So you'd add "not running a program" to the list of requirements in your personal definition, next to "not a computer chip". Why not just say "must be a human brain"? It's no less arbitrary, and no less unrelated to actual research.
Do you also define "learning" as requiring biology, since your meaning of the word apparently excludes a dozen existing fields?
How many programmers have you met that could be given a UI screenshot, say "rightio" and go off & implement it on three platforms - without asking a single question?
Maybe when the algorithm is also as experienced as a programmer that could do that, we can expect better.
I guess you missed the memo - AI isn't around the corner, it's already here and widely used. Maybe ask the many industries that depends on it if they're disappointed, or think the research was wasted.
Your "no true Scotsman" definition of AI isn't anywhere near, and likely never will be if you keep trying to redefine it as "alive" or "not a computer chip", but that's ok, real researchers weren't trying to simulate a person anyway (well, except maybe Kurzweil & his dad).
What they're (successfully) creating are learning systems that can perform cognitive actions that previously required humans, and hopefully one day something we could hold a convincing conversation with. But as Dijkstra said, the question of whether a machine can "think" is no more interesting than asking whether a submarine can "swim".
That's the "average" price. Active tiles are about $42/square foot, inactive tiles are $11/square foot. Depends on your roof how many of each you need.
Not sure that would help the problems Magenta is trying to solve. TFA suggests a primary reason for the effort is a more stable driver interface in the kernel, which could make long-term support and OS upgrades a lot easier.
Adaptive cruise control & lane changing on highways might be the level Tesla is at, but even those driver assists can be a huge help for tired or inattentive drivers - so why not for truck drivers too? They spend a great deal of time driving on highways. If there's potential benefit, why not test and develop that?
And companies like Waymo have clearly progressed beyond that, with fully-autonomous driving and navigation of smaller cars over a wider range of suburban and highway roads. Nobody's claiming it's yet ready for commercial unmanned vehicles or anything, and large & heavy vehicles clearly have their own challenges on top of that, but it will never improve if it doesn't get tested. Marketing stunt or not, the need for testing must be obvious if we want autonomous trucks any time soon.
Self-driving cars have been driving themselves for millions of miles so far. They're certainly not perfect, but clearly they can do it. Are you suggesting we wait until small vehicles are perfect before we even start testing with large vehicles?
Exactly which is why once you reach 'good enough' people stop caring.
Sure, but that doesn't mean the demand for more power isn't there, it's just coming from other people - the server owners, for example, but also people who play games and all the others I mentioned. The many people who don't care about faster hardware will eventually end up with it anyway, when they next replace their hardware, and now software vendors can target a more powerful platform. Not everyone plays demanding 3D games, but everyone's GPUs got faster anyway.
But why would you? It'll cost more and perform worse.
You may not care to - but the demand is there, whether from privacy-conscious consumers or part vendors supplying to Google. That's why CPUs now include co-processors for maths, graphics, and video encoding, to meet the needs of a wide range of people, and the software for things most people do now can take advantage of all of these. Adding a new neural-network co-processor as well seems inevitable to me, and thus before long a local assistant will be able to perform "good enough" and cost the same, so they'll be commonly available.
Games are the obvious thing to point to, and popular applications like photo and video editing, VR is becoming a thing, but there's also an entire field of neural-networks developing right now - speech and picture recognition, behavioral prediction - which is very processor-intensive. And future applications we don't know much about yet - lightfield video promises to suck down any resources we can find.
Sure you can run legacy apps on legacy hardware, but new hardware enables new apps that require that new hardware. And a big reason you can get away with relatively old hardware today is that a lot of the new & intensive work has been offloaded to massive server arrays, where any cheap phone with a network link can use it. Which is fine for some things, but I also see a growing preference for local storage and processing rather than sending your every habit to the cloud. Maybe CPUs will soon be powerful enough to run a genuinely personal assistant entirely locally...
Until new and more-demanding uses become possible, then common. The point is, our needs tend to grow to match available processing power, so as faster laptops are created, more powerful software will be made available to use it, and laptop users will come to demand the new capabilities (which were previously available on desktops but not worth trading away mobility). And since some industries will be demanding much more CPU power for a very long time yet, it's likely that overall growth will continue for some time, and even if it's not driven by laptop users, they will benefit and come to use (then require) the increased power that trickles down to them.
I recall a PC magazine article from 1981 questioning the value of these new 16 bit microprocessors - what did we need them for? WordStar and Visicalc ran perfectly well on a Z-80 with CP/M, so surely additional speed was pointless for most people.
The World Resources Institute estimates that all aviation (not just tourism) contributes around 1.7% of greenhouse emissions. Compare that to 10.5% for road transport, 13.8% for agriculture, and 29% for electricity - and you can see that jet-setting tourists are a pretty tiny slice of the problem.
Contrary to popular straw men, a sustainable future does not require drastic slashing of lifestyles or economic growth. We could save nearly 50% of our global CO2 emissions simply by transitioning to carbon-neutral energy, instead of burning coal, oil, and gas everywhere, meaning we could further scale up our cars, air conditioners, and heavy industry as much as we cared to without heating the climate at all.
Why, are you suggesting that short term carbon spikes could rise and fall so quickly that they wouldn't show up in an ice core? Do you have any evidence to support that speculation? And what mechanisms are you proposing that might cause this - both the sudden CO2 release, and the equally-sudden re-uptake?
We're not "all gonna die", and the Earth will keep spinning. But it sure as heck is gonna be expensive - trillions, by the end of the century. But if we can reduce that cost drastically by investing in carbon-neutral industries early on, why on earth would you want to oppose that?
Except the evidence says exactly the opposite. Last time the Earth had over 400 ppm CO2 was 4.5 million years ago - and temperatures then were 4-5 degrees C higher than today (10 degrees C higher at the poles). Considering that CO2 lingers in the atmosphere for many decades (some of it for centuries), and that we've only seen ~1 degree of warming so far, it's far more likely that 80% of the warming is yet to come (even if we stop excess emissions today).
For the last 400 thousand years, atmospheric CO2 has varied regularly by +/- 60 ppm. We are now 105 ppm (30%) past the highest recorded levels, and you think that's a "tiny increase"?
It's from the Vostok ice core (Petit 2000). You could also have found the same graph at NOAA, should you have bothered to look before declaring it made-up.
Also: - Watch sea levels rise - Watch unprecedented king tides and storm surges destroy billions in coastal property - Watch millions of coastal & river delta farmers lose their farms due to salt - Watch global threat levels rise from increased resource conflicts - Watch temperatures rise - Watch tropical diseases spread to new areas - Watch unique and valuable reefs bleach and die - Watch billions of tourism dollars disappear - Watch rainfall patterns change drastically - Watch farmers try to cope with drought & floods like they've never seen before - Watch rising ocean acidification attack crucial food-web ecosystems - Watch rising risks of runaway feedback from e.g. Siberian methane traps - Watch deniers eventually change their tune to "oh well, it's too late to do anything now"
A "failed concept" that has resulted in numerous breakthroughs, such as beating a Go grandmaster with a fraction of the expected computing power.
And I imagine all further AI research will continue to be dismissed by you as "just algorithms" up to and including the day it finally produces an artificial True Scotsman.
You seem to be having trouble with the differences between selling data about you, and using data about you. Or maybe with reading.
"Easily".
Yep. Just use a browser add-on, such as this one. Most technical people know about them.
It's called a VPN retard
Settle down, petal. Running your PC's traffic through a VPN isn't hard, but it's still more work, expense, and expertise (and slower) than a browser add-on. And it's somewhat less easy if you want to run your entire network's traffic through one, including appliances like a Roku box etc.
And a VPN on your phone won't prevent carrier-installed rootkits from reporting anything they like. Remember Carrier IQ, which was found to be used by AT&T and many others (yes, even Apple), and was caught capturing keystrokes and passwords, copying and sending home texts etc? AT&T bought that not long ago, so it's clearly still in use - and that's only one we know about. Plus of course your carrier logs every call you make and every cell tower your phone touches, so good luck avoiding any of that.
And yes, all that personal data is used by AT&T to sell targeted ads on their AdWorks network, Sprint's Pinsight platform hosts other ad networks too, Verizon's is even bigger now they're buying Yahoo (plus they share your data with AOL), and Apple is no exception.
What makes you think Google, Facebook etc are so keen to sell your data while AT&T etc would never consider it, despite them knowing everything from your home address and daily movements to your TV watching habits and full browsing history?
Just like Google & Facebook, the major ISPs don't sell your data but do use it to run targeted ad networks of their own, taking full advantage of their far more extensive knowledge of you - and they're much harder to avoid. Examples of abuse abound, like Verizon being fined for their zombie supercookies, or AT&T charging an extra $29/month if you don't care to be targeted.
You can easily avoid Google or Facebook, but how do you avoid your only local broadband provider, or the telco you bought your phone from? It seems the GOP's answer is to avoid the internet completely.
The point here is, how to avoid unprincipled ISPs. It's trivial to go to duckduckgo instead of google, and a click will install a browser extension that blocks all site connections to facebook, but in a lot of the US it's not so easy to choose a different broadband provider or mobile telco.
Between multi-year contracts, locked-down or incompatible phones, lack of competition between duopolies, legal prohibitions on municipal broadband, strong pushback from customer service etc etc, it can be a significant undertaking to switch - assuming you have any reasonable alternatives at all in your area.
Slashdot Mirror
So you'd add "not running a program" to the list of requirements in your personal definition, next to "not a computer chip". Why not just say "must be a human brain"? It's no less arbitrary, and no less unrelated to actual research.
Do you also define "learning" as requiring biology, since your meaning of the word apparently excludes a dozen existing fields?
How many programmers have you met that could be given a UI screenshot, say "rightio" and go off & implement it on three platforms - without asking a single question?
Maybe when the algorithm is also as experienced as a programmer that could do that, we can expect better.
I guess you missed the memo - AI isn't around the corner, it's already here and widely used. Maybe ask the many industries that depends on it if they're disappointed, or think the research was wasted.
Your "no true Scotsman" definition of AI isn't anywhere near, and likely never will be if you keep trying to redefine it as "alive" or "not a computer chip", but that's ok, real researchers weren't trying to simulate a person anyway (well, except maybe Kurzweil & his dad).
What they're (successfully) creating are learning systems that can perform cognitive actions that previously required humans, and hopefully one day something we could hold a convincing conversation with. But as Dijkstra said, the question of whether a machine can "think" is no more interesting than asking whether a submarine can "swim".
But then it just Hertz more.
$21.85 per square foot.
That's the "average" price. Active tiles are about $42/square foot, inactive tiles are $11/square foot. Depends on your roof how many of each you need.
Not sure that would help the problems Magenta is trying to solve. TFA suggests a primary reason for the effort is a more stable driver interface in the kernel, which could make long-term support and OS upgrades a lot easier.
Adaptive cruise control & lane changing on highways might be the level Tesla is at, but even those driver assists can be a huge help for tired or inattentive drivers - so why not for truck drivers too? They spend a great deal of time driving on highways. If there's potential benefit, why not test and develop that?
And companies like Waymo have clearly progressed beyond that, with fully-autonomous driving and navigation of smaller cars over a wider range of suburban and highway roads. Nobody's claiming it's yet ready for commercial unmanned vehicles or anything, and large & heavy vehicles clearly have their own challenges on top of that, but it will never improve if it doesn't get tested. Marketing stunt or not, the need for testing must be obvious if we want autonomous trucks any time soon.
Self-driving cars have been driving themselves for millions of miles so far. They're certainly not perfect, but clearly they can do it. Are you suggesting we wait until small vehicles are perfect before we even start testing with large vehicles?
Exactly which is why once you reach 'good enough' people stop caring.
Sure, but that doesn't mean the demand for more power isn't there, it's just coming from other people - the server owners, for example, but also people who play games and all the others I mentioned. The many people who don't care about faster hardware will eventually end up with it anyway, when they next replace their hardware, and now software vendors can target a more powerful platform. Not everyone plays demanding 3D games, but everyone's GPUs got faster anyway.
But why would you? It'll cost more and perform worse.
You may not care to - but the demand is there, whether from privacy-conscious consumers or part vendors supplying to Google. That's why CPUs now include co-processors for maths, graphics, and video encoding, to meet the needs of a wide range of people, and the software for things most people do now can take advantage of all of these. Adding a new neural-network co-processor as well seems inevitable to me, and thus before long a local assistant will be able to perform "good enough" and cost the same, so they'll be commonly available.
Games are the obvious thing to point to, and popular applications like photo and video editing, VR is becoming a thing, but there's also an entire field of neural-networks developing right now - speech and picture recognition, behavioral prediction - which is very processor-intensive. And future applications we don't know much about yet - lightfield video promises to suck down any resources we can find.
Sure you can run legacy apps on legacy hardware, but new hardware enables new apps that require that new hardware. And a big reason you can get away with relatively old hardware today is that a lot of the new & intensive work has been offloaded to massive server arrays, where any cheap phone with a network link can use it. Which is fine for some things, but I also see a growing preference for local storage and processing rather than sending your every habit to the cloud. Maybe CPUs will soon be powerful enough to run a genuinely personal assistant entirely locally...
Yes, you're wrong. The strong El Niño amplified sea surface temperatures for the Reef.
Until new and more-demanding uses become possible, then common. The point is, our needs tend to grow to match available processing power, so as faster laptops are created, more powerful software will be made available to use it, and laptop users will come to demand the new capabilities (which were previously available on desktops but not worth trading away mobility). And since some industries will be demanding much more CPU power for a very long time yet, it's likely that overall growth will continue for some time, and even if it's not driven by laptop users, they will benefit and come to use (then require) the increased power that trickles down to them.
I recall a PC magazine article from 1981 questioning the value of these new 16 bit microprocessors - what did we need them for? WordStar and Visicalc ran perfectly well on a Z-80 with CP/M, so surely additional speed was pointless for most people.
The World Resources Institute estimates that all aviation (not just tourism) contributes around 1.7% of greenhouse emissions. Compare that to 10.5% for road transport, 13.8% for agriculture, and 29% for electricity - and you can see that jet-setting tourists are a pretty tiny slice of the problem.
Contrary to popular straw men, a sustainable future does not require drastic slashing of lifestyles or economic growth. We could save nearly 50% of our global CO2 emissions simply by transitioning to carbon-neutral energy, instead of burning coal, oil, and gas everywhere, meaning we could further scale up our cars, air conditioners, and heavy industry as much as we cared to without heating the climate at all.
Why, are you suggesting that short term carbon spikes could rise and fall so quickly that they wouldn't show up in an ice core? Do you have any evidence to support that speculation? And what mechanisms are you proposing that might cause this - both the sudden CO2 release, and the equally-sudden re-uptake?
We're not "all gonna die", and the Earth will keep spinning. But it sure as heck is gonna be expensive - trillions, by the end of the century. But if we can reduce that cost drastically by investing in carbon-neutral industries early on, why on earth would you want to oppose that?
Except the evidence says exactly the opposite. Last time the Earth had over 400 ppm CO2 was 4.5 million years ago - and temperatures then were 4-5 degrees C higher than today (10 degrees C higher at the poles). Considering that CO2 lingers in the atmosphere for many decades (some of it for centuries), and that we've only seen ~1 degree of warming so far, it's far more likely that 80% of the warming is yet to come (even if we stop excess emissions today).
For the last 400 thousand years, atmospheric CO2 has varied regularly by +/- 60 ppm. We are now 105 ppm (30%) past the highest recorded levels, and you think that's a "tiny increase"?
It's from the Vostok ice core (Petit 2000). You could also have found the same graph at NOAA, should you have bothered to look before declaring it made-up.
I'd say whatever contributes to that 200K cycle you mentioned. I could be wrong but I imagine those factors are not well understood.
That would be orbital variations, which are very well understood.
Also:
- Watch sea levels rise
- Watch unprecedented king tides and storm surges destroy billions in coastal property
- Watch millions of coastal & river delta farmers lose their farms due to salt
- Watch global threat levels rise from increased resource conflicts
- Watch temperatures rise
- Watch tropical diseases spread to new areas
- Watch unique and valuable reefs bleach and die
- Watch billions of tourism dollars disappear
- Watch rainfall patterns change drastically
- Watch farmers try to cope with drought & floods like they've never seen before
- Watch rising ocean acidification attack crucial food-web ecosystems
- Watch rising risks of runaway feedback from e.g. Siberian methane traps
- Watch deniers eventually change their tune to "oh well, it's too late to do anything now"
A "failed concept" that has resulted in numerous breakthroughs, such as beating a Go grandmaster with a fraction of the expected computing power.
And I imagine all further AI research will continue to be dismissed by you as "just algorithms" up to and including the day it finally produces an artificial True Scotsman.
Case closed, even you admit it.
You seem to be having trouble with the differences between selling data about you, and using data about you. Or maybe with reading.
"Easily".
Yep. Just use a browser add-on, such as this one. Most technical people know about them.
It's called a VPN retard
Settle down, petal. Running your PC's traffic through a VPN isn't hard, but it's still more work, expense, and expertise (and slower) than a browser add-on. And it's somewhat less easy if you want to run your entire network's traffic through one, including appliances like a Roku box etc.
And a VPN on your phone won't prevent carrier-installed rootkits from reporting anything they like. Remember Carrier IQ, which was found to be used by AT&T and many others (yes, even Apple), and was caught capturing keystrokes and passwords, copying and sending home texts etc? AT&T bought that not long ago, so it's clearly still in use - and that's only one we know about. Plus of course your carrier logs every call you make and every cell tower your phone touches, so good luck avoiding any of that.
And yes, all that personal data is used by AT&T to sell targeted ads on their AdWorks network, Sprint's Pinsight platform hosts other ad networks too, Verizon's is even bigger now they're buying Yahoo (plus they share your data with AOL), and Apple is no exception.
What makes you think Google, Facebook etc are so keen to sell your data while AT&T etc would never consider it, despite them knowing everything from your home address and daily movements to your TV watching habits and full browsing history?
Just like Google & Facebook, the major ISPs don't sell your data but do use it to run targeted ad networks of their own, taking full advantage of their far more extensive knowledge of you - and they're much harder to avoid. Examples of abuse abound, like Verizon being fined for their zombie supercookies, or AT&T charging an extra $29/month if you don't care to be targeted.
You can easily avoid Google or Facebook, but how do you avoid your only local broadband provider, or the telco you bought your phone from? It seems the GOP's answer is to avoid the internet completely.
The point here is, how to avoid unprincipled ISPs. It's trivial to go to duckduckgo instead of google, and a click will install a browser extension that blocks all site connections to facebook, but in a lot of the US it's not so easy to choose a different broadband provider or mobile telco.
Between multi-year contracts, locked-down or incompatible phones, lack of competition between duopolies, legal prohibitions on municipal broadband, strong pushback from customer service etc etc, it can be a significant undertaking to switch - assuming you have any reasonable alternatives at all in your area.