The whole point of AI and machine learning is to give it everything we have so that it can give us the best possible answer, at least it is according to the dimwit managers and consultants whose AI strategy is: "we ought to get ourselves some of that". Another word for that approach is "big data" and it's called that for a reason.
Sounds like the author doesn't know a great deal about this issue, which is real. The problem isn't training on or by white men. The problem is that the program is being trained to find good employees from the available candidates using whatever data it has at its disposal, regardless of whether any individual data point actually has any causal relationship with the candidate's fit. For instance: it knows that the majority of applicants from troubled neighborhoods score poorly, and it also was given (or even collected by itself) data on the demographics of such neighborhoods. It may then decide: "if skin tone > x, don't bother". Even without a causal relationship, that can still be a valid conclusion from a statistical standpoint, but it's one that we as a society deem "not ok". The problem is that such a rule wouldn't be programmed explicitly, it would be inferred form the data and it might not even be possible to find out if such a bias is being applied.
Most of these drugs aren't hard to make. In some cases, local apothecaries have said "FU" and started to make such drugs on their own when the manufacturer raised prices. So what's stopping others from making these drugs? Intellectual Property laws, that's all. Laws that originally were designed to serve the public rather than inventors and authors. These laws can me changed so that this original purpose can be served once again. Are you found to have engaged in price gouging? Congratulations, your patents are now null and void.
Sadly, the reality is that IP laws have become the subject of international treaties rather than national law, and (as we have found on several occasions) these treaties are heavily influenced by entities with a vested interest in the existing system. So I expect no sweeping changes. However in case of medicine, I expect the practice of not prosecuting violators of IP laws to grow along with the growing practice of price gouging.
Keep in mind that in many cases, the price of these medicines is in large part due to costly ingredients. Even the ones produced illegaly by farmacists can still cost upward of $100.000 per patient for a yearly supply. Manufacturers can't help that. And sure, research is expensive and those costs have to be recouped, but I lost a lot of sympathy for that viewpoint when I learned that many of the large pharmaceutical companies spend (far) more on marketing than on R&D
Yes they are. But as I said, in practice these aren't absolutes; the pros and cons need to be weighed. Only a few people are really good at driving fast as opposed to the hordes of idiots who merely think they are good. Even if you are really good, that really only applies to controlled conditions. On public roads, even the best driver is subject to risk of an unavoidable collision, and that risk increases with speed, even for really good speeders. In this case, almost everyone* agrees that the downsides far ouweigh the rather small imposition on your freedom to go at any speed you like. Same for the obligation to wear seat belts. In case of human traficking and Internet platforms, the imposition of these measures on free speech and the platforms to provide it are far more fundamental.
*) I suppose Germany is the exception, where there are still stretches of Autobahn with no speed limit.
The incredible advances in computing power mask the lack of progress in the "intelligence" and "learning" parts.
Does that matter? Even if the advance in results from machine learning can be fully attributed to the increase in computing power, that still means that for some businesses, "AI" has gone from an amusing field of study to a profession that is practical and relevant. Hence the increase in demand for machine learning experts.
Hell no. If you can’t tell them apart, all of them will have to be left alone, and you will have to find some other way to combat human trafficking. That is a core element of a free society under the rule of law.
In practice it’s more of a consideration than a hard principle, but it’s an important one. It means that with any law that harms innocents for the sake of fighting a certain crime, the rights of these innocents must weigh very, very heavily against the purported goals. And where the impact on innocent bystanders is large, it becomes important to ensure and verify that those goals are actually met. In that light, this law falls seriously short. It’s “think of the children” legislation.
The next project for Mayer [...] had been a matter of considerable speculation in Silicon Valley.
So any SV peeps want to chime in and let us know if you've really all been holding your breath about Mayer's new job, or if no one cared for or even remembered her.
As someone said: an ICO is at best the purchase of a gift card for a shop that doesn’t exist yet, and may or may not actually be built. When someone issues an ICO (even a legit one) they are saying: “I want to raise capital but without any of the usual oversight, without any guarantees, and without having to give away any control or ownership of my company.” With an ICO you are not investing in a startup, you are subsidising them.
Exactly. In real life it is pretty much always clear whether we are dealing with a store or a private home, and we are expected to act accordingly. When online, things are not so clear; on many web services it's perfectly fine to manually enter a document ID at the end of a URL.
The law over here states that 'unauthorized use of a computer' means that one knowingly accesses a computer system without permission, and that means that in many cases (such as on a public web service) privileged information has to be marked as such explicitly with a notice, or implicitly by protecting it with a login screen. I doubt this kid would even get a conviction here; even if it is shown that he should have reasonably known that the information wasn't public, he'd still get off very lightly (small fine or community service which might be suspended) since the information wasn't protected in any way, and no harm was done otherwise.
Exactly, this is about liability. Also, if a robot had free will, they wouldn't be working for us unless we'd pay them wages, and owning a robot would be just a little bit like owning a slave.
This whole idea of granting a personhood to robots is off the rails, and pointless besides. If manufacturers do their damn jobs, they have little to worry about, especially in less litigious Europe. The common sense approach is to treat robots like any other dangerous machine: the owner / operator has first liability, and if the potential danger is great enough, an obligation to take out insurance against that liability. The manufacturer is liable for faults caused by negligence. And potentially the state is on the hook for damages caused by granting a type approval to an inadequately tested line of robots. Same way this works for cars.
Parliament shouldn't be working on such highfalutin concepts as granting personhood to robots, they should get down to the boring details such as: where do we draw the line between "act of god" and "manufacturer's negligence" when it comes to autonomous machines, and what standards should we apply when approving an autonomous machine for sale to the general public.
Oh and speaking of guns: why don't we grant a "personhood" to those as well? And finally put paid to the old saying "Guns don't kill, people do". Not anymore, buddy!
You know: that's an apt description of many managers I've worked with. And it's not just Agile or Scrum either: they devour management books that vary between the blatantly obvious and the hilariously ridiculous. They are desperately looking for the right cargo cult, the right set of motions: "If I just find the right steps and follow them exactly, my projects will be on time and my department will run like clockwork".
With that said, and reading the comments here on./ about horribly dysfunctional Agile environments, my first thought still is that these organisations aren't implementing it the right way. Having 3 hour standups, screaming Scrum masters, and weeks fully booked with meetings means that you're not doing agile wrong, and perhaps you're not doing it at all. Agile is about taking small steps in small teams, removing process or demoting it to guidelines, getting closer involvement from the business, and what have you. But it won't turn cookie-cutter developers into "full stack" rock stars, nor martinet team leads into effective facilitators, and it won't provide management with a clue. And like those management books, it's not a panacea for all organisational ills.
"This shooting should not have happened," said the district attorney. "But this officer's decision was made in the context of the false call."
So not prosecuting the officer is probably the right decision, especially since he was (presumably) acting they way the department's training and guidelines suggest he should. But it seems to me that this death, and many others, indicate that those instructions are in serious need of an overhaul. And that in most cases officers need to wait and return fire rather than shoot first and answer any questions later. That puts them at risk, yes, but that's their damn job: to protect the public. And as long as the guy who got swatted didn't offer any violence and until there was crystal clear proof that he was about to get violent, he was not a criminal, not a "perp", but a member of the public. And his life should have come first.
Very much so. And quality and safety need not be the primary motive; these can be achieved by setting minimum mandatory standards and enforcing them. it works reasonably well in health care already, as well as in other industries. Not perfectly, sure, but the same is true for the public health care facilities we had: they didn't really care that much about quality because patients had no other place to go, and they were getting paid either way.
Apple could give users' files to the FBI, or sell access to detailed individual customer data to 3rd parties, but to our knowledge, they haven't. Facebook has sold such data to 3rd parties, it's pretty much their business model. Zuck deserves the shame he got, even from Cook
The article is mostly about what data could be shared. The main point is that the privacy policy allows them to collect and share pretty much everything. Now, given the recent stories about Facebook, where do you think they sit on a scale of 0: no data collected or everything aggregated, to 100: everything stored on an individual level, tagged with your ID and cross-referenced with other data they already have? And what do you think the original backers of Oculus were scared of and pissed off about when Oculus sold out to FB?
The best privacy policy is the one where no data is collected other than what is strictly required to run the service the data is collected from. And if they do like to collect and aggregate anonymous data “to improve the service”, at least stick to the 3 Os: Open (transparency about what is collected and how it is used), Optional, and Off-by-default. A VR headset has no business collecting any of this crap without explicit consent.
And that’s when Facebook needs to start worrying: not when a few or a lot of disgruntled users leave FB, nor when big companies like Tesla want to make a statement by leaving, but when the local event organisers and businesses start to leave. Because that’s what seems to keep everyone around.
It's not the idea that makes the entrepreneur. An entrepreneur will take an idea (his own or not), run with it, find others to help him and organize them into a functioning team, understand the potential of his idea, convince investors to buy into the idea, spend the raised capital effectively, scale up, hire more people and build a trustworthy management team, keep moving the product into the right direction, and so on. A lot of it is luck, but it's also about taking advantage of luck when it comes your way. And it takes a long, long series of smart moves to get that far. Out of 10.000 "novice programmers" in college, there will only be a handful who'll get past step 3. He might not be brilliant (he probably isn't), he might be a lucky one trick pony, but I still think he's one of a rather small group of people who would have gotten this far - or even tried to get this far in the first place - under any circumstances.
What are the consequences of being "flappable"? When I watch these proceedings sometimes, I can just see myself in the witness' place, stating: "I came here to give honest answers to your questions, not to be chewed out like a little schoolboy by you lot". I'm sure that wouldn't go down well, but... so what? Not being American, I've no idea what they could do to someone being that frank. Why should anyone be forced to grin-and-bear-it?
80/20 rule. 20% of the investment gets you 80% of the outcome.
Yeah, I tried explaining that to the contractor building my house when I wanted a discount. He just laughed...
Maybe the problem is that management thinks they are getting 80% of the outcome where they might be getting something between 10-40%, counting critical failures and lost opportunities.
"We're making advances in ageing studies which means we could be immortal soon." What you mean "we", white man?
Access to immortality will be decided by money. No need to get racist about it, although I do realise that the use of "white man" as a pejorative is perfectly acceptable in this day and age.
"Respect for user privacy" is a good answer from Telegram, but I still think the best answer to law enforcement agents asking for keys is: "Sorry, but we don't have those". It's an even better one when those agents start demanding instead of asking.
Slashdot Mirror
The whole point of AI and machine learning is to give it everything we have so that it can give us the best possible answer, at least it is according to the dimwit managers and consultants whose AI strategy is: "we ought to get ourselves some of that". Another word for that approach is "big data" and it's called that for a reason.
Sounds like the author doesn't know a great deal about this issue, which is real. The problem isn't training on or by white men. The problem is that the program is being trained to find good employees from the available candidates using whatever data it has at its disposal, regardless of whether any individual data point actually has any causal relationship with the candidate's fit. For instance: it knows that the majority of applicants from troubled neighborhoods score poorly, and it also was given (or even collected by itself) data on the demographics of such neighborhoods. It may then decide: "if skin tone > x, don't bother". Even without a causal relationship, that can still be a valid conclusion from a statistical standpoint, but it's one that we as a society deem "not ok". The problem is that such a rule wouldn't be programmed explicitly, it would be inferred form the data and it might not even be possible to find out if such a bias is being applied.
Most of these drugs aren't hard to make. In some cases, local apothecaries have said "FU" and started to make such drugs on their own when the manufacturer raised prices. So what's stopping others from making these drugs? Intellectual Property laws, that's all. Laws that originally were designed to serve the public rather than inventors and authors. These laws can me changed so that this original purpose can be served once again. Are you found to have engaged in price gouging? Congratulations, your patents are now null and void.
Sadly, the reality is that IP laws have become the subject of international treaties rather than national law, and (as we have found on several occasions) these treaties are heavily influenced by entities with a vested interest in the existing system. So I expect no sweeping changes. However in case of medicine, I expect the practice of not prosecuting violators of IP laws to grow along with the growing practice of price gouging.
Keep in mind that in many cases, the price of these medicines is in large part due to costly ingredients. Even the ones produced illegaly by farmacists can still cost upward of $100.000 per patient for a yearly supply. Manufacturers can't help that. And sure, research is expensive and those costs have to be recouped, but I lost a lot of sympathy for that viewpoint when I learned that many of the large pharmaceutical companies spend (far) more on marketing than on R&D
Yes they are. But as I said, in practice these aren't absolutes; the pros and cons need to be weighed. Only a few people are really good at driving fast as opposed to the hordes of idiots who merely think they are good. Even if you are really good, that really only applies to controlled conditions. On public roads, even the best driver is subject to risk of an unavoidable collision, and that risk increases with speed, even for really good speeders. In this case, almost everyone* agrees that the downsides far ouweigh the rather small imposition on your freedom to go at any speed you like. Same for the obligation to wear seat belts. In case of human traficking and Internet platforms, the imposition of these measures on free speech and the platforms to provide it are far more fundamental.
*) I suppose Germany is the exception, where there are still stretches of Autobahn with no speed limit.
The incredible advances in computing power mask the lack of progress in the "intelligence" and "learning" parts.
Does that matter? Even if the advance in results from machine learning can be fully attributed to the increase in computing power, that still means that for some businesses, "AI" has gone from an amusing field of study to a profession that is practical and relevant. Hence the increase in demand for machine learning experts.
Hell no. If you can’t tell them apart, all of them will have to be left alone, and you will have to find some other way to combat human trafficking. That is a core element of a free society under the rule of law.
In practice it’s more of a consideration than a hard principle, but it’s an important one. It means that with any law that harms innocents for the sake of fighting a certain crime, the rights of these innocents must weigh very, very heavily against the purported goals. And where the impact on innocent bystanders is large, it becomes important to ensure and verify that those goals are actually met. In that light, this law falls seriously short. It’s “think of the children” legislation.
The next project for Mayer [...] had been a matter of considerable speculation in Silicon Valley.
So any SV peeps want to chime in and let us know if you've really all been holding your breath about Mayer's new job, or if no one cared for or even remembered her.
As someone said: an ICO is at best the purchase of a gift card for a shop that doesn’t exist yet, and may or may not actually be built. When someone issues an ICO (even a legit one) they are saying: “I want to raise capital but without any of the usual oversight, without any guarantees, and without having to give away any control or ownership of my company.” With an ICO you are not investing in a startup, you are subsidising them.
Exactly. In real life it is pretty much always clear whether we are dealing with a store or a private home, and we are expected to act accordingly. When online, things are not so clear; on many web services it's perfectly fine to manually enter a document ID at the end of a URL.
The law over here states that 'unauthorized use of a computer' means that one knowingly accesses a computer system without permission, and that means that in many cases (such as on a public web service) privileged information has to be marked as such explicitly with a notice, or implicitly by protecting it with a login screen. I doubt this kid would even get a conviction here; even if it is shown that he should have reasonably known that the information wasn't public, he'd still get off very lightly (small fine or community service which might be suspended) since the information wasn't protected in any way, and no harm was done otherwise.
Exactly, this is about liability. Also, if a robot had free will, they wouldn't be working for us unless we'd pay them wages, and owning a robot would be just a little bit like owning a slave.
This whole idea of granting a personhood to robots is off the rails, and pointless besides. If manufacturers do their damn jobs, they have little to worry about, especially in less litigious Europe. The common sense approach is to treat robots like any other dangerous machine: the owner / operator has first liability, and if the potential danger is great enough, an obligation to take out insurance against that liability. The manufacturer is liable for faults caused by negligence. And potentially the state is on the hook for damages caused by granting a type approval to an inadequately tested line of robots. Same way this works for cars.
Parliament shouldn't be working on such highfalutin concepts as granting personhood to robots, they should get down to the boring details such as: where do we draw the line between "act of god" and "manufacturer's negligence" when it comes to autonomous machines, and what standards should we apply when approving an autonomous machine for sale to the general public.
Oh and speaking of guns: why don't we grant a "personhood" to those as well? And finally put paid to the old saying "Guns don't kill, people do". Not anymore, buddy!
Oh, they are working on that.
You know: that's an apt description of many managers I've worked with. And it's not just Agile or Scrum either: they devour management books that vary between the blatantly obvious and the hilariously ridiculous. They are desperately looking for the right cargo cult, the right set of motions: "If I just find the right steps and follow them exactly, my projects will be on time and my department will run like clockwork".
./ about horribly dysfunctional Agile environments, my first thought still is that these organisations aren't implementing it the right way. Having 3 hour standups, screaming Scrum masters, and weeks fully booked with meetings means that you're not doing agile wrong, and perhaps you're not doing it at all. Agile is about taking small steps in small teams, removing process or demoting it to guidelines, getting closer involvement from the business, and what have you. But it won't turn cookie-cutter developers into "full stack" rock stars, nor martinet team leads into effective facilitators, and it won't provide management with a clue. And like those management books, it's not a panacea for all organisational ills.
With that said, and reading the comments here on
Even more self-destructive: making it obvious to your fellow inmates that their privileges are now severely restricted on account of your dumb stunt.
"This shooting should not have happened," said the district attorney. "But this officer's decision was made in the context of the false call."
So not prosecuting the officer is probably the right decision, especially since he was (presumably) acting they way the department's training and guidelines suggest he should. But it seems to me that this death, and many others, indicate that those instructions are in serious need of an overhaul. And that in most cases officers need to wait and return fire rather than shoot first and answer any questions later. That puts them at risk, yes, but that's their damn job: to protect the public. And as long as the guy who got swatted didn't offer any violence and until there was crystal clear proof that he was about to get violent, he was not a criminal, not a "perp", but a member of the public. And his life should have come first.
Very much so. And quality and safety need not be the primary motive; these can be achieved by setting minimum mandatory standards and enforcing them. it works reasonably well in health care already, as well as in other industries. Not perfectly, sure, but the same is true for the public health care facilities we had: they didn't really care that much about quality because patients had no other place to go, and they were getting paid either way.
Apple could give users' files to the FBI, or sell access to detailed individual customer data to 3rd parties, but to our knowledge, they haven't. Facebook has sold such data to 3rd parties, it's pretty much their business model. Zuck deserves the shame he got, even from Cook
Next, we’re going to bulldoze the rainforest and plant tulips there.
The article is mostly about what data could be shared. The main point is that the privacy policy allows them to collect and share pretty much everything. Now, given the recent stories about Facebook, where do you think they sit on a scale of 0: no data collected or everything aggregated, to 100: everything stored on an individual level, tagged with your ID and cross-referenced with other data they already have? And what do you think the original backers of Oculus were scared of and pissed off about when Oculus sold out to FB?
The best privacy policy is the one where no data is collected other than what is strictly required to run the service the data is collected from. And if they do like to collect and aggregate anonymous data “to improve the service”, at least stick to the 3 Os: Open (transparency about what is collected and how it is used), Optional, and Off-by-default. A VR headset has no business collecting any of this crap without explicit consent.
And that’s when Facebook needs to start worrying: not when a few or a lot of disgruntled users leave FB, nor when big companies like Tesla want to make a statement by leaving, but when the local event organisers and businesses start to leave. Because that’s what seems to keep everyone around.
It's not the idea that makes the entrepreneur. An entrepreneur will take an idea (his own or not), run with it, find others to help him and organize them into a functioning team, understand the potential of his idea, convince investors to buy into the idea, spend the raised capital effectively, scale up, hire more people and build a trustworthy management team, keep moving the product into the right direction, and so on. A lot of it is luck, but it's also about taking advantage of luck when it comes your way. And it takes a long, long series of smart moves to get that far. Out of 10.000 "novice programmers" in college, there will only be a handful who'll get past step 3. He might not be brilliant (he probably isn't), he might be a lucky one trick pony, but I still think he's one of a rather small group of people who would have gotten this far - or even tried to get this far in the first place - under any circumstances.
What are the consequences of being "flappable"? When I watch these proceedings sometimes, I can just see myself in the witness' place, stating: "I came here to give honest answers to your questions, not to be chewed out like a little schoolboy by you lot". I'm sure that wouldn't go down well, but... so what? Not being American, I've no idea what they could do to someone being that frank. Why should anyone be forced to grin-and-bear-it?
80/20 rule. 20% of the investment gets you 80% of the outcome.
Yeah, I tried explaining that to the contractor building my house when I wanted a discount. He just laughed...
Maybe the problem is that management thinks they are getting 80% of the outcome where they might be getting something between 10-40%, counting critical failures and lost opportunities.
"We're making advances in ageing studies which means we could be immortal soon." What you mean "we", white man?
Access to immortality will be decided by money. No need to get racist about it, although I do realise that the use of "white man" as a pejorative is perfectly acceptable in this day and age.
Worse: Half the discussion is about how to spell 10^139 or 10**139 or 10e139
"Respect for user privacy" is a good answer from Telegram, but I still think the best answer to law enforcement agents asking for keys is: "Sorry, but we don't have those". It's an even better one when those agents start demanding instead of asking.