Artificial Intelligence

#21
I have been surprised by the rush to implement "self-driving vehicles". I understand the economic drive. But my windows computer still crashes at least once a month. My android phone is more reliable, but not 100%. The computer in my tv system crashes daily. My server farm is better, but I dont know how you do cloud technology in all weather vehicle.

I expect lawsuits in the U.S.A. will eventually provide a counter to those businesses in a rush.
 
#22
But my windows computer still crashes at least once a month.
I've noticed the people most wary of depending on technology for activities where Life hangs in the balance tend to be people like me who are more familiar with what a House of Cards the whole thing actually is.

I'm still amazed that computers even work at all because I know about the thousands of "right-on-the-edge" things flying around inside them every second.
 
#24
My main worry about AI has been the apocalyptic scenario where rather than our expiring accidentally because of AI programming mistakes, we become slaves or extinct because they decide to take over. I found a good article that I think explodes that worry, at http://www.rawstory.com/2016/03/a-n...ts-will-never-have-consciousness-like-humans/ :

Some of today’s top techies and scientists are very publicly expressing their concerns over apocalyptic scenarios that are likely to arise as a result of machines with motives. Among the fearful are intellectual heavyweights like Stephen Hawking, Elon Musk, and Bill Gates, who all believe that advances in the field of machine learning will soon yield self-aware A.I.s that seek to destroy us—or perhaps just dispose of us, much like scum getting obliterated by a windshield wiper. In fact, Dr. Hawking told the BBC, “The development of full artificial intelligence could spell the end of the human race.”

Indeed, there is little doubt that future A.I. will be capable of doing significant damage. For example, it is conceivable that robots could be programmed to function as tremendously dangerous autonomous weapons unlike any seen before. Additionally, it is easy to imagine an unconstrained software application that spreads throughout the Internet, severely mucking up our most efficient and relied upon medium for global exchange.

But these scenarios are categorically different from ones in which machines decide to turn on us, defeat us, make us their slaves, or exterminate us. In this regard, we are unquestionably safe. On a sadder note, we are just as unlikely to someday have robots that decide to befriend us or show us love without being specifically prompted by instructions to do so.

This is because such intentional behavior from an A.I. would undoubtedly require a mind, as intentionality can only arise when something possesses its own beliefs, desires, and motivations. The type of A.I. that includes these features is known amongst the scientific community as “Strong Artificial Intelligence”. Strong A.I., by definition, should possess the full range of human cognitive abilities. This includes self-awareness, sentience, and consciousness, as these are all features of human cognition.

On the other hand, “Weak Artificial Intelligence” refers to non-sentient A.I. The Weak A.I. Hypothesis states that our robots—which run on digital computer programs—can have no conscious states, no mind, no subjective awareness, and no agency. Such A.I. cannot experience the world qualitatively, and although they may exhibit seemingly intelligent behavior, it is forever limited by the lack of a mind.

A failure to recognize the importance of this strong/weak distinction could be contributing to Hawking and Musk’s existential worries, both of whom believe that we are already well on a path toward developing Strong A.I. (a.k.a. Artificial General Intelligence). To them it is not a matter of “if”, but “when”.

But the fact of the matter is that all current A.I. is fundamentally Weak A.I., and this is reflected by today’s computers’ total absence of any intentional behavior whatsoever. Although there are some very complex and relatively convincing robots out there that appear to be alive, upon closer examination they all reveal themselves to be as motiveless as the common pocket calculator.
 
#25
They have a person sitting in the car because they are testing / developing the technology. Eventually they will not have a driver, just passengers and or cargo.
I assume they are racking up the hours until they can go to the government and say: "Our autonomous vehicles have a better safety record than human drivers, it's time to let us operate without a human ready to take control." (A human taking control would make it less safe!). Pick a factor 2x 10x 100x at some point the safety record of autonomous vehicles will be so many times better than humans the pressure will be to restrict human drivers not autonomous vehicles. It is inevitable. Resistance is futile.
 
Last edited:
#26
China's lack of privacy concerns will make it the Leader in the "Race for A.I."

http://www.globaltimes.cn/content/1094094.shtml

At the annual meeting of the Association for the Advancement of AI held in New Orleans last month, China submitted 1,242 papers, 25 percent more than the US.

...The rapid development of AI in China is in part due to the country's huge market, large population and fewer privacy concerns among the public, which means it's easier to collect data.
 
#27
I have been surprised by the rush to implement "self-driving vehicles". I understand the economic drive. But my windows computer still crashes at least once a month. My android phone is more reliable, but not 100%. The computer in my tv system crashes daily. My server farm is better, but I dont know how you do cloud technology in all weather vehicle.

I expect lawsuits in the U.S.A. will eventually provide a counter to those businesses in a rush.
I think "their" goal was to get self-driving vehicles mainstream by 2020, but I agree that they are rushing it and there could be a lot of pushback from accidents... but they'll come out with the statistics and propaganda and say... "yes, occasionally, the AI makes a horrible mistake, but it makes far fewer mistakes than human drivers" and then insurance costs will be significantly cheaper for self-driving vehicles and all the insurance companies will start charging an arm and a leg for anyone that wants to keep driving an old-fashioned vehicle with a steering wheel so people will eventually be forced to accept it via economics. There will be more terrorist attacks where mentally deranged known-wolves mow down crowds of people with their vehicles and eventually anyone who wants to keep a steering wheel will be viewed with suspicion. City people will be early adopters and exceptions will be made for people who live out in the country, but eventually even farmers will be forced into it.

There will be hold outs... backwards-looking gun-toting people like me with quaint ideas about personal freedom and individual responsibility who believe giving up the steering wheel will make them slaves (because the government and corporations will have back doors to all vehicles allowing them to remotely pilot them or shut them down or prohibit them from entering certain spaces or crossing boundaries), but even now all the new vehicles are able to be remotely piloted so it will be increasingly difficult to find a vehicle that meets emissions regulations and that is unable to be remotely hijacked.

Regarding the reliability issues... my assumption is you build reliability with redundancy... multiple CPU's running the same calculations and then voting on a solution.

And I would assume you don't want any safety critical information on the cloud because: what if you're in a bad service area or bad weather, like you said.

Safety will improve with more networking as more vehicles and devices on the road and in people's pockets can rapidly link together forming local ad hoc networks that share critical information. I believe that is how the swarms of drones (like displayed in the 2018 winter Olympic ceremonies) manage to avoid hitting each other. I can imagine future generations of smart phones coming with a special chip that will instantly link with any nearby vehicles communicating information about the pedestrian or bike rider that will help the vehicle predict their path and avoid them. I can imagine "smart intersections" where all vehicles within x-distance of the intersection feed their information to the intersection AI traffic controller which then feeds back instructions to the vehicles that optimizes traffic flow and safety.

I'm wondering if they'll ever get the magnetically levitated ball-wheel technology rolling for the vehicles of the future:
 
#28
It will be really interesting to see how driverless cars work out - particularly in complicated, cluttered streets.

For example, there are many narrow minor roads in Britain that have parked cars on both sides of the road, and drivers find the need to negotiate using hand signals!

Imagine if the recent victim had been a child. When a person sees a child, they can understand a lot about the situation - is the kid under supervision, is he carrying a ball, is he talking to someone else across the street. AI typically gets part of the way and then stalls - think of automatic translation.

David
 
#29
but they'll come out with the statistics and propaganda and say... "yes, occasionally, the AI makes a horrible mistake, but it makes far fewer mistakes than human drivers" and then insurance costs will be significantly cheaper for self-driving vehicles and all the insurance companies will start charging an arm and a leg for anyone that wants to keep driving an old-fashioned vehicle with a steering wheel so people will eventually be forced to accept it via economics.
I am interested in how insurance companies view autonomous vehicles. But wrongful death lawsuits in US can be done by jury and going after companies like UBER have potential for very high settlements.

Having worked with technology for 30 years of my career, I am very skeptical about the safety of computer controlled cars and the reliability of consumer priced electronics.
 
#32
There is now a new Google assistant impersonating a human on the phone. It's scary to me too.

"Google could soon have a feature that lets your phone impersonate people. Called Duplex, it's intended to make people's lives easier by handling standard phone calls that are necessary, but not especially personal.

In examples Google demonstrated on stage during the I/O keynote, Google Assistant called a hair stylist to arrange an appointment and called a restaurant to get information about a reservation, using a voice that sounds a little less robotic than the standard Google Assistant (whether that voice is the user's or a standard Google Duplex voice has not been made clear)."

"(This is) absolutely terrifying. For the sci-fi fans it's terrifying because it's one step closer to the kind of AI we put in robots we enslave and who eventually rebel."
Endeavoring to produce exactly perfect impersonations of human beings is a very bad idea. Even not considering the "strong AI" implications of this development, it's still a very bad idea. Apparently Google is showing little inclination to consider the negative side of their new developments. Full speed ahead. What about combining this technology with the hyper-realistic robots covered in the last post?

"The near future terror of this project has to do with how it could be used to further erode your privacy and security. Google has access to a lot of your information. It knows everything you browse on Chrome, and places you go on Google Maps. If you have an Android device it knows who you call. If you use Gmail it knows how regularly you skip chain emails from your mum. Giving an AI that pretends to be human access to all that information should terrify you."
 
Last edited:
#33
There is now a new Google assistant impersonating a human on the phone. It's scary to me too
I guess arranging an appointment is a relatively simple job - particularly as the person on the other end presumably recognises instantly that she is talking to a robot and simplifies her reply to "ten thirty" or whatever!

I just don't think this technology will be of any real use - roughly like the paper clip that used to jump about in Word for Windows!

Let's change that to an appointment with the doctor:

Robot: My owner has asked me to make an appointment with my doctor between 10 and 12 on Tuesday.

Assistant: I am sorry we have no appointments available then unless it is an emergency!

Robot: OK - I will get back to my owner and contact you again shortly!

Robot: My owner has asked me to make an appointment with my doctor between 10 and 12 on Tuesday because he is suffering from indigestion.

Assistant: I don't think that can be classed as an emergency, I can offer you an appointment late next week.

Robot: OK - I will get back to my owner and contact you again shortly!

Owner: Hello, I am so sorry about that wretched GOOGLE assistant thing - I actually wanted to renew my Omeprazole prescription, but it got confused.

Assistant: That was the 67 th call of that sort this week, we are now activating the latest version of "Apple get lost" software on your line that will screen out calls from GOOGLE Assistant!

David
 
#34
I guess arranging an appointment is a relatively simple job - particularly as the person on the other end presumably recognises instantly that she is talking to a robot and simplifies her reply to "ten thirty" or whatever!

I just don't think this technology will be of any real use - roughly like the paper clip that used to jump about in Word for Windows!

Let's change that to an appointment with the doctor:

Robot: My owner has asked me to make an appointment with my doctor between 10 and 12 on Tuesday.

Assistant: I am sorry we have no appointments available then unless it is an emergency!

Robot: OK - I will get back to my owner and contact you again shortly!

Robot: My owner has asked me to make an appointment with my doctor between 10 and 12 on Tuesday because he is suffering from indigestion.

Assistant: I don't think that can be classed as an emergency, I can offer you an appointment late next week.

Robot: OK - I will get back to my owner and contact you again shortly!

Owner: Hello, I am so sorry about that wretched GOOGLE assistant thing - I actually wanted to renew my Omeprazole prescription, but it got confused.

Assistant: That was the 67 th call of that sort this week, we are now activating the latest version of "Apple get lost" software on your line that will screen out calls from GOOGLE Assistant!

David

That's good. It's an example of why natural language conversation is so hard (or impossible) for current AI - the need for, among many other things, a very wide array of associated information that is inherently available to the human speaker. I think that given the effort apparently put into it, this Google Duplex assistant probably is designed to query the user about at least some details, before it actually makes the call. Probably, even if this enables the Google system to successfully complete the call, it takes the user longer to put the necessary data in than it would have for him to have made the call himself. Of course, the system probably saves the data and attempts to learn from each call. Anyway, it still makes me uneasy.
 
#35
That's good. It's an example of why natural language conversation is so hard (or impossible) for current AI - the need for, among many other things, a very wide array of associated information that is inherently available to the human speaker. I think that given the effort apparently put into it, this Google Duplex assistant probably is designed to query the user about at least some details, before it actually makes the call. Probably, even if this enables the Google system to successfully complete the call, it takes the user longer to put the necessary data in than it would have for him to have made the call himself. Of course, the system probably saves the data and attempts to learn from each call. Anyway, it still makes me uneasy.
Yes - there is real technology behind the speech recognition, but it isn't really AI in the classical sense. Of course, we are not normally conscious of recognising what is said to us - our brains just feed us the words. Actually having the 2-way conversation is the AI task - in away it would look remarkably like some old AI programs from about 30 years ago.

The problem is, querying the user extensively before each call makes the thing useless in practice, and guessing what to carry over from previous calls is probably going to be unreliable. The thing is, AI researchers seem to love their software like a parent might love an ugly child. Who in their right mind would have wanted others to listen to this?

http://digg.com/video/ai-christmas-carol

David
 
#36
It's old news, but for a while the military had a project to develop a tactical robot that could "live off the land" so to speak, by harvesting and consuming "biomass" where it could find it. This robot was providentially given an appropriate acronym, EATR (Energetically Autonomous Tactical Robot). It was supposed to be entirely vegan (wood and other plant matter). Apparently the project was cancelled after the word got out and there was a lot of concern and negative PR that EATR might graduate to carnivory or to at least harvesting the bodies of the dead. After all, what about your vegan or vegetarian who after having 5 drinks then orders a meat pizza? Then, when such "food" is scarce, coudn't EATR graduate to procuring its own bodies? Learning algorithms have a lot of applications.

Not to indulge in alarmist fantasy, but here is yet another potential way AI could become our enemy.
 
#37
It's old news, but for a while the military had a project to develop a tactical robot that could "live off the land" so to speak, by harvesting and consuming "biomass" where it could find it. This robot was providentially given an appropriate acronym, EATR (Energetically Autonomous Tactical Robot). It was supposed to be entirely vegan (wood and other plant matter). Apparently the project was cancelled after the word got out and there was a lot of concern and negative PR that EATR might graduate to carnivory or to at least harvesting the bodies of the dead. After all, what about your vegan or vegetarian who after having 5 drinks then orders a meat pizza? Then, when such "food" is scarce, coudn't EATR graduate to procuring its own bodies? Learning algorithms have a lot of applications.

Not to indulge in alarmist fantasy, but here is yet another potential way AI could become our enemy.
I get the feeling that nowadays a lot of money gets spent on quite absurd projects. Leaving aside the AI aspects, what would the robot do with the food it foraged. I suppose the best I can imagine would be to ferment it for methane gas - but I'll bet even that would be a lot trickier than it sounds. It would probably expire from constipation!

David
 
#38
A Feb. 2018 update on current progress toward practical self-driving cars (Autonomous Vehicle Safety Myths And Facts), from Piekniewski's Blog on the limits of deep learning and where to go next with AI. This expert looks at the facts, not all the hype and wishful thinking.

Will the future bring AV's safer than humans that everyone loves to talk about? Probably yes, but looking at the current data, it will likely take years and require several changes in approach and groundbreaking discoveries (particularly in the space of AI). I think the disengagement line has to fall well below human crash rate, before the safety claim can be seriously made. And when that happens, I will gladly acknowledge it. But not yet.
 
#39
A Feb. 2018 update on current progress toward practical self-driving cars (Autonomous Vehicle Safety Myths And Facts), from Piekniewski's Blog on the limits of deep learning and where to go next with AI. This expert looks at the facts, not all the hype and wishful thinking.
Interesting - but I suspect that plateauing will prove very hard to solve - the better AI that he says is necessary basically means "something that we haven't thought of"!

Furthermore, for automatic cars to be useful, I think they absolutely have to do the entire job. People want them to take kids, things, or incapacitated individuals from door to door in more or less all weathers.

Tesla is rapidly losing money, so the only thing we may hear is that the project was shelved for the foreseeable future for financial reasons!

Nobody mentions the 1980's

That's good. It's an example of why natural language conversation is so hard (or impossible) for current AI - the need for, among many other things, a very wide array of associated information that is inherently available to the human speaker. I think that given the effort apparently put into it, this Google Duplex assistant probably is designed to query the user about at least some details, before it actually makes the call. Probably, even if this enables the Google system to successfully complete the call, it takes the user longer to put the necessary data in than it would have for him to have made the call himself. Of course, the system probably saves the data and attempts to learn from each call. Anyway, it still makes me uneasy.
I think it is useful to imagine a conversational AI system that has at its disposal a huge collection of actual conversations trawled from the internet. Let us assume the conversation is about some general subject, such as art, rather than something of immediate practical concern.

As the human / AI conversation progresses, the AI may be able to match (exactly or approximately) the conversation to one stored in this enormous database - and thus find its next utterance from this conversation. Clearly such a technique might be made to work fairly well with clever pattern matching - yet it would embody no knowledge of the subject matter at all! In other words, to be a worthy AI, we would have to have some idea how it obtains its results.

David
 
#40
I think it is useful to imagine a conversational AI system that has at its disposal a huge collection of actual conversations trawled from the internet. Let us assume the conversation is about some general subject, such as art, rather than something of immediate practical concern.

As the human / AI conversation progresses, the AI may be able to match (exactly or approximately) the conversation to one stored in this enormous database - and thus find its next utterance from this conversation. Clearly such a technique might be made to work fairly well with clever pattern matching - yet it would embody no knowledge of the subject matter at all! In other words, to be a worthy AI, we would have to have some idea how it obtains its results.

David
That's a good variation on Searle's Chinese Room argument against the possibility of AI achieving actual consciousness or mind through some massive and elaborate scheme of data processing.

The Turing test of course isn't really a test for consciousness, so we can't use that. Knowing how the conversational AI got it's (convincing) results certainly wouldn't answer whether the system really had consciousness, it would just confirm that some sort of logical algorithmic data processing mechanism achieved the appearance of it. This would presumably also confirm that the conversational AI had no consciousness, because of the "hard problem" - a mechanism of any possible kind or the functioning of it is of a different, lower order of being than qualia.

It seems to me the only way to truly determine consciousness in such a system would be to find a psychic with powerful telepathic or other powers, who could communicate psychically with the AI. This conjecture is probably even more science fantasy speculation than the expectation of conscious AI.
 
Last edited:
Top