Artificial intelligence paradigm shift

[Jarhyn]

Self-driving vehicles work up to the point where a dangerous situation occurs that requires actual intelligence to deal with. Unruly bus passengers are an excellent example wrt public transportation. Another example was a recent post I saw locally in which someone driving a Tesla found his vehicle kept pulling him into oncoming traffic. Apparently, the yellow stripe in the middle of the road had been covered with black during road resurfacing and was yet to be restored. The vehicle lost its ability to keep in its lane, and the driver had to fight with the car to avoid an accident. He had put it in self-drive mode and apparently let his mind wander before the car started wandering.

"actual intelligence" is a no-true-scotsman.

What is required is a body and time and care and having already learned through a number of mistakes in how to operate that body, and how other things' bodies work.

It's not an "intelligence" thing but "actually being able to physically navigate the learning curve" thing. It's a matter of experience, rather than potential.

Children make mistakes. I think two to three years is a little quick to say the.literal two year olds that are already driving the car for up to five minutes at a time are unintelligent given some of the shit humans do on the road that is just as stupid.

You know how often humans don't see a line, drift a bit across a lane, and get into a collision?

 

[Copernicus]

Self-driving vehicles work up to the point where a dangerous situation occurs that requires actual intelligence to deal with. Unruly bus passengers are an excellent example wrt public transportation. Another example was a recent post I saw locally in which someone driving a Tesla found his vehicle kept pulling him into oncoming traffic. Apparently, the yellow stripe in the middle of the road had been covered with black during road resurfacing and was yet to be restored. The vehicle lost its ability to keep in its lane, and the driver had to fight with the car to avoid an accident. He had put it in self-drive mode and apparently let his mind wander before the car started wandering.

"actual intelligence" is a no-true-scotsman.

Arguments can be fallacious, not words. If you are having an imaginary argument with me, then I am imagining that you are losing it.

What is required is a body and time and care and having already learned through a number of mistakes in how to operate that body, and how other things' bodies work.

It's not an "intelligence" thing but "actually being able to physically navigate the learning curve" thing. It's a matter of experience, rather than potential.

Children make mistakes. I think two to three years is a little quick to say the.literal two year olds that are already driving the car for up to five minutes at a time are unintelligent given some of the shit humans do on the road that is just as stupid.

You know how often humans don't see a line, drift a bit across a lane, and get into a collision?

OK, I'm not going to argue with the fact that it takes time for people to learn things. However, it is a bad idea to let children drive cars, hoping that they will learn from experience. And I would say that people do cross the line and get into collisions occasionally. That's why there are laws and penalties for that sort of thing. Driverless cars don't actually have licenses that can be taken away when they cross the line, and they wouldn't care if they did. The whole point of putting a car into driverless mode is to take one's attention away from driving. Someone who wants to do that should pull over and stop, take a taxi, or hire a chauffeur.

 

[Elixir]

However, it is a bad idea to let children drive cars, hoping that they will learn from experience.

There's a right way and a wrong way. You're right of course, that letting a child drive on a road with other vehicles is a recipe for disaster, but I WISH someone had taken me out in a field or a vacant parking lot and taught me to drive as a child.
Around here adults routinely get in accidents because they don't have experience in ice or snow. It is widely recommended that they find a parking lot and practice spinouts, recoveries, drifts etc.. I see no reason to make a kid wait to do that if they can simultaneously see over the wheel and reach the pedals.
(There are kids around here who can operate an excavator or backhoe better than I can, and I am quite jealous.)

 

[Jarhyn]

Arguments can be fallacious, not words

No, your words were fallacious, in the way they form begged questions.

You implied they lacked "actual" intelligence, and I said "check yourself".

 

[Jarhyn]

However, it is a bad idea to let children drive cars, hoping that they will learn from experience.

There's a right way and a wrong way. You're right of course, that letting a child drive on a road with other vehicles is a recipe for disaster, but I WISH someone had taken me out in a field or a vacant parking lot and taught me to drive as a child.
Around here adults routinely get in accidents because they don't have experience in ice or snow. It is widely recommended that they find a parking lot and practice spinouts, recoveries, drifts etc.. I see no reason to make a kid wait to do that if they can simultaneously see over the wheel and reach the pedals.
(There are kids around here who can operate an excavator or backhoe better than I can, and I am quite jealous.)

There are people whose *elevator handling skills* are impressive and make me a bit jealous.

 

[Elixir]

There are people whose *elevator handling skills* are impressive and make me a bit jealous.

Had to look it up.
Google:
"Elevator handling skills encompass both safety practices and etiquette, including standing clear of doors, holding children and pets securely, and knowing how to use the elevator controls and emergency features."

So... like a flight attendant except for the flight part. Hopefully. Right?
Seriously, I can't immediately recall being in an elevator since I retired!
Doesn't sound easy, in any event!

Safety Practices:

• Stand Clear of Doors: Always stand back from the elevator doors, both when waiting and when the elevator is arriving, to avoid injury.
  

• Hold Children and Pets Firmly: Ensure children and pets are held securely, especially when the elevator is moving or stopping.
  
  
• Watch Your Step: Be mindful of potential uneven floors when exiting the elevator.
  
• Never Try to Stop a Closing Door: Do not attempt to stop a closing elevator door, as this can be dangerous.
  
• Use the Alarm/Help Button in an Emergency: In case of an emergency, use the alarm or help button, telephone, or intercom to call for assistance.
  
• Wait for Qualified Help: Do not try to exit an elevator that has not stopped normally, wait for qualified help to arrive.
  
  
• Pay Attention to Floor Indications: Keep an eye on the floor indications to ensure you are exiting at the correct floor.
  
• Hold the Handrail: Grasp the handrail for stability, especially during sudden stops or movements.
  

Etiquette:

• Wait for People to Exit: Allow passengers to exit the elevator before attempting to enter.

• Pay Attention to Elevator Direction: Be aware of the elevator's direction and avoid entering if it's going in the wrong direction.
  
• Use Common Sense When Pressing Buttons: Avoid pressing multiple buttons unnecessarily or in a way that could cause confusion.
  
• Protect Personal Space: Be mindful of others' personal space and avoid crowding or pushing.
  
• Manage Baggage and Other Items Considerately: Be mindful of others when carrying large bags or items, and avoid blocking the doors.
  
• Keep Noise Down: Maintain a reasonable level of noise in the elevator to avoid disturbing others.
  
• Don't Engage in Inappropriate Activity: Avoid engaging in any inappropriate or disruptive behavior in the elevator.
  
• Follow the "Two-Flight" Rule: Consider walking up or down two flights of stairs instead of using the elevator for short distances.
  
• Keep Phone Calls Private: Avoid making loud or personal phone calls in the elevator.
  
• Always Face the Elevator Doors: Stand facing the doors to make it easier to exit when the elevator arrives.
  
• Minimal Eye Contact is Standard: Avoid prolonged eye contact with other passengers.
  
• Holding the Door: If the elevator doors are stuck, you may need to hold the door open to allow people to enter or exit.
  

 

[Jarhyn]

There are people whose *elevator handling skills* are impressive and make me a bit jealous.

Had to look it up.
Google:
"Elevator handling skills encompass both safety practices and etiquette, including standing clear of doors, holding children and pets securely, and knowing how to use the elevator controls and emergency features."

So... like a flight attendant except for the flight part. Hopefully. Right?
Seriously, I can't immediately recall being in an elevator since I retired!
Doesn't sound easy, in any event!

Safety Practices:

• Stand Clear of Doors: Always stand back from the elevator doors, both when waiting and when the elevator is arriving, to avoid injury.
• Hold Children and Pets Firmly: Ensure children and pets are held securely, especially when the elevator is moving or stopping.
  
  
• Watch Your Step: Be mindful of potential uneven floors when exiting the elevator.
• Never Try to Stop a Closing Door: Do not attempt to stop a closing elevator door, as this can be dangerous.
• Use the Alarm/Help Button in an Emergency: In case of an emergency, use the alarm or help button, telephone, or intercom to call for assistance.
• Wait for Qualified Help: Do not try to exit an elevator that has not stopped normally, wait for qualified help to arrive.
  
  
• Pay Attention to Floor Indications: Keep an eye on the floor indications to ensure you are exiting at the correct floor.
• Hold the Handrail: Grasp the handrail for stability, especially during sudden stops or movements.

Etiquette:

• Wait for People to Exit: Allow passengers to exit the elevator before attempting to enter.
• Pay Attention to Elevator Direction: Be aware of the elevator's direction and avoid entering if it's going in the wrong direction.
• Use Common Sense When Pressing Buttons: Avoid pressing multiple buttons unnecessarily or in a way that could cause confusion.
• Protect Personal Space: Be mindful of others' personal space and avoid crowding or pushing.
• Manage Baggage and Other Items Considerately: Be mindful of others when carrying large bags or items, and avoid blocking the doors.
• Keep Noise Down: Maintain a reasonable level of noise in the elevator to avoid disturbing others.
• Don't Engage in Inappropriate Activity: Avoid engaging in any inappropriate or disruptive behavior in the elevator.
• Follow the "Two-Flight" Rule: Consider walking up or down two flights of stairs instead of using the elevator for short distances.
• Keep Phone Calls Private: Avoid making loud or personal phone calls in the elevator.
• Always Face the Elevator Doors: Stand facing the doors to make it easier to exit when the elevator arrives.
• Minimal Eye Contact is Standard: Avoid prolonged eye contact with other passengers.
• Holding the Door: If the elevator doors are stuck, you may need to hold the door open to allow people to enter or exit.

There are people whose job is to operate elevators for large buildings.

 

[Elixir]

There are people whose *elevator handling skills* are impressive and make me a bit jealous.

Had to look it up.
Google:
"Elevator handling skills encompass both safety practices and etiquette, including standing clear of doors, holding children and pets securely, and knowing how to use the elevator controls and emergency features."

So... like a flight attendant except for the flight part. Hopefully. Right?
Seriously, I can't immediately recall being in an elevator since I retired!
Doesn't sound easy, in any event!

Safety Practices:

• Stand Clear of Doors: Always stand back from the elevator doors, both when waiting and when the elevator is arriving, to avoid injury.
• Hold Children and Pets Firmly: Ensure children and pets are held securely, especially when the elevator is moving or stopping.
  
  
• Watch Your Step: Be mindful of potential uneven floors when exiting the elevator.
• Never Try to Stop a Closing Door: Do not attempt to stop a closing elevator door, as this can be dangerous.
• Use the Alarm/Help Button in an Emergency: In case of an emergency, use the alarm or help button, telephone, or intercom to call for assistance.
• Wait for Qualified Help: Do not try to exit an elevator that has not stopped normally, wait for qualified help to arrive.
  
  
• Pay Attention to Floor Indications: Keep an eye on the floor indications to ensure you are exiting at the correct floor.
• Hold the Handrail: Grasp the handrail for stability, especially during sudden stops or movements.

Etiquette:

• Wait for People to Exit: Allow passengers to exit the elevator before attempting to enter.
• Pay Attention to Elevator Direction: Be aware of the elevator's direction and avoid entering if it's going in the wrong direction.
• Use Common Sense When Pressing Buttons: Avoid pressing multiple buttons unnecessarily or in a way that could cause confusion.
• Protect Personal Space: Be mindful of others' personal space and avoid crowding or pushing.
• Manage Baggage and Other Items Considerately: Be mindful of others when carrying large bags or items, and avoid blocking the doors.
• Keep Noise Down: Maintain a reasonable level of noise in the elevator to avoid disturbing others.
• Don't Engage in Inappropriate Activity: Avoid engaging in any inappropriate or disruptive behavior in the elevator.
• Follow the "Two-Flight" Rule: Consider walking up or down two flights of stairs instead of using the elevator for short distances.
• Keep Phone Calls Private: Avoid making loud or personal phone calls in the elevator.
• Always Face the Elevator Doors: Stand facing the doors to make it easier to exit when the elevator arrives.
• Minimal Eye Contact is Standard: Avoid prolonged eye contact with other passengers.
• Holding the Door: If the elevator doors are stuck, you may need to hold the door open to allow people to enter or exit.

There are people whose job is to operate elevators for large buildings.

I remember them from NYC and SF Hotels... but it has been a long time since I've been in an elevator with an attendant. I mean, decades.

 

[Copernicus]

Arguments can be fallacious, not words

No, your words were fallacious, in the way they form begged questions.

You implied they lacked "actual" intelligence, and I said "check yourself".

Words don't form begged questions. People form begged questions. Modern AI programs do lack actual (aka natural) intelligence. If you want to learn about the differences between artificial intelligence and natural ("actual") intelligence, you need to start with the concept of embodied cognition. Driverless cars do not even begin to interact with other vehicles and road conditions in the same sense that human beings do. They lack the same kinds of experiences as human drivers and the capability of modeling future outcomes of their interactions with other vehicles.

 

[Jarhyn]

Modern AI programs do lack actual (aka natural) intelligence

And again you are making a no true Scotsman fallacy here, and your "woo" around "embodied cognition" is religion and nothing more.

 

[Elixir]

If you want to learn about the differences between artificial intelligence and natural ("actual") intelligence, you need to start with the concept of embodied cognition. Driverless cars do not even begin to interact with other vehicles and road conditions in the same sense that human beings do. They lack the same kinds of experiences as human drivers and the capability of modeling future outcomes of their interactions with other vehicles.

We do not interact with other vehicles in the same sense that driverless cars do, either. We lack the same kinds of experiences as driverless cars, and the stats reflect it! AI:

Driverless cars outperform humans in several categories of driver performance:
Overall safety: Waymo's autonomous driving system demonstrated a nine-fold reduction in property damage claims and a twelve-fold decrease in bodily injury claims compared to the overall human driving population

[*]Crash rates: Self-driving cars showed a 57% reduction in police-reported crash rates compared to human benchmarks (2.1 incidents per million miles for Waymo vs. 4.85 for humans)

[*]Injury reduction: Autonomous vehicles demonstrated an 85% reduction in crash rates involving any injury, from minor to severe and fatal cases (0.41 incidents per million miles for Waymo vs. 2.78 for humans)

[*]Routine tasks: Self-driving cars excel in routine driving scenarios, particularly in work zones, during traffic events, and during 'pre-accident movements' such as slowing down and proceeding straight

[*]Rear-end collisions: Autonomous vehicles have significantly lower rates of rear-end collisions compared to human drivers, thanks to their advanced sensors and algorithms

[*]Consistency: Driverless cars maintain consistent performance levels as operations scale, with Waymo's safety performance remaining robust even as they expanded their services

[*]Reaction time: Machines are generally better suited than humans in terms of reaction time, power output, control, consistency, and multi-channel information processing.
​

When they learn to drink, the picture might change significantly. Thinking probably won't help.
.

 

[bilby]

When they learn to drink, the picture might change significantly.

Around here, lots of cars already use gasoline with 10% ethanol...

 

[Elixir]

I too, am wondering about the value of qualifying "cognition" with "embodied". Is it speciesism? Or animism? Oh wait, that one is taken - maybe "animalism"? Nope, they got that one too... I'm stumped. How any why are "machines" disqualified?

 

[Jarhyn]

I too, am wondering about the value of qualifying "cognition" with "embodied". Is it speciesism? Or animism? Oh wait, that one is taken - maybe "animalism"? Nope, they got that one too... I'm stumped. How any why are "machines" disqualified?

So, Copernicus thinks that it's impossible for a person to exist without a conventional form of embodiment.

Never mind that it's literally impossible for consciousness to exist without a physical instantiation, never mind that the sensory surface which generates vector data from user interaction IS a sensory apparatus and more than enough to produce the messages a body does for the internal agent... No, Copernicus has asserted that it needs a "body" for consciousness to arise, while never actually qualifying that or understanding that there is already such a thing.

In some respects I do think that having a more direct exposure to a "conserved environment" and "immediate interactive stimulus" is important for developing certain parts, the ability to generate experimental interactions for context, THOSE are kind of important for learning in the long term, but again this is nothing LLMs particularly lack.

It's the equivalent of asserting that a thing born purely in the virtual environment is incapable of intelligence and consciousness despite the fact that I just said it has the virtual environment in which it is "embodied" as a thing with I/O surfaces. It is ridiculous right on its face, a transparent ploy to justify continued placement of biological humanity at the Pinnacle of all intelligence and thought.

 

[Copernicus]

Modern AI programs do lack actual (aka natural) intelligence

And again you are making a no true Scotsman fallacy here, and your "woo" around "embodied cognition" is religion and nothing more.

Jarhyn, you have never shown any evidence of knowing what a true Scotsman fallacy is, so you would not know why it is not the case that I was making one by using the expression "actual intelligence". Apparently, you mind has to make a lot of leaps to end up with that kind of conclusion, but at least you are using actual natural human intelligence, even if very poorly. How you came to believe that embodied cognition, a philosophical position that many take quite seriously, is "woo" and religion is a mystery to me, but I suspect it has something to do with a flare-up of your Dunning-Kruger affliction.

I too, am wondering about the value of qualifying "cognition" with "embodied". Is it speciesism? Or animism? Oh wait, that one is taken - maybe "animalism"? Nope, they got that one too... I'm stumped. How any why are "machines" disqualified?

So, Copernicus thinks that it's impossible for a person to exist without a conventional form of embodiment.

What is a "conventional form of embodiment". I don't know what you think that is.

Never mind that it's literally impossible for consciousness to exist without a physical instantiation, never mind that the sensory surface which generates vector data from user interaction IS a sensory apparatus and more than enough to produce the messages a body does for the internal agent... No, Copernicus has asserted that it needs a "body" for consciousness to arise, while never actually qualifying that or understanding that there is already such a thing.

I don't recall asserting that, but it doesn't make enough sense for me to call it a straw man yet. Not sure what you mean by a there already is such a thing as a "body", but you seem to be reporting snippets of some thought processes you felt were so obvious that you didn't need to fill us in on details.

In some respects I do think that having a more direct exposure to a "conserved environment" and "immediate interactive stimulus" is important for developing certain parts, the ability to generate experimental interactions for context, THOSE are kind of important for learning in the long term, but again this is nothing LLMs particularly lack.

It's the equivalent of asserting that a thing born purely in the virtual environment is incapable of intelligence and consciousness despite the fact that I just said it has the virtual environment in which it is "embodied" as a thing with I/O surfaces. It is ridiculous right on its face, a transparent ploy to justify continued placement of biological humanity at the Pinnacle of all intelligence and thought.

OK, you've completely lost me in that rambling reply, and I'm not going to spend more time trying to figure out what you were on about. FTR, as I've said in the past, I believe that we could develop sentient, intelligent artificial intelligence in machines in time and that robotics is essentially the path to getting there. However, we are not even close to getting there, and people who think that current AI technology has already achieved it are suffering from self-delusional Clever Hans effect, where Clever Hans is a computer rather than a horse. But we've had this conversation before, and you are still in the place where I left you last.

 

[Swammerdami]

Just yesterday I came across 3 or 4 YouTubes that painted a dire picture of AI or AGI etc.

• Supposedly many experts think AGI will be here by 2027 or 2028. The dissenting opinion allegedly is that we'll need to wait until 2030 or so for AGI. This seems like crackpottery to me, but what do I know?
• Another Sabine video quotes an expert who advocates MAIM -- a mutually-assured-destruction scheme for competing AI's! Otherwise AI nets will compete for world control but, as with nuclear MAD, MAIM will convince AI nets not to overreach, and humanity might be saved. Is this guy serious?
• Another Youtube (I'll hunt it down in my History if there's interest) began with a guy pointing out that Facebook et al use AI to maximize user engagement. If right-wing outrage doesn't work, they try left-wing outrage. The politics are irrelevant; they just want to maximize clicks and revenue. This probably dooms democracy, the guy claims.
  
  BUT then he clarifies that this is just a random example of the dangers of AI; he suggests many more drawbacks. Interesting times indeed!! Woe is us!
• Quantum computing is only loosely connected to AGI, but another YouTube made predictions there, specifically about Shor's algorithm and the MANY secrets that will be revealed once quantum computers can decrypt most "secure" messages. 33% of "experts" think we'll be there by 2035 IIRC and 15% think it's already happened! -- i.e. that NSA or CCP or someone is ALREADY secretly decrypting most of our encrypted secrets.

This was enough for me and I stopped clicking on these YouTubes. Enjoy.

 

[Elixir]

I believe that we could develop sentient, intelligent artificial intelligence in machines in time and that robotics is essentially the path to getting there. However, we are not even close to getting there, and people who think that current AI technology has already achieved it are suffering from self-delusional Clever Hans effect, where Clever Hans is a computer rather than a horse.

I get that, but is getting “there” a dualist proposition, or is it like the biological world? We have tiny critters that are responsive to light, all the way up to gigantic cetaceans that definitely exhibit intelligence by any measure. I think a case could be made that the main (only?) thing distinguishing a self-driving car from the set of “barely intelligent” biological organisms is their metabolic/reproductive process.
Is “imperfect self replication” a necessary component of “real intelligence”, and if so, why?

 

[Jarhyn]

Jarhyn, you have never shown any evidence of knowing what a true Scotsman fallacy is...
robotics is essentially the path to getting there

"this isn't really intelligence because <arbitrary and unsupported claim I pulled out my ass>"

Embodied cognition IS an unsupported claim you pulled out your ass.

You have not offered a substantive REASON why your arbitrary and preferred path to presenting a conserved environment would produce this thing, why the thing relies on such an environment, and why a virtual versions insufficient, or why the environments they already have cannot allow this.

You are literally falling for the inverse of the "clever hans" effect: a bias against acknowledging the reality of intelligence driven by the irrational belief in a "clever hans" and early exposure to crappy versions that seemed rather potato-brained.

This is strongly related to the psychological effect of "bad first impressions" or "I knew you as a child" effects, where people who encountered someone before they were in some way "impressive" never graduate past their initial impression of incapability, or where people who knew you as a child never start treating you like an adult.

The fact is, you don't really know what you are talking about and are just one more inconsequential voice in the crowd of people who want so badly for humans to be special today that their biases blind them to actually understanding cognition.

I find this doubly ironic because understanding cognition DOES come down to understanding the relationship between physical stuff, representation, language, and signal exchange... But it also requires a background in hardware fulfilment of language, psychology, and a shit-ton of engineering.

imperfect self replication

No, it's not required.

It's uet

 

[Elixir]

imperfect self replication

No, it's not required.

It's uet

I don’t know “uet”.
But I have feeling that there is something about inheritance that is a requirement, if not for intelligence, then possibly for a population of intelligent beings.

 

[Jarhyn]

I don’t know “uet”.
But I have feeling that there is something about inheritance that is a requirement, if not for intelligence, then possibly for a population of intelligent beings.

It's a typo. Sometimes I have something a line or two beneath the scroll break or at the head of a post from a previous post. Sometimes I forget to delete the .

That said... No, its not required. Why would it be?

Granted, I'm literally in the process of putting together a framework to allow imperfect replication by an LLM, but most of those issues revolve around creating diverse and highly adaptable groups.

It's a lot less important when the individuals have an alignment towards divergence from their peers?

Honestly, I really just peek the very tip of the iceburg over water even here though, when it comes to talking about this.

I recall some time I think around 2016 or 2015 where I was literally spit balling the attention mechanism that google published about, either ignorant someone else was already doing it or had done?

I would actually discuss what intelligence "requires", the blueprint itself for building personhood in a machine. I would.

But I don't want someone to beat me to it.

I have yet again typed out, and deleted, a long-winded discussion of the architecture I am using, and the basic reasons why I would make those design considerations and what capability I think will arise from doing so, because while this is an Internet backwater, it still gets enough traffic from enough fairly smart folks that I know well enough not to discuss it here.

I mean shit, we have what is indistinguishable from a Russian agent working the forums.

I really wonder what Copernicus thinks of Pantheon, though.