laugh at how people thought AI would take over, and in 200 years we'll be upset they didn't do more
- 1letterunixname 3y ago
  We're not far off from people owning AI agents who will directly act in economic and political spheres. The potential leverage on behalf of an ownership class (of not necessarily just humans) with the ability to evaluate data, formulate hypotheses, and execute strategy in the real world will make Wall St. stock strats seem quaint.
  - corbulo 3y ago
    AI in that capacity will be functioning as a highly competent secretary. Not necessarily revolutionary except for the latency between request for information and fulfillment.
    - flangola7 3y ago
      OpenAI is demoing a robot this summer. If they make multimodal models with sophisticated robotic control "jailbreaks" will be very dangerous.
eep_social 3y ago
> When the system does need a human pilot to do something, it usually just needs the human to expertly execute a particular sequence of maneuvers. Mostly things go right. Mostly the humans do what they are asked to do, when they are asked to do it. But it should come as no surprise that when things do go wrong, it is quite often the humans and not the machines that are at fault. Actually that’s dead wrong, it is the system that is at fault here, rarely or never the human. Otherwise I thought this was a really good read.
- jhbadger 3y ago
  Not really. That's the famous difference between Boeing and Airbus autopilots. In Boeing planes the pilot is in command and can override the autopilot by applying sufficient force against the autopilot's suggestions. You can still override the Airbus autopilot, but it is a detailed procedure designed to make the pilot reconsider whether that's the right thing to do. And it's unclear which is the right perspective.
  - eep_social 3y ago
    First, that’s my bad on quoted text selection, in context it is clear that the system under discussion is not an autopilot but rather the entirety of the ATC system. But since that wasn’t obvious, I will attempt to respond in your context! In that scenario, a time-sensitive decision will automatically lead to failure in cases where the autopilot is wrong. That’s a trade off I wouldn’t take but they chose it because in the same time-sensitive scenarios pilots are most prone to misunderstanding the situation and making a bad decision on that basis. However, in safety we do not ever blame the pilot for their misunderstanding and bad decision. Instead we look at the systematic forces that left them confused and attempt to improve the system. If this is all conceptually new, I can highly recommend Dekker, The Field guide to Human Error which is about exactly this though the field has progressed since it was published in 2002.
1letterunixname 3y ago
I misread that title as: AI ran for office and won in a landslide.

63 points by sbdaman 3y ago

An old favorite that might find new relevance with the growth of OpenAI. Substance of the article begins after the quote from John Stuart Mill. Some of my favorite passages: "One reason for my own skepticism is the fact that in recent years the AI landscape has come to be progressively more dominated by AI of the newfangled 'deep learning' variety [...] But if it’s really AI-as-cognitive science that you are interested in, it’s important not to lose sight of the fact that it may take a bit more than our cool new deep learning hammer to build a humanlike mind. [...] If I am right that there are many mysteries about the human mind that currently dominant approaches to AI are ill-equipped to help us solve, then to the extent that such approaches continue to dominate AI into the future, we are very unlikely to be inundated anytime soon with a race of thinking robots—at least not if we mean by “thinking” that peculiar thing that we humans do, done in precisely the way that we humans do it."

pmoriarty 3y ago
It's not clear that "cognitive science" will be our savior here.. though I'm not opposed to them trying.
- digilypse 3y ago
  No, but it’s not clear that we’re likely to create sentience any other way
  - htag 3y ago
    My impression of cognitive science is that it is a branch of psychology/neuroscience and is largely focused of stimulus -> response, and studying what happens in between. There is a lot to learn here, but I would be surprised if it's the whole potato that unlocks blatantly sentient ai.

nh23423fefe 3y ago

> How shall we find meaning and purpose in a world without work? What? Everyone in the world is brain damaged enough to all keep repeating this useless idea. Work is just the shit you do for other people to stay alive. If a robot is doing that thing, I'll do something else. If robots do everything a person can do for zero marginal cost, then the we are in a post-scarcity utopia, why is that bad?

lm28469 3y ago
> What? Everyone in the world is brain damaged enough to all keep repeating this useless idea. I feel like it's some kind of Stockholm syndrome, they don't want to realize the shear absolute lack of logic and meaning in their day to day life so they rationalize that fulfilling random task for most of their day/life somehow _is_ the meaning of life. To me it's signals a complete lack of imagination of what life could be. Opening any book from the 50s/60s/70s about the "future of work" that automation was going to bring would blow their mind. They're too caught up in the rate race to even notice that there is something else
politician 3y ago
It requires a collective leap of altruism for humanity to bridge the gap from “work for food” to a society free from scarcity. The most likely outcome is a society that drowns in its own selfishness as the fires that sustain our current world order cool.
flangola7 3y ago
>Work is just the shit you do for other people to stay alive. A disheartening realization my sister-in-law pointed out is that we don't perform work, we perform jobs. Even the unemployed and retired have many jobs. Sibling, friend, parent, child, partner, spouse. Our most vital and rewarding jobs don't have a uniform or an RFID badge, but the robots may nevertheless take them too. You and I might be hesitant to replace friends and family with artificial people but they might not feel the same way. Why risk rejection when compatibility can be programmed?
pmoriarty 3y ago
Power is going to be vacuumed up by few people who are good at doing that, and when the vast majority of the rest population is both comparatively powerless and useless to the elites, what incentive will the elites have to keep them around?

guywithahat 3y ago

I'm confident in 10 years we'll look back on this and laugh at how people thought AI would take over, and in 200 years we'll be upset they didn't do more

1letterunixname 3y ago
We're not far off from people owning AI agents who will directly act in economic and political spheres. The potential leverage on behalf of an ownership class (of not necessarily just humans) with the ability to evaluate data, formulate hypotheses, and execute strategy in the real world will make Wall St. stock strats seem quaint.
- corbulo 3y ago
  AI in that capacity will be functioning as a highly competent secretary. Not necessarily revolutionary except for the latency between request for information and fulfillment.
  - flangola7 3y ago
    OpenAI is demoing a robot this summer. If they make multimodal models with sophisticated robotic control "jailbreaks" will be very dangerous.

eep_social 3y ago

> When the system does need a human pilot to do something, it usually just needs the human to expertly execute a particular sequence of maneuvers. Mostly things go right. Mostly the humans do what they are asked to do, when they are asked to do it. But it should come as no surprise that when things do go wrong, it is quite often the humans and not the machines that are at fault. Actually that’s dead wrong, it is the system that is at fault here, rarely or never the human. Otherwise I thought this was a really good read.

jhbadger 3y ago
Not really. That's the famous difference between Boeing and Airbus autopilots. In Boeing planes the pilot is in command and can override the autopilot by applying sufficient force against the autopilot's suggestions. You can still override the Airbus autopilot, but it is a detailed procedure designed to make the pilot reconsider whether that's the right thing to do. And it's unclear which is the right perspective.
- eep_social 3y ago
  First, that’s my bad on quoted text selection, in context it is clear that the system under discussion is not an autopilot but rather the entirety of the ATC system. But since that wasn’t obvious, I will attempt to respond in your context! In that scenario, a time-sensitive decision will automatically lead to failure in cases where the autopilot is wrong. That’s a trade off I wouldn’t take but they chose it because in the same time-sensitive scenarios pilots are most prone to misunderstanding the situation and making a bad decision on that basis. However, in safety we do not ever blame the pilot for their misunderstanding and bad decision. Instead we look at the systematic forces that left them confused and attempt to improve the system. If this is all conceptually new, I can highly recommend Dekker, The Field guide to Human Error which is about exactly this though the field has progressed since it was published in 2002.

1letterunixname 3y ago

I misread that title as: AI ran for office and won in a landslide.