One of the most important problems that the nuclear-fission pioneers faced was precisely to decide the elements to work with. After long discussions, it was impossible to come to an agreement, that's why they decided the following: to hire a random person (he responded to an ad in a newspaper, I think), who was expected to arbitrarily select the 10 elements to be used in all the future fission reactors (the process was simple: a dart, a blindfold and a periodic table with all the known-at-that-time chemical elements). Unfortunately, this guy chose the less stable elements in the table! (I think that he committed suicide some years later. A quite sad story!). The radioactive ones, whose permanent instability is precisely what we call "radiation" (a very dangerous phenomenon which becomes geometrically more active after the fission process, which basically consists in throwing particles to atoms of these elements to break their nuclear bonds. And that makes them even more unstable?! Isn’t this incredible?). Some scientists didn’t like these results and proposed to better use other elements which were less dangerous (like what you are saying). In fact, most of the community thought that this was a good idea, but they have signed a contract with various providers and well.... Long story short: they had to stick with the most dangerous elements in the period table. And all the fission reactors have been fueled by radioactive materials until today!
Nah... I am joking, it is because of what PvtVoid suggests (the binding energy): you want heavy enough atoms whose particles can be easily separated (i.e., radioactive ones). This is precisely the opposite than what is required in fusion, where the lightest (and safest) elements are the ideal candidates; although this process is much more complicated and that’s why we firstly tried fission.
Even If this is the case, it shouldn't be acceptable either. This kind of statements coming from an organisation linked to the government (to national security; precisely assumed to have a good enough understanding about certain issues and to be a trustworthy source of information) shouldn't ever be tolerated or excused.
I quote from the video "...they knew how to use encrypted communications because of the E. Snowden revelations..."!!
It is not about defending/attacking Snowden or what he did. It is not even about a person working for a governmental agency publically and arbitrarily blaming someone for the Paris attacks by using a so strong language.
For me, the main problem here is that extremely important actions, like properly understanding/analysing/making decisions, are performed by clueless individuals. A person delivering the aforementioned nonsense should be immediately fired. People coming to so nonsensical conclusions are certainly responsible for lots of bad things.
It seems that your only intention is criticising SO/SE without properly understanding the situation and that's why you wrote two contradicting arguments: - "30% are shutdown because some Nazirator...". - "...encourage mediocracy" + "scam to promote the clueless"
By applying your example of sticking a screwdriver into an electric socket, a certainly clueless attitude which shouldn't be promoted, you would complain about the nazirator who is censoring such an attitude (or about the unfair downvotes or similar).
There are quite a few things which I don't like about SE (actually, I use SO almost exclusively); in fact, I have stopped participating in SO for over one year precisely because of not feeling like dealing with certain people. Additionally, I have seen (and not liked) quite a few attitudes on the lines of “Poor soul! Here you have my upvote to mood you up such that you don’t feel so bad about your nonsensical question”.
In any case, the global picture is very clear: the outputs are certainly worthy and SO has indeed filled an important gap in the online programming knowledge base.
It is impossible to behave as humans better than what humans can do. As far as predicting is basically simulating the given behaviour, there is nothing which can predict human behaviour better than what humans can do.
Translation of the previous paragraph to a more-appealing language (c-biased pseudo-code):
There are many (partial) solutions, but all of them have consequences and people (and corporations and governments) don't want to pay the consequences. Internal-combustion engines are optimised-over-100-years pieces of engineering which cannot be replaced right away (not in cars; but much less in trucks, ships and further bigger-and-more-contaminant engines). To not mention that there are quite a few lobbies which will make such a transition as hard as possible.
In any case, I do think that we will be gradually moving to electric engines, although this process will take quite a few years. But this will not be a perfect solution either: most of electricity comes from power plants which generate lots of pollution too.
Some year ago, I worked at an engineering consultancy specialising in IC-engine emissions. Back then, people in the industry were quite sceptical about meeting the strict targets in the upcoming regulations.
Engine manufacturers are making tremendous efforts to reduce emissions, but the whole situation is crazily hard: people want their cars to be more powerful but less contaminant?! This is impossible! Everyone has lied here: manufacturers, regulators (by promising what cannot be delivered; or perhaps by making important decisions without the required knowledge) and even people to themselves (by seriously expecting more powerful & less-contaminant cars).
I look forward to seeing all the liars taking responsibility for their actions and, hopefully, learning from their errors.
This is the second article about Live Coding here; in fact, I wrote a comment to the previous article. Back then, I said that the idea seemed interesting and was about to test it for self-promotional purposes. Finally, the experience turned to be HORRIBLE.
I was expecting it to be a business run by/addressed to programmers, genuinely interested in promoting programming; an assumption which was quickly proven wrong. Although I never felt any kind of interest in watching someone writing code, I believed ideas on the lines of "newbies find it very helpful".
This experiment didn’t last for too long and that's why I am not completely sure about the target audience. In any case, I met quite a few completely-clueless people not knowing anything about the given programming language, even about programming in general; asking random things and pointing out irrelevant problems. There were also cheerleaders, trying to trigger participation in the less-appealing-to-me way possible (e.g., talking about random things without caring about the code being written at all). There were also some people interested in knowing what was going on, but they were a minority.
As far as my intention was just taking this as an excuse to have videos where I was coding (like a picture of mine: a complement to my online references), I didn't mind too much the aforementioned issues. The worst part was the site itself (and their staff). They are certainly not programmers, but want to manage the whole thing their way. The result? I found the overall service/site very unappealing at different level (support, features, control on your videos, etc.). Imagine that Slashdot is being completely managed by Dice people (including moderation)!
My recommendation to any programmer planning to go down that road (mainly with Live Coding): this is not a site/concept managed by programmers. Accept this point and you might even enjoy the experience (so many people, so many minds).
Yes. As commented above, I am quite new in the comments section, but am liking what I have seen so far pretty much (there are always some exceptions; like the special guy above. I mean... I want to help everyone, but sometimes feel like getting a bit more relaxed and using humour without having to explain every single bit; currently living in a remote area where finding people properly understanding anything is quite difficult).
The AC two comments above has written a quite good summary about the Haselton issue; and all the ideas on this front are quite clear already. I certainly look forward to continue participating in this community.
It was a joke, apparently a bit too difficult for you. Sorry for not having written a simple enough set of ideas suitable for readers of any "background". Please, feel free to ask me anything you need to know; also I will try to avoid complex ideas/humour and use as simple words/concepts as I can.
I have been participating (= reading/writing comments) in Slashdot during just a few months, but am starting to like the overall attitude here quite a lot. It was just an innocent joke highlighting something curious (first time I have seen a so unanimous behaviour against something/someone); although I never doubted that there might be quite good reasons for it.
After quickly skimming through this article and even before reading the comments, my opinion was: too lengthy, unnecessarily detailed and describing something which is not news (tons of filtering algorithms are built every second). Additionally, the target behaviour (plagiarism of jokes) didn't seem particularly interesting.
My opinion about this guy continues being the same: I don’t know him and that's why cannot have a valid opinion; although I have certainly got a pretty bad first impression.
The situation you describe is the most logical output in this kind of operations. Companies only want profits and will never change what is already generating them.
Murdoch will certainly not affect the underlying ideas of National Geographic. A different story is the quality: reduction of expenses is usually the first step when moving from non-profit to for-profit.
Unfortunately, it is not 5 ms of properly-speaking power, like electricity you might use; this time refers just to one of the multiple steps required to generate electricity. By bearing in mind that even in case of holding such a step for months, as a conventional power plant does, lots of further problems would popup; for example: converting the unacceptably high temperatures into what is required.
From the point of view of generating electricity (converting fusion into an actually usable form of energy), 5 ms is even more ridiculously insignificant than what seems at a first sight.
OK. All this makes even more sense (-> I should have chosen this alternative the first time, rather than wasting my time answering someone like you).
PS: I have heard something about you (I am new here; in fact this has been my first modded-down comment), Elon fans, and the legends seem to be true. Although what I don’t understand is why you didn't modded down my other comment too. It basically implies that this whole thing is a scam proposed by a liar (too subtle perhaps?).
Compare that to a jet plane, and then STFU, FUD troll.
Highlighting a reality which you don't like is being a troll?!
What has 9/11 to do with all this? If a plane has a TECHNICAL problem, all the passengers might die, but most likely nobody else as far as might happen in a remote area (ocean). Additionally, planes have been tested for almost 100 years (and there are still accidents; quite a few of them during the last years) and run on a no-obstacle area (open air).
Is it so difficult to see that reaching plane-like speeds at the ground level is likely to provoke many more problems? We are talking about 50% faster than the fastest train ever (around 600 km/h)! By bearing in mind that the usual speed for high-speed trains is in the 300 km/h range!
A never-tested-before technology aims to be notably faster than the fastest ground transportation ever built and you don't see any problem with that? It is weird because you seem a pretty sensible person.
Additionally, are we sure that Elon Musk hasn't try this kind of things before? Are we sure that he is not related to Lyle Lanley? Because their speeches seem quite similar https://www.youtube.com/watch?...
Kind of surprised because of not seeing a single comment mentioning the recent SpaceX accident (+ delay in finding the problems + not too convincing problem reports). Not implying that Elon's companies are likely to deliver faulty technology, but that building extremely complex (transportation) systems is so difficult that cannot be mastered even after years of experience.
Hyperloop involves lots of never-tested-before features and is addressed to normal people (not to trained astronauts, perfectly aware of the risks). According to Wikipedia (https://en.wikipedia.org/wiki/Hyperloop), it is expected to have an average speed of 962 km/h (and a top one of 1220 km/h), like a plane at the ground level! And it would run across (densely) populated areas! What could possibly go wrong?
Recursively using simple parts to create more complex ones is a basic idea underlying any algorithm (it even explains how computers work). For example: a computer program can only understand a concept like movement by relying on simpler ideas like initial position, final position and time (which are also defined on account of other simpler ideas and so on until reaching pure 0s and 1s, only things which a computer can truly understand). "Detecting pedestrians" is so important that all what you are proposing (e.g., predicting pedestrian behaviour) cannot happen without it.
Your first paragraph explains what I have said in some other comments: this is a (pedestrian-)recognition algorithm and thus it has only to take care of performing proper recognitions (not what the proposed figures indicate). Properly recognising a pedestrian is the pre-step to perform many further actions (e.g., calling the whattodowithhumans() method); if a so basic feature fails, lots of further issues would also fail. In your own words "Once categorized as a pedestrian", represents the starting point for quite a few other decisions and doesn't still work fine. There is no problem in building all the remaining parts before (e.g., "guess whether the pedestrians are going to enter the roadway and take defensive driving precautions to avoid hitting them"), but you shouldn't say that the system is almost working when very important steps are still too faulty.
On the other hand, I don't think that traffic safety authorities will agree with your "even those who enter the roadway will generally be trying to avoid being hit by cars" and let autonomous cars run free. Also note that your "which may help explain why Google cars haven't been hitting 26% of the pedestrians they pass but instead 0%" is quite misleading as far as no single fully-autonomous car has ever been released in public streets (not as per my knowledge); in all the tests performed by Google, the cars had drivers. This is equivalently to what happens since day 1 when a person learns to drive; would you trust in the 0-day-experience novice driver ability to deal with any situation autonomously?
Regarding your "in other words, this isn't quite as basic of a problem as you make it out to be", I don’t agree with you and think that this is almost the same than trying to write a book by relying on an alphabet which has various missing letters (i.e., something completely unacceptable). If your intention is getting all the required permissions to allow such a system to be commercialised, it would certainly not happen (the whole system is still very far away from a first not-too-bad version). On the other hand, if your intention is just showing how cool and forward-thinking your company is, I guess that it is OK; your promotional-with-driver cars might continue running for other 5 years, before people getting bored of them.
All what you say after "even without this, autonomous cars are likely to be safer than human drivers" seems to show a knowledge-from-sci-fi-movies about which I don't feel like discussing.
One last comment, though: your "we're using a neural net rather than an algorithmic solution in the first place" seems to imply that you work at Google, not sure if in this exact project (not even sure if in actually-technical issues, as far as your whole speech seems too managerial to me). In any case, you should be happy as far as most likely will never have any money-related problem (I mean money/work-stability is a major concern for most of people and I guess that also for you) independently upon the future of this project.
PS: sorry about my late reply, but I am still new in Slashdot and haven’t got properly how notifications work. Apparently, I am only notified about logged-in users replying to my comments. I guess that you are called Anonymous Cowards for something (i.e., Slashdot does not take you too seriously).
When analysing the reliability of a given approach, you have to consider the worst scenario conditions. In this case, not being able to adequately recognise a pedestrian might be irrelevant or a tremendous problem. It is impossible to know where people or objects would be located with respect to the car and thus you cannot count on "perhaps I fail to recognise a person who is far away enough" (and another algorithm will take care of this determination; the recognition algorithm just has to worry about having an as high as possible success ratio). Additionally, bear in mind that the future safety tests, which this kind of approaches will have to pass before becoming commercially available, will certainly be based on ideas on these lines (toughest conditions; lots of tests and very small error margins).
On the other hand, note that the global figures published in technical papers are usually very positive (i.e., worst-scenario conditions are rarely tested, unless expressly mentioned in the paper), what makes much more sense with this partially-promotional attempt. I am quite sure that their 30% of errors would become much higher under tough-enough testing conditions. I am also quite sure that if they were completely certain about the reliability of their methodology, a much more detailed set of results would have been published.
In any case, we cannot know anything for sure without having access to detailed enough information.
Slashdot Mirror
One of the most important problems that the nuclear-fission pioneers faced was precisely to decide the elements to work with. After long discussions, it was impossible to come to an agreement, that's why they decided the following: to hire a random person (he responded to an ad in a newspaper, I think), who was expected to arbitrarily select the 10 elements to be used in all the future fission reactors (the process was simple: a dart, a blindfold and a periodic table with all the known-at-that-time chemical elements). Unfortunately, this guy chose the less stable elements in the table! (I think that he committed suicide some years later. A quite sad story!). The radioactive ones, whose permanent instability is precisely what we call "radiation" (a very dangerous phenomenon which becomes geometrically more active after the fission process, which basically consists in throwing particles to atoms of these elements to break their nuclear bonds. And that makes them even more unstable?! Isn’t this incredible?). Some scientists didn’t like these results and proposed to better use other elements which were less dangerous (like what you are saying). In fact, most of the community thought that this was a good idea, but they have signed a contract with various providers and well.... Long story short: they had to stick with the most dangerous elements in the period table. And all the fission reactors have been fueled by radioactive materials until today!
Nah... I am joking, it is because of what PvtVoid suggests (the binding energy): you want heavy enough atoms whose particles can be easily separated (i.e., radioactive ones). This is precisely the opposite than what is required in fusion, where the lightest (and safest) elements are the ideal candidates; although this process is much more complicated and that’s why we firstly tried fission.
Even If this is the case, it shouldn't be acceptable either. This kind of statements coming from an organisation linked to the government (to national security; precisely assumed to have a good enough understanding about certain issues and to be a trustworthy source of information) shouldn't ever be tolerated or excused.
I quote from the video "...they knew how to use encrypted communications because of the E. Snowden revelations..."!!
It is not about defending/attacking Snowden or what he did. It is not even about a person working for a governmental agency publically and arbitrarily blaming someone for the Paris attacks by using a so strong language.
For me, the main problem here is that extremely important actions, like properly understanding/analysing/making decisions, are performed by clueless individuals. A person delivering the aforementioned nonsense should be immediately fired. People coming to so nonsensical conclusions are certainly responsible for lots of bad things.
It seems that your only intention is criticising SO/SE without properly understanding the situation and that's why you wrote two contradicting arguments:
- "30% are shutdown because some Nazirator...".
- "...encourage mediocracy" + "scam to promote the clueless"
By applying your example of sticking a screwdriver into an electric socket, a certainly clueless attitude which shouldn't be promoted, you would complain about the nazirator who is censoring such an attitude (or about the unfair downvotes or similar).
There are quite a few things which I don't like about SE (actually, I use SO almost exclusively); in fact, I have stopped participating in SO for over one year precisely because of not feeling like dealing with certain people. Additionally, I have seen (and not liked) quite a few attitudes on the lines of “Poor soul! Here you have my upvote to mood you up such that you don’t feel so bad about your nonsensical question”.
In any case, the global picture is very clear: the outputs are certainly worthy and SO has indeed filled an important gap in the online programming knowledge base.
What has been your involvement in SO/SE/discourse.org at a programming level? (Kudos anyway. The results are certainly impressive).
Watch the video until the end. The pea is actually being pulled.
Nice video. But this is a levitation (objects are pushed) rather than a tractor (objects are pulled) beam.
The first part of the linked bbc.co.uk video is a bit unclear, watch it until the end to see the pulling effect.
It is impossible to behave as humans better than what humans can do. As far as predicting is basically simulating the given behaviour, there is nothing which can predict human behaviour better than what humans can do.
Translation of the previous paragraph to a more-appealing language (c-biased pseudo-code):
Function main()
{
behaviour = doBehaviour(true);
predAlgorithm = doBehaviour(false);
predHuman = doBehaviour(true);
countAlgorithm = 0;
countHuman = 0;
while (all elements in behavior/predAlgorithm/predHuman)
{
if(behavior_item == predAlgorithm_item) countAlgorithm ++;
else if(behavior_item == predHuman_item) countHuman ++;
}
If(countAlgorithm > countHuman) print “Review the whole algorithm because this result doesn’t make any sense”;
}
function doBehaviour(isHuman)
{
if(isHuman) return behave();
else return guessHowHumansBehave();
}
There are many (partial) solutions, but all of them have consequences and people (and corporations and governments) don't want to pay the consequences. Internal-combustion engines are optimised-over-100-years pieces of engineering which cannot be replaced right away (not in cars; but much less in trucks, ships and further bigger-and-more-contaminant engines). To not mention that there are quite a few lobbies which will make such a transition as hard as possible.
In any case, I do think that we will be gradually moving to electric engines, although this process will take quite a few years. But this will not be a perfect solution either: most of electricity comes from power plants which generate lots of pollution too.
Some year ago, I worked at an engineering consultancy specialising in IC-engine emissions. Back then, people in the industry were quite sceptical about meeting the strict targets in the upcoming regulations.
Engine manufacturers are making tremendous efforts to reduce emissions, but the whole situation is crazily hard: people want their cars to be more powerful but less contaminant?! This is impossible! Everyone has lied here: manufacturers, regulators (by promising what cannot be delivered; or perhaps by making important decisions without the required knowledge) and even people to themselves (by seriously expecting more powerful & less-contaminant cars).
I look forward to seeing all the liars taking responsibility for their actions and, hopefully, learning from their errors.
This is the second article about Live Coding here; in fact, I wrote a comment to the previous article. Back then, I said that the idea seemed interesting and was about to test it for self-promotional purposes. Finally, the experience turned to be HORRIBLE.
I was expecting it to be a business run by/addressed to programmers, genuinely interested in promoting programming; an assumption which was quickly proven wrong. Although I never felt any kind of interest in watching someone writing code, I believed ideas on the lines of "newbies find it very helpful".
This experiment didn’t last for too long and that's why I am not completely sure about the target audience. In any case, I met quite a few completely-clueless people not knowing anything about the given programming language, even about programming in general; asking random things and pointing out irrelevant problems. There were also cheerleaders, trying to trigger participation in the less-appealing-to-me way possible (e.g., talking about random things without caring about the code being written at all). There were also some people interested in knowing what was going on, but they were a minority.
As far as my intention was just taking this as an excuse to have videos where I was coding (like a picture of mine: a complement to my online references), I didn't mind too much the aforementioned issues. The worst part was the site itself (and their staff). They are certainly not programmers, but want to manage the whole thing their way. The result? I found the overall service/site very unappealing at different level (support, features, control on your videos, etc.). Imagine that Slashdot is being completely managed by Dice people (including moderation)!
My recommendation to any programmer planning to go down that road (mainly with Live Coding): this is not a site/concept managed by programmers. Accept this point and you might even enjoy the experience (so many people, so many minds).
Yes. As commented above, I am quite new in the comments section, but am liking what I have seen so far pretty much (there are always some exceptions; like the special guy above. I mean... I want to help everyone, but sometimes feel like getting a bit more relaxed and using humour without having to explain every single bit; currently living in a remote area where finding people properly understanding anything is quite difficult).
The AC two comments above has written a quite good summary about the Haselton issue; and all the ideas on this front are quite clear already. I certainly look forward to continue participating in this community.
It was a joke, apparently a bit too difficult for you. Sorry for not having written a simple enough set of ideas suitable for readers of any "background". Please, feel free to ask me anything you need to know; also I will try to avoid complex ideas/humour and use as simple words/concepts as I can.
Thanks for sharing all this information
I have been participating (= reading/writing comments) in Slashdot during just a few months, but am starting to like the overall attitude here quite a lot. It was just an innocent joke highlighting something curious (first time I have seen a so unanimous behaviour against something/someone); although I never doubted that there might be quite good reasons for it.
After quickly skimming through this article and even before reading the comments, my opinion was: too lengthy, unnecessarily detailed and describing something which is not news (tons of filtering algorithms are built every second). Additionally, the target behaviour (plagiarism of jokes) didn't seem particularly interesting.
My opinion about this guy continues being the same: I don’t know him and that's why cannot have a valid opinion; although I have certainly got a pretty bad first impression.
The situation you describe is the most logical output in this kind of operations. Companies only want profits and will never change what is already generating them.
Murdoch will certainly not affect the underlying ideas of National Geographic. A different story is the quality: reduction of expenses is usually the first step when moving from non-profit to for-profit.
Not really. Actually, I haven't ever heard about this guy before today.
I just wanted to feel special for a moment. Because there is not even a single comment (out of 58) showing any kind of appreciation for that guy!!
Unfortunately, it is not 5 ms of properly-speaking power, like electricity you might use; this time refers just to one of the multiple steps required to generate electricity. By bearing in mind that even in case of holding such a step for months, as a conventional power plant does, lots of further problems would popup; for example: converting the unacceptably high temperatures into what is required.
From the point of view of generating electricity (converting fusion into an actually usable form of energy), 5 ms is even more ridiculously insignificant than what seems at a first sight.
OK. All this makes even more sense (-> I should have chosen this alternative the first time, rather than wasting my time answering someone like you).
PS: I have heard something about you (I am new here; in fact this has been my first modded-down comment), Elon fans, and the legends seem to be true. Although what I don’t understand is why you didn't modded down my other comment too. It basically implies that this whole thing is a scam proposed by a liar (too subtle perhaps?).
Compare that to a jet plane, and then STFU, FUD troll.
Highlighting a reality which you don't like is being a troll?!
What has 9/11 to do with all this? If a plane has a TECHNICAL problem, all the passengers might die, but most likely nobody else as far as might happen in a remote area (ocean). Additionally, planes have been tested for almost 100 years (and there are still accidents; quite a few of them during the last years) and run on a no-obstacle area (open air).
Is it so difficult to see that reaching plane-like speeds at the ground level is likely to provoke many more problems? We are talking about 50% faster than the fastest train ever (around 600 km/h)! By bearing in mind that the usual speed for high-speed trains is in the 300 km/h range!
A never-tested-before technology aims to be notably faster than the fastest ground transportation ever built and you don't see any problem with that? It is weird because you seem a pretty sensible person.
Additionally, are we sure that Elon Musk hasn't try this kind of things before? Are we sure that he is not related to Lyle Lanley? Because their speeches seem quite similar
https://www.youtube.com/watch?...
Kind of surprised because of not seeing a single comment mentioning the recent SpaceX accident (+ delay in finding the problems + not too convincing problem reports). Not implying that Elon's companies are likely to deliver faulty technology, but that building extremely complex (transportation) systems is so difficult that cannot be mastered even after years of experience.
Hyperloop involves lots of never-tested-before features and is addressed to normal people (not to trained astronauts, perfectly aware of the risks). According to Wikipedia (https://en.wikipedia.org/wiki/Hyperloop), it is expected to have an average speed of 962 km/h (and a top one of 1220 km/h), like a plane at the ground level! And it would run across (densely) populated areas! What could possibly go wrong?
Recursively using simple parts to create more complex ones is a basic idea underlying any algorithm (it even explains how computers work). For example: a computer program can only understand a concept like movement by relying on simpler ideas like initial position, final position and time (which are also defined on account of other simpler ideas and so on until reaching pure 0s and 1s, only things which a computer can truly understand). "Detecting pedestrians" is so important that all what you are proposing (e.g., predicting pedestrian behaviour) cannot happen without it.
Your first paragraph explains what I have said in some other comments: this is a (pedestrian-)recognition algorithm and thus it has only to take care of performing proper recognitions (not what the proposed figures indicate). Properly recognising a pedestrian is the pre-step to perform many further actions (e.g., calling the whattodowithhumans() method); if a so basic feature fails, lots of further issues would also fail. In your own words "Once categorized as a pedestrian", represents the starting point for quite a few other decisions and doesn't still work fine. There is no problem in building all the remaining parts before (e.g., "guess whether the pedestrians are going to enter the roadway and take defensive driving precautions to avoid hitting them"), but you shouldn't say that the system is almost working when very important steps are still too faulty.
On the other hand, I don't think that traffic safety authorities will agree with your "even those who enter the roadway will generally be trying to avoid being hit by cars" and let autonomous cars run free. Also note that your "which may help explain why Google cars haven't been hitting 26% of the pedestrians they pass but instead 0%" is quite misleading as far as no single fully-autonomous car has ever been released in public streets (not as per my knowledge); in all the tests performed by Google, the cars had drivers. This is equivalently to what happens since day 1 when a person learns to drive; would you trust in the 0-day-experience novice driver ability to deal with any situation autonomously?
Regarding your "in other words, this isn't quite as basic of a problem as you make it out to be", I don’t agree with you and think that this is almost the same than trying to write a book by relying on an alphabet which has various missing letters (i.e., something completely unacceptable). If your intention is getting all the required permissions to allow such a system to be commercialised, it would certainly not happen (the whole system is still very far away from a first not-too-bad version). On the other hand, if your intention is just showing how cool and forward-thinking your company is, I guess that it is OK; your promotional-with-driver cars might continue running for other 5 years, before people getting bored of them.
All what you say after "even without this, autonomous cars are likely to be safer than human drivers" seems to show a knowledge-from-sci-fi-movies about which I don't feel like discussing.
One last comment, though: your "we're using a neural net rather than an algorithmic solution in the first place" seems to imply that you work at Google, not sure if in this exact project (not even sure if in actually-technical issues, as far as your whole speech seems too managerial to me). In any case, you should be happy as far as most likely will never have any money-related problem (I mean money/work-stability is a major concern for most of people and I guess that also for you) independently upon the future of this project.
PS: sorry about my late reply, but I am still new in Slashdot and haven’t got properly how notifications work. Apparently, I am only notified about logged-in users replying to my comments. I guess that you are called Anonymous Cowards for something (i.e., Slashdot does not take you too seriously).
When analysing the reliability of a given approach, you have to consider the worst scenario conditions. In this case, not being able to adequately recognise a pedestrian might be irrelevant or a tremendous problem. It is impossible to know where people or objects would be located with respect to the car and thus you cannot count on "perhaps I fail to recognise a person who is far away enough" (and another algorithm will take care of this determination; the recognition algorithm just has to worry about having an as high as possible success ratio). Additionally, bear in mind that the future safety tests, which this kind of approaches will have to pass before becoming commercially available, will certainly be based on ideas on these lines (toughest conditions; lots of tests and very small error margins).
On the other hand, note that the global figures published in technical papers are usually very positive (i.e., worst-scenario conditions are rarely tested, unless expressly mentioned in the paper), what makes much more sense with this partially-promotional attempt. I am quite sure that their 30% of errors would become much higher under tough-enough testing conditions. I am also quite sure that if they were completely certain about the reliability of their methodology, a much more detailed set of results would have been published.
In any case, we cannot know anything for sure without having access to detailed enough information.
I meant that the target expectation succeeded (i.e., in this case, proportion of failures).