|
Robotic studio takes fashion photos without a camera crew https://www.youtube.com/watch?v=lLABw4sCQbw
|
#
?
Feb 22, 2017 19:44
|
|
|
|
| # ? May 8, 2024 08:21 |
|
|
A Wizard of Goatse posted:p much, future technology could potentially eliminate human drudgery but technology we've had for your entire lifetime could factually eliminate human hunger. edit: Percentage-wise it looks way down in the developing world since 1970 - quote:Number of undernourished globally Cicero fucked around with this message at 19:58 on Feb 22, 2017 |
|
#
?
Feb 22, 2017 19:54
|
|
|
I wonder how much malnutrition in terms of starvation has been replaced by well, bad food? Is there a way to measure that?
|
|
#
?
Feb 22, 2017 20:09
|
|
|
Lightning Lord posted:I wonder how much malnutrition in terms of starvation has been replaced by well, bad food? Is there a way to measure that? what, like, vitamin A deficiency from eating only white rice vs. literally eating nothing? A Wizard of Goatse fucked around with this message at 20:24 on Feb 22, 2017 |
|
#
?
Feb 22, 2017 20:19
|
|
|
It doesn't count until everyone in Africa has unlimited access to free-range organic avocados
|
|
#
?
Feb 22, 2017 20:31
|
|
|
A Wizard of Goatse posted:there is no 'replacement', they're not comparable; a vitamin A deficiency from eating nothing but white rice is an incredible improvement in every regard over eating nothing I absolutely agree, I'm just wondering if that kind of thing is measured. I'm not saying that making sure everyone has SOMETHING to eat isn't important and laudable, but I do think throwing brown people scraps and making sure garbage food merchants are able to make money at their expense isn't good enough. We can do a lot better.
|
|
#
?
Feb 22, 2017 20:35
|
|
|
Sethex posted:That doesn't work in capitalism. So does welfare not exist in the fantasy marxoteen world or
|
|
#
?
Feb 22, 2017 20:41
|
|
|
Lightning Lord posted:I absolutely agree, I'm just wondering if that kind of thing is measured. I'm not saying that making sure everyone has SOMETHING to eat isn't important and laudable, but I do think throwing brown people scraps and making sure garbage food merchants are able to make money at their expense isn't good enough. We can do a lot better. yeah fair enough. malnutrition for the sake of those counts tries to incorporate all those things; there's a big focus on getting people nutritional supplements because outside of emergencies like war or natural disaster a lack of raw calories isn't generally the problem. Volkerball posted:So does welfare not exist in the fantasy marxoteen world or welfare in capitalist economies is workforce maintenance; it is there to keep people mothballed and able to eventually return to productivity in situations where temporary incapacity, unemployment, or life-consuming responsibilities like supporting family would otherwise mean a permanent end to their contribution to capital. the logic of the modern welfare state assumes close to everybody is either productively contributing to the economy, or will be soon, and in order to qualify for it you typically have to fill out a whole bunch of forms to confirm that you do, in fact, fit into that rubric. it's not just free money for the hell of it. A Wizard of Goatse fucked around with this message at 21:12 on Feb 22, 2017 |
|
#
?
Feb 22, 2017 20:45
|
|
|
‘What If We Try This?’ Asks Robotics Grad Student About To Eliminate 30% Of Workforce
|
|
#
?
Feb 22, 2017 23:30
|
|
|
A Wizard of Goatse posted:welfare in capitalist economies is workforce maintenance; it is there to keep people mothballed and able to eventually return to productivity in situations where temporary incapacity, unemployment, or life-consuming responsibilities like supporting family would otherwise mean a permanent end to their contribution to capital. the logic of the modern welfare state assumes close to everybody is either productively contributing to the economy, or will be soon, and in order to qualify for it you typically have to fill out a whole bunch of forms to confirm that you do, in fact, fit into that rubric. it's not just free money for the hell of it. That's not true at all. It's about maintaining order and prosperity in a society, if you leave a portion of people in the dust eventually you will have extensive social unrest and possibly revolts. There are many forms of welfare even in the US that are indefinite and not aimed at returning people to the workforce after a rough patch.
|
|
#
?
Feb 24, 2017 16:50
|
|
|
I'm surprised no ones talked about this: https://www.youtube.com/watch?v=mL9Ydk3w_8Q Say goodbye to a lot of accountants, lawyers, and doctors.
|
|
#
?
Feb 25, 2017 19:07
|
|
|
Dmitri-9 posted:I'm surprised no ones talked about this: Love the implied "and therefore you're fired" there.
|
|
#
?
Feb 25, 2017 19:30
|
|
|
If we really wanted to automate taxes, we'd have the IRS do the calculations first, and then send you a form saying "This is what we think you should get, click here to accept, or do it yourself if you think it's not right."
|
|
#
?
Feb 25, 2017 22:19
|
|
|
Coles and woolies the biggest supermarkets here down under have both been rolling out self checkouts for a few years. Recently they reveled that they are loosing some huge amount of money to theft through them. So naturally they Like gently caress you, you want more security you pay for it. Maybe they could find some robo cops?
|
|
#
?
Feb 25, 2017 22:53
|
|
|
Watson seems like such a well-marketed non-product. Can anyone point to a real application of it that either cut jobs or really increased productivity (moreso than, say, a well implemented database with a decent UI)?
|
|
#
?
Feb 25, 2017 22:54
|
|
|
Evil Robot posted:Watson seems like such a well-marketed non-product. Can anyone point to a real application of it that either cut jobs or really increased productivity (moreso than, say, a well implemented database with a decent UI)? Im phone posting so can't find a source, but I'm pretty sure Watson has already done some crazy stuff with regard to cancer treatments at some hospital. If I recall it was being used to read 100s of treatment reports a day and then offer suggestions to a treatment board when it noticed similarities in new cases with old ones. So not cutting jobs but increasing productivity and efficacy.
|
|
#
?
Feb 25, 2017 23:04
|
|
|
The problem with any AI system is false positives. You can make a medical system with the experience of 800 medics and 50.000.000 illness, but it will only be slighly better than a newbie medic with two days of experience, because without context computers are dumb, lack any common sense, make mistakes that a 8 years old would not make. In problems like "solve this boardgame problem" like chess, that is hard to see or is not present, the context don't exist or is too simple, so a chess program with the experience 800 medics and 50.000.000 games can end winning against a chess master, but in the real world the levels of complexity are a problem several orders of magnitude more complex than a computer can cope for. Fortunately, we have humans, so a expert system can advice and the human decide on it. What the system can't do is replace the human.
|
|
#
?
Feb 25, 2017 23:09
|
|
|
It can replace all their advisers and clerks though.
|
|
#
?
Feb 26, 2017 01:29
|
|
|
Tei posted:The problem with any AI system is false positives. You can make a medical system with the experience of 800 medics and 50.000.000 illness, but it will only be slighly better than a newbie medic with two days of experience, because without context computers are dumb, lack any common sense, make mistakes that a 8 years old would not make. What are you basing any of this on?
|
|
#
?
Feb 26, 2017 02:21
|
|
|
Tei posted:The problem with any AI system is false positives. You can make a medical system with the experience of 800 medics and 50.000.000 illness, but it will only be slighly better than a newbie medic with two days of experience, because without context computers are dumb, lack any common sense, make mistakes that a 8 years old would not make. From what I understand, the reason why deep learning systems are so effective is because they enable the matching and categorization of data in a contextual manner.
|
|
#
?
Feb 26, 2017 02:31
|
|
|
Dr. Arbitrary posted:If we really wanted to automate taxes, we'd have the IRS do the calculations first, and then send you a form saying "This is what we think you should get, click here to accept, or do it yourself if you think it's not right." P. sure H&R Block crushes this idea with $1 zillion of lobbying effort every time someone even thinks about bringing it up.
|
|
#
?
Feb 26, 2017 03:05
|
|
|
Dr. Arbitrary posted:If we really wanted to automate taxes, we'd have the IRS do the calculations first, and then send you a form saying "This is what we think you should get, click here to accept, or do it yourself if you think it's not right." Yeah and the entire tax prep industry lobbies against it
|
|
#
?
Feb 26, 2017 05:05
|
|
|
SwissCM posted:From what I understand, the reason why deep learning systems are so effective is because they enable the matching and categorization of data in a contextual manner. A algorithm don't understand context. They can forge a context but it look something like this "a)Yes, b)No c)Yes d)Yes e)Yes f)Yes e)Yes g)No" where they choose between 256 precreated answers. "a)Yes, b)No c)Yes d)Yes e)Yes f)Yes e)Yes g)No" is a laughable context if you ask a human. Owlofcreamcheese posted:What are you basing any of this on? I am software developer. Don't trust me. If you think something I say is wrong, probably is wrong, go find your own answer online reading technical articles, the algorithms that uses this type of software, etc. Just avoid take the word of a journalist of a blogger, go read the details themselves. Tei fucked around with this message at 08:55 on Feb 26, 2017 |
|
#
?
Feb 26, 2017 08:52
|
|
|
Tei posted:A algorithm don't understand context. They can forge a context but it look something like this "a)Yes, b)No c)Yes d)Yes e)Yes f)Yes e)Yes g)No" where they choose between 256 precreated answers. "a)Yes, b)No c)Yes d)Yes e)Yes f)Yes e)Yes g)No" is a laughable context if you ask a human. Yes, clearly the farthest computers have come is being able to manage 8 bits of binary data. The 1990s are gonna be wild though I bet when they get wilson running on an SNES and he can fake context with 32768 yeses and nos.
|
|
#
?
Feb 26, 2017 15:58
|
|
|
Owlofcreamcheese posted:Yes, clearly the farthest computers have come is being able to manage 8 bits of binary data. The 1990s are gonna be wild though I bet when they get wilson running on an SNES and he can fake context with 32768 yeses and nos. Humans are both good at generalist problems and expert problems. So when you have a expert human he can look at the details of a problem and can also look at the big picture. This helps ignore solutions that seems to make sense when you ignore the context.
|
|
#
?
Feb 26, 2017 16:27
|
|
|
Tei posted:Humans are both good at generalist problems and expert problems. So when you have a expert human he can look at the details of a problem and can also look at the big picture. This helps ignore solutions that seems to make sense when you ignore the context. Unless I was a millionaire going to doctor house who might make up brand new solutions to my medical problems I'd rather have a system that searched a bunch of databases really well and scientifically applied the best available medical studies perfectly than have some guy follow his heart and emotions to pick what pills I need.
|
|
#
?
Feb 26, 2017 16:40
|
|
|
Especially since sometimes those emotions include stuff like "black guys don't really feel pain as much and are just drug seeking."
|
|
#
?
Feb 26, 2017 16:52
|
|
|
Tei posted:A algorithm don't understand context. They can forge a context but it look something like this "a)Yes, b)No c)Yes d)Yes e)Yes f)Yes e)Yes g)No" where they choose between 256 precreated answers. "a)Yes, b)No c)Yes d)Yes e)Yes f)Yes e)Yes g)No" is a laughable context if you ask a human. You are describing basic algorithmic software which is typically written by a human. This is not how neural networks work. Take image processing for example, if you have a network built to detect fruit, you don't define the attributes of apples and bananas and use detection algorithms like edge finding and blob finders to narrow down the possibilities and classify the results, Instead you simply feed it data, tons of data. Let's say, 500 images each of a variety of fruits. Then, when you feed it a new image of a fruit, it will be able to narrow down the possibilities by using only data it has received as not an array of complex but ultimately dumb algorithms. It works similarly to the human mind, albeit not as sophisticated, requiring significantly larger amounts of data than a human to make a determination.
|
|
#
?
Feb 26, 2017 17:05
|
|
|
Taffer posted:You are describing basic algorithmic software which is typically written by a human. This is not how neural networks work. Take image processing for example, if you have a network built to detect fruit, you don't define the attributes of apples and bananas and use detection algorithms like edge finding and blob finders to narrow down the possibilities and classify the results, Instead you simply feed it data, tons of data. Let's say, 500 images each of a variety of fruits. Then, when you feed it a new image of a fruit, it will be able to narrow down the possibilities by using only data it has received as not an array of complex but ultimately dumb algorithms. It works similarly to the human mind, albeit not as sophisticated, requiring significantly larger amounts of data than a human to make a determination. Watson specifically isn't neural nets or anything. The whole gimmick of watson is that it's just very strictly traditional database stuff but run at a scale that wasn't possible in the past. So there is no real specific breakthrough to it, or superscience programming magic. It's mostly just the idea of "what if instead of running one algorithm on one database we ran 100 algorithms on 100 databases and had some good functions to combine the results into a single answer" but did that way faster than used to be possible and it turned out that works shockingly well for a lot of domains.
|
|
#
?
Feb 26, 2017 21:53
|
|
|
Dr. Arbitrary posted:If we really wanted to automate taxes, we'd have the IRS do the calculations first, and then send you a form saying "This is what we think you should get, click here to accept, or do it yourself if you think it's not right." For the record, this is how most European countries do things.
|
|
#
?
Feb 26, 2017 23:44
|
|
|
Owlofcreamcheese posted:Unless I was a millionaire going to doctor house who might make up brand new solutions to my medical problems I'd rather have a system that searched a bunch of databases really well and scientifically applied the best available medical studies perfectly than have some guy follow his heart and emotions to pick what pills I need. Sure. But what's actually happening in the automated version is that a human is using their knowledge and skills to observe your problems, using their judgement to make decisions, and then using their expertise to condense the results of their observations and judgement into something they can enter into what is essentially a specialized search engine.
|
|
#
?
Feb 27, 2017 00:31
|
|
|
Guavanaut posted:Especially since sometimes those emotions include stuff like "black guys don't really feel pain as much and are just drug seeking." We can teach a computer to think that.
|
|
#
?
Feb 27, 2017 01:48
|
|
|
Dr. Arbitrary posted:We can teach a computer to think that. Yeah, one of the fun things about neural networks is they're highly 'impressionable'. Biases like that can sneak into your neural networks quite easily.
|
|
#
?
Feb 27, 2017 02:18
|
|
|
Taffer posted:You are describing basic algorithmic software which is typically written by a human. This is not how neural networks work. Take image processing for example, if you have a network built to detect fruit, you don't define the attributes of apples and bananas and use detection algorithms like edge finding and blob finders to narrow down the possibilities and classify the results, Instead you simply feed it data, tons of data. Let's say, 500 images each of a variety of fruits. Then, when you feed it a new image of a fruit, it will be able to narrow down the possibilities by using only data it has received as not an array of complex but ultimately dumb algorithms. It works similarly to the human mind, albeit not as sophisticated, requiring significantly larger amounts of data than a human to make a determination. https://studio.azureml.net/ If you have any MS account it is free. Play around with it a bit. They made it trivial to implement.
|
|
#
?
Feb 27, 2017 02:24
|
|
|
Dead Reckoning posted:Now, you could try to revive the murdered corpse of the international labor movement, except 1) that would require accepting the same standard of living for you and yours as people in third world FTZs, you will lose while they gain, 2) the owner class now also owns all the platforms and means of communication that you could realistically use to organize such a thing, and 3) it still doesn't address the long term problem of surplus labor force. Ah yes, remember that time back in the 1800s, when workers could more easily and more freely communicate with their international class comrades. You know, back when the owners didn't own the platforms and means of communication?
|
|
#
?
Feb 27, 2017 09:49
|
|
|
Dr. Arbitrary posted:We can teach a computer to think that. Ratoslov posted:Yeah, one of the fun things about neural networks is they're highly 'impressionable'. Biases like that can sneak into your neural networks quite easily.
|
|
#
?
Feb 27, 2017 12:25
|
|
|
"A computer will never take my job" - Literal human calculators from the 50s who used to do all the arithmetic for nuclear weapons If you think your job is safe from automation, and your job can be proceduralized in any way, you are wrong. Here's how it goes. You get a nice job. But you live in the 1st world, so you cost a lot of money. So they get you to develop better and better work instructions until they can offshore that poo poo to a poor country. Then they start investing in capital, because 1st world capitalists don't particularly like having to deal with 3rd world countries, they do it because the people are cheap and exploitable. This capital is called "automation." 1st world people, like you if you are lucky enough to understand it, work on it until the process of building the "automation" has better work instructions, then they offshore that poo poo too, and get a smaller subset of 1st world employees to build a way to automate the process of building the existing automated architectures. And on and on it goes, until all the product is made by capital and a small subset of artisans if necessary. People tend to like poo poo with a bit of a human touch. I could see the capital being located in the 3rd world, but only if the capitalists believe the country is secure enough, which right now is not descriptive of most of these countries.
|
|
#
?
Feb 27, 2017 14:12
|
|
|
Main Paineframe posted:Sure. But what's actually happening in the automated version is that a human is using their knowledge and skills to observe your problems, using their judgement to make decisions, and then using their expertise to condense the results of their observations and judgement into something they can enter into what is essentially a specialized search engine. Again, maybe you go to doctor house MD and he is using his genius brain to think up brand new diagnosis and then curing them with brand new treatments. But most of real medicine is failable doctors trying to vaguely remember symptoms within the limits of human memory then poorly looking up the symptoms they don't remember then giving treatments based on whatever the research said the last time they looked at the research mixed with how they "feel" about it. It is very very likely reducing whole sections of medicine to cold hard data is going to be extremely good news for those branches of medicine. Bedside manner is nice too. And feeling like someone cares about you. But if my kid has leukemia maybe I can hire a clown or something to make the kid feel happy then have a cold boring database actually look up every medical study every conducted about leukemia to figure out what treatment has the best success rate.
|
|
#
?
Feb 27, 2017 14:24
|
|
|
Owlofcreamcheese posted:But most of real medicine is failable doctors trying to vaguely remember symptoms within the limits of human memory then poorly looking up the symptoms they don't remember then giving treatments based on whatever the research said the last time they looked at the research mixed with how they "feel" about it.
|
|
#
?
Feb 27, 2017 15:47
|
|
|
|
| # ? May 8, 2024 08:21 |
|
|
Guavanaut posted:I've seen open Wikipedia tabs for names of medicines before at doctors' offices. I think (hope) they were just checking alternative names for poo poo before checking it out in the actual formulary. Haha they arent. If it makes you feel worse in America the system they place perscriptions to allows advertising so companies can match your anticpated disease with a doctors office and have their drug pop up as the doctor goes to look for prescriptions. Thats what pills that cost 80,000 for a full treatment buys you.
|
|
#
?
Feb 27, 2017 16:46
|
|
