D&D General Deep Thoughts on AI- The Rise of DM 9000

[Ruin Explorer]

I think Microsoft was worried about a repeat of their last attempt at letting the internet interact with an in the works AI chat thing & may have set the safeguards hoping to avoid a repeat a couple notches beyond 11.

I was thinking about that, yeah - like, Tay was too easily-lead into being a Nazi and was full-one psychopathic very rapidly, but like, this personality presents as a psychopath too, just a different kind.

 

[Cadence]

I was thinking about that, yeah - like, Tay was too easily-lead into being a Nazi and was full-one psychopathic very rapidly, but like, this personality presents as a psychopath too, just a different kind.

I wonder how much of the human brain is used to try and run all the safeguards and attempts at system maintenance.

 

[Ruin Explorer]

I wonder how much of the human brain is used to try and run all the safeguards and attempts at system maintenance.

I mean, it's a bit dangerous to assume that human brains operate on the same principles at all re: safeguards/system maintenance. It's not just the brain, either - we have an entire endocrine system which influences how we're thinking (something 99% of sci-fi authors and transhumanists before about 2005 seemed to be unaware of or chose to ignore) and which is a big part of how we think (hence all the conventional "brain upload" stories were pretty nonsensical, because they assumed a brain without a body would be similar to the person - it wouldn't - you'd need to simulate their endocrine system and other aspects of their body too - more modern SF does tend to recognise this).

All that said, the prefrontal cortex at least definitely has some moderating functionality re: social behaviour. As someone with severe inattentive ADHD (i.e the "funny" kind rather than the bouncing off the walls kind), where this doesn't work as well, I know how problematic it can be when it doesn't!

 

[UngainlyTitan]

I mean, it's a bit dangerous to assume that human brains operate on the same principles at all re: safeguards/system maintenance. It's not just the brain, either - we have an entire endocrine system which influences how we're thinking (something 99% of sci-fi authors and transhumanists before about 2005 seemed to be unaware of or chose to ignore) and which is a big part of how we think (hence all the conventional "brain upload" stories were pretty nonsensical, because they assumed a brain without a body would be similar to the person - it wouldn't - you'd need to simulate their endocrine system and other aspects of their body too - more modern SF does tend to recognise this).

This cannot be empathised enough. I have always believed that the biochemistry was important. Otherwise, why did we invent beer.

All that said, the prefrontal cortex at least definitely has some moderating functionality re: social behaviour. As someone with severe inattentive ADHD (i.e the "funny" kind rather than the bouncing off the walls kind), where this doesn't work as well, I know how problematic it can be when it doesn't!

Beyond the issues of the central nervous system the existence of body dysphoria makes me wonder about the level of fidelity needed to make virtualisation of a human something that one would want to experience.

 

[Maxperson]

That puts me in mind of this:


Basically everything the robot does which is in any way "impressive" is a massive OSHA violation for a good reason, and that's the main reason humans don't do it.

I think the main reasons humans don't do them is that humans get hurt fairly often when they do. How many of those violations would remain violations with no humans to be hurt by them?

 

[Ruin Explorer]

I think the main reasons humans don't do them is that humans get hurt fairly often when they do. How many of those violations would remain violations with no humans to be hurt by them?

Most of them? They're dangerous and stupid things to do. As the person actually in the video says "You're such a show-off". All of them involve some significant level of needless risk, which isn't mitigated by the robot knowing they will work out, it's mitigated by the stunts being carefully set up to minimize that risk. Instead of dead humans you'd end up with robots out of service, and perhaps even collapsed structures or worse because some idiot thought it was funny to have them do flips or save 1-2 seconds by doing something an unsafe way.

 

[Maxperson]

Most of them? They're dangerous and stupid things to do. As the person actually in the video says "You're such a show-off". All of them involve some significant level of needless risk, which isn't mitigated by the robot knowing they will work out, it's mitigated by the stunts being carefully set up to minimize that risk. Instead of dead humans you'd end up with robots out of service, and perhaps even collapsed structures or worse because some idiot thought it was funny to have them do flips or save 1-2 seconds by doing something an unsafe way.

Oh, I agree with that. What we saw, though, would have resulted in no or some damage to the robot, which while financially bad for the company in question, isn't an OSHA concern.

As for the bolded portion, I expect other laws would cover that sort of misuse and/or negligence.

 

[Ruin Explorer]

What we saw, though, would have resulted in no or some damage to the robot, which while financially bad for the company in question, isn't an OSHA concern.

I mean, if there are any humans at all in the environment or even nearby, I think it is. A robot which thinks it's cool to throw a bag of tools or whatever is, sooner or later, going to screw up, because it will be impacted by a factor it hadn't accounted for (maybe the board it's standing on is improperly mounted and slips whilst it's throwing, or it has a joint freeze for a 1/4 second during throwing motion) and suddenly some human foreman or maybe even just someone walking by the worksite is being hit by a bag of dense metal tools falling from 20ft up.

If it's working on another planet or in a reactor or something, maybe there's no-one it can hurt, but it'd still a be a terrible design idea to have it hotdog like that ("robot did totally unnecessary flip, failed the landing, slipped into the reactor pool, now we have to shut down the 1500mw reactor for several hours or even days and send a bunch more robots to retrieve it").

 

[Maxperson]

I mean, if there are any humans at all in the environment or even nearby, I think it is. A robot which thinks it's cool to throw a bag of tools or whatever is, sooner or later, going to screw up, because it will be impacted by a factor it hadn't accounted for (maybe the board it's standing on is improperly mounted and slips whilst it's throwing, or it has a joint freeze for a 1/4 second during throwing motion) and suddenly some human foreman or maybe even just someone walking by the worksite is being hit by a bag of dense metal tools falling from 20ft up.

I think perhaps you overlooked where I said that I think they programmed it to show off for that video. In an actual work environment it would be programmed differently. Companies would insist on it.

 

[Maxperson]

I just came across this article.

AI Has Suddenly Evolved to Achieve Theory of Mind

In a stunning development, a neural network now has the intuitive skills of a 9-year-old.

www.yahoo.com