Agreed. But Ford is much likelier to deliver a commercial product than Google is. As they ship TJA in 2017, I'd expect Ford to announce, "In 5 more years (2022), we'll ship a car that's fully automatic".
Of course by 2017, Mercedes and Lexus and even Hyundai will probably ship autocars of their own.
RTFA. Ford isn't promising full autonomy. Their "Traffic Jam Assist" is pretty close to what Mercedes already offers -- the ability to trail along behind another car and automatically adapt your speed to theirs. TJA only adds the ability to track the car ahead and steer with it. To me that seems quite achievable within 5 years.
Sebastian Thrun and Google have already done much more wuth the Google Autocar. I woudn't be surprised if by 2017 the GA will be fully and reliably autonomous. The challenge probably isn't the algorithms but the instrumentation. Somehow the production cars will need to spray out several light and radar beams and make reliable sense of the reflection, all within the shape of a car that looks normal and withstands snow coverage and the incomplete removal thereof. That typical continuing level of everyday soccer mom abuse will limit full autonomy for a while yet, but at no fault of Ford (or Google).
Great advice. I think techy types can especialy see the value in: 2) keep climbing up and 3) take ownership. These two foci show that you're keeping your skills up to date, perhaps even advancing them beyond the state-of-the-art, and then applying them where it counts -- to the bottom line. If you can do that, any decent business or manager should appreciate that you're doing all you can to make a real impact where it matters most.
I agree it's also helpful (though annoying) to add 4) self promotion. But if you've already "taken ownership" of problems or processes, then at least you clearly deserve the recognition for whatever you're promoting. By focusing on the advancement of your tech skills and then manifesting them via task ownership, it shouldn't be too painful to self-promote those results in politically effective yet palatable forms.
Finally, I'd also propose goal 6) Promote Others In Your Team. If you can join hands with coworkers toward the other goals you've listed, management is likelier to appreciate the value of your work, much more than if you do it solo. Others in the team will be more willing to sing your praises if they think it'll help to float their boat too.
I guarantee, if a biological process can be replaced with a computer simulation, it has been. Computer simulations of drug pharmacodynamics and pharmacokinetics are MUCH faster and cheaper than animal testing. But they're also seriously inadequate in measuring drug efficacy or toxicity. If they worked better, you'd hear much more about their use by EVERY biology research team in academia or medicine, or pharmas. The fact that you don't should tell you something.
For the forseeable future, nothing can replace controlled (and blinded) trials on animals. The most that computational scientists who work at pharmas (like me) can hope for is to minimize the cost and damage that's incurred in the hunt for better drugs.
You don't agree? Then the next time someone you love gets sick, be sure to withhold that antibiotic.
As a computer scientist, I work for a large pharma in a group that builds computational models, does image analysis, and uses machine learning. We supplement the work of scientists in wet labs and reduce the amount and severity and cost of animal experimentation. Personally, when I was a student, I volunteered for several Phase I drug studies. I think I'm not unusual here. I believe in what I do.
Unfortunately animal testing is nowhere near good enough to evaluate the efficacy or toxicity of a new drug, but it's the best tool we have. Computational methods are much more limited than using animals, and that will continue for many years to come. Life is incredibly complex and wondrously hard to reduce to simple mechanisms for simulation. We use quant models mostly for evaluating potential dose ranges of candidate drugs before First-In-Man / Phase-I studies. We also use them to find and develop new biomarkers (new computational metrics which correlate with the biological outcomes that we need to predict or improve upon). Quant models have shown some value in pharma, but for now and a long time yet, if you want new drugs, you're going to have to test them on animals.
I can't speak for everyone in pharma, but I think we all want to do as much as we can to minimize suffering -- for humans as well as test animals. Remember too that in the wild, the life of a mouse, rat, or rabbit will be scary, brutal, and short. At a pharma that life will still be short, but it'll also be boring and painless. That's a tradeoff that a lot of prey animals like these just might prefer.
If Gallup's pollsters call only land lines and no cell phones, then they're going to reach an older and poorer sample than they should. I suspect a high fraction of college grads under 40 don't even own land line phones any more. I don't.
Biasing the sample downward in this way could help to explain the poll's rather bizarre results.
A/B focus testing is about observing how customers or users choose between two alternatives based on their qualitative sense of aesthetics. ML is about classifying data based on quantifying the data into defined classes or toward optimal values.
Predicting the outcome of a focus group is a completely different problem than multi arm slot machines. In focus groups there is no objective metric, so focus group problems are not amenable to machine learning unless your machine can define, measure, and perhaps predict aesthetic criteria.
Now THAT I'd like to see.
Re:True AI would dominate the world. YAY!
on
Where's HAL 9000?
·
· Score: 1
I disagree. Domination of the human race by AI might be the best thing ever. To quote Lennon, "Imagine there's no country..."
The petty squabbles that we humans are so good at turning into mechanized genocides would end. No machine overlord would tolerate such crap.
Every politician would be out of work, since AI would manage resources equitably and hundreds of times more efficiently.
Police forces would cease to exist. It'd be trivial for AI to monitor everyone's movements and quickly find, isolate, and neutralize every human threat.
Every court case would be tried by a scrupulously fair arbiter. No more need for ambitous judges or juries of fools.
Everyone would be out of a job... on permanent vacation. Money would disappear. All our needs would be provided by automated production systems. For free.
No more hunger, crime, war, or injustice. And you think that's a bad thing? O Ye of little faith...
So will True AI actually lead to this utopia? Who the hell knows? But I do know that scifi literatue is a damned silly place to look for the answer.
Look at the 10 states that did best. Almost all are rural: Massachusetts, Montana, North Dakota, Utah, Vermont, Colorado, Minnesota, New Hampshire, South Dakota, Virginia.
Texas and Montana did better than NY or CA?
Time to ask, 'What and who were tested?' I suspect the sample was far from uniformly distributed across all US 8th graders.
> Until they do that, it'll still be just videotapes of college lectures.
's/just/free/g' That's a huge mod.
It's silly to compare the curent crop of on-line FREE courses with tuition-based in-class courses. Better to compare them with tuition-based on-line courses, like those of Columbia CVN or Stanford SEE. By that standard, the free courses I've seen are as good or better. And remember, each of the freebies costs $2500 less. That's a world of difference.
Will on-line courses ever approach the learning experience provided by in-class lectures or face-to-face study groups? Possibly. But given the GREATLY reduced cost, the ease of time shifting, the opportunity to learn from the very best profs (eventually), cut-rate e-learning isn't going away. In fact it's hard to imagine that it won't eventually outcompete the present college system, given the unreal cost difference.
Given the rise of Walmart and Amazon, you should never underestimate the power of delivering a product by combining 'sufficient', 'convenient', and 'dirt cheap'.
This article would be more credible if the author weren't Norm Matloff, a statistics prof who's been bemoaning the invasion of H1Bs into the software business for over a decade. Now that the demand for cheap labor has left the building, I guess he's turned to stomping sour grapes, "You shouldn't have wanted the job anyway".
Not so different from a May 2010 pioece in IEEE Spectrum where they built a laser targeting system for backyard mosquitos.
What we really need is the marriage of both products -- a laser canon that automatically tracks and vaporizes squirrels. And starlings. And grackles. And cowbirds...
So if the camera fails to see your license plate you get no gas? Clever. I'm sure that another car or truck will never obstruct the camera's view, that snow will never obscure the plate, that fog will never blur the plate letters, that the plate will always be adequately illuminated, that the cameras will never break down, that the license database will always be up-to-date and on-line. No flies in THAT ointment, no sir.
All this fal-de-ral just to make sure that a few people pay their vehicle tax? Why not simply require everyone to pay their tax annually when they register their vehicle? Put a sticker on the windshield showing that the tax was paid, LIKE THEY DO EVERYWHERE ELSE.
Or if you must monitor everyone's tax status minute-by-minute, have everyone carry a tax-paid UPC fob that is scannable by a credit card swiper (or an attendant) when you pay for your gas? Would that cost, oh perhaps, a BILLION pounds less than buying and wiring up multiple spy cameras for every service station in the UK?
Who comes up with ideas this overcomplex, ineffective, and brain damaged? Newt Gingrich's british cousin?
Actually, breaking into a strong safe exactly like breaking strong encryption. Both will require greater-than-average expenditures of time and money to crack, but given enough resources, both are possible.
If the state chooses not to expend the resources, that doesn't change the role of the defendant. It only makes it more likely that the prosecutors will whine to the judge about the unfairness of it all, and the judge will then throw the defendant's ass in jail until s/he capitulates and opens up the safe or removes the encryption.
The relevance of the 5th amendment to encryption seems to be, what evidence is necessary to justify the search of your computer? If 1) contraband or your acquisition of contraband has been observed, and 2) containers that *might* contain contraband are observed in your possession, then you're screwed. If 1 but not 2, this ruling says you're OK. If 2 but not 1, then you're OK as long as you wait out the imprisonment for contempt because you should eventually win your appeal of the search warrant since it lacked probable cause.
So the moral of the story is: make sure you hide your encrypted containers.
CNN has a nice article that puts "The Fair Labor Association" in proper perspective:
"Apple's major move has been to announce that it has joined an organization called the Fair Labor Association, which will "audit" Apple's factories. According to Apple, the Fair Labor Association is an independent watchdog that will work tenaciously to hold Apple and its suppliers accountable.
Unfortunately, while there are some fine people at the association, the organization is not the independent watchdog Apple claims it to be. Indeed, most of its money -- millions of dollars per year -- comes from the very companies whose labor practices it is supposed to scrutinize. Although Apple has not disclosed its financial relationship with the Fair Labor Association, it is likely now the organization's largest funder. Moreover, on the association's board of directors sit executives of major corporations such as Nike, Adidas and agribusiness giant Syngenta. The job of these executives is to represent the interests of other member companies, such as Apple. Under the Fair Labor Association's rules, the company representatives on the board exercise veto power over major decisions."
No. Large or academically motivated tech employers strongly prefer degrees from bigger name schools. There is no comparing the employment prospects of someone from MIT vs some low brow state school. Huge difference, fair or not.
Engineering master's are essential for advancement to technical project management. If the degree follows a few years of commercial experience, they're worth more on the floor than a PhD, since the latter targets research and not production.
An MBA who lacks engineering chops is useless toward leading a tech team to deliver a product that works and arrives on time. And useless for pretty much everything else too.
> 'I think it will actually be a long time, maybe never, when actual Stanford degrees would be given for fully online work...
Lots of fine universities already offer degrees entirely on-line (especially MS in engineering). Unlike the dean, I am certain Stanford will do this. But it'll cost... full tuition.
I doubt Thrun intends to offer a few courses and stop there. I think he'll offer an entire CS curriculum within maybe 3 years, and offer some soft of CS degree program soon thereafter.
It seems like you could offer other degrees using this same technology -- probably all engineerings, physics, probably math and statistics, maybe biology (but without labs).
Not only would the degrees be FREE (a huge thing for the poor in the third world and BRIC countries), but they'd be FAST. By excluding all the non-essentials, the equivalent of a BS in CS could be completed three times faster, in no more than 1.5 years.
Based on what I've seen from Thrun so far, I bet the degree will be widely respected, and frankly, better than 3/4 of today's CS degrees.
Universities beware. You're about to run smack into The Innovator''s Dilemma. And in my humble opinion, it's about damned time.
Slashdot Mirror
Agreed. But Ford is much likelier to deliver a commercial product than Google is. As they ship TJA in 2017, I'd expect Ford to announce, "In 5 more years (2022), we'll ship a car that's fully automatic".
Of course by 2017, Mercedes and Lexus and even Hyundai will probably ship autocars of their own.
RTFA. Ford isn't promising full autonomy. Their "Traffic Jam Assist" is pretty close to what Mercedes already offers -- the ability to trail along behind another car and automatically adapt your speed to theirs. TJA only adds the ability to track the car ahead and steer with it. To me that seems quite achievable within 5 years.
Sebastian Thrun and Google have already done much more wuth the Google Autocar. I woudn't be surprised if by 2017 the GA will be fully and reliably autonomous. The challenge probably isn't the algorithms but the instrumentation. Somehow the production cars will need to spray out several light and radar beams and make reliable sense of the reflection, all within the shape of a car that looks normal and withstands snow coverage and the incomplete removal thereof. That typical continuing level of everyday soccer mom abuse will limit full autonomy for a while yet, but at no fault of Ford (or Google).
Great advice. I think techy types can especialy see the value in: 2) keep climbing up and 3) take ownership. These two foci show that you're keeping your skills up to date, perhaps even advancing them beyond the state-of-the-art, and then applying them where it counts -- to the bottom line. If you can do that, any decent business or manager should appreciate that you're doing all you can to make a real impact where it matters most.
I agree it's also helpful (though annoying) to add 4) self promotion. But if you've already "taken ownership" of problems or processes, then at least you clearly deserve the recognition for whatever you're promoting. By focusing on the advancement of your tech skills and then manifesting them via task ownership, it shouldn't be too painful to self-promote those results in politically effective yet palatable forms.
Finally, I'd also propose goal 6) Promote Others In Your Team. If you can join hands with coworkers toward the other goals you've listed, management is likelier to appreciate the value of your work, much more than if you do it solo. Others in the team will be more willing to sing your praises if they think it'll help to float their boat too.
I guarantee, if a biological process can be replaced with a computer simulation, it has been. Computer simulations of drug pharmacodynamics and pharmacokinetics are MUCH faster and cheaper than animal testing. But they're also seriously inadequate in measuring drug efficacy or toxicity. If they worked better, you'd hear much more about their use by EVERY biology research team in academia or medicine, or pharmas. The fact that you don't should tell you something.
For the forseeable future, nothing can replace controlled (and blinded) trials on animals. The most that computational scientists who work at pharmas (like me) can hope for is to minimize the cost and damage that's incurred in the hunt for better drugs.
You don't agree? Then the next time someone you love gets sick, be sure to withhold that antibiotic.
As a computer scientist, I work for a large pharma in a group that builds computational models, does image analysis, and uses machine learning. We supplement the work of scientists in wet labs and reduce the amount and severity and cost of animal experimentation. Personally, when I was a student, I volunteered for several Phase I drug studies. I think I'm not unusual here. I believe in what I do.
Unfortunately animal testing is nowhere near good enough to evaluate the efficacy or toxicity of a new drug, but it's the best tool we have. Computational methods are much more limited than using animals, and that will continue for many years to come. Life is incredibly complex and wondrously hard to reduce to simple mechanisms for simulation. We use quant models mostly for evaluating potential dose ranges of candidate drugs before First-In-Man / Phase-I studies. We also use them to find and develop new biomarkers (new computational metrics which correlate with the biological outcomes that we need to predict or improve upon). Quant models have shown some value in pharma, but for now and a long time yet, if you want new drugs, you're going to have to test them on animals.
I can't speak for everyone in pharma, but I think we all want to do as much as we can to minimize suffering -- for humans as well as test animals. Remember too that in the wild, the life of a mouse, rat, or rabbit will be scary, brutal, and short. At a pharma that life will still be short, but it'll also be boring and painless. That's a tradeoff that a lot of prey animals like these just might prefer.
If Gallup's pollsters call only land lines and no cell phones, then they're going to reach an older and poorer sample than they should. I suspect a high fraction of college grads under 40 don't even own land line phones any more. I don't.
Biasing the sample downward in this way could help to explain the poll's rather bizarre results.
You're right. My criticism was misdirected. The article is fine; it's not about ML or focus groups but minimizing trial size.
It was the Slashdot summary that somehow saw it as 'ML Replaces Focus Groups'. Thee-a-culpa.
A/B focus testing is about observing how customers or users choose between two alternatives based on their qualitative sense of aesthetics. ML is about classifying data based on quantifying the data into defined classes or toward optimal values.
Predicting the outcome of a focus group is a completely different problem than multi arm slot machines. In focus groups there is no objective metric, so focus group problems are not amenable to machine learning unless your machine can define, measure, and perhaps predict aesthetic criteria.
Now THAT I'd like to see.
I disagree. Domination of the human race by AI might be the best thing ever. To quote Lennon, "Imagine there's no country..."
The petty squabbles that we humans are so good at turning into mechanized genocides would end. No machine overlord would tolerate such crap.
Every politician would be out of work, since AI would manage resources equitably and hundreds of times more efficiently.
Police forces would cease to exist. It'd be trivial for AI to monitor everyone's movements and quickly find, isolate, and neutralize every human threat.
Every court case would be tried by a scrupulously fair arbiter. No more need for ambitous judges or juries of fools.
Everyone would be out of a job... on permanent vacation. Money would disappear. All our needs would be provided by automated production systems. For free.
No more hunger, crime, war, or injustice. And you think that's a bad thing? O Ye of little faith...
So will True AI actually lead to this utopia? Who the hell knows? But I do know that scifi literatue is a damned silly place to look for the answer.
Or do you get extra credit for "Lying about your past"?
Look at the 10 states that did best. Almost all are rural: Massachusetts, Montana, North Dakota, Utah, Vermont, Colorado, Minnesota, New Hampshire, South Dakota, Virginia.
Texas and Montana did better than NY or CA?
Time to ask, 'What and who were tested?' I suspect the sample was far from uniformly distributed across all US 8th graders.
> Until they do that, it'll still be just videotapes of college lectures.
's/just/free/g' That's a huge mod.
It's silly to compare the curent crop of on-line FREE courses with tuition-based in-class courses. Better to compare them with tuition-based on-line courses, like those of Columbia CVN or Stanford SEE. By that standard, the free courses I've seen are as good or better. And remember, each of the freebies costs $2500 less. That's a world of difference.
Will on-line courses ever approach the learning experience provided by in-class lectures or face-to-face study groups? Possibly. But given the GREATLY reduced cost, the ease of time shifting, the opportunity to learn from the very best profs (eventually), cut-rate e-learning isn't going away. In fact it's hard to imagine that it won't eventually outcompete the present college system, given the unreal cost difference.
Given the rise of Walmart and Amazon, you should never underestimate the power of delivering a product by combining 'sufficient', 'convenient', and 'dirt cheap'.
This article would be more credible if the author weren't Norm Matloff, a statistics prof who's been bemoaning the invasion of H1Bs into the software business for over a decade. Now that the demand for cheap labor has left the building, I guess he's turned to stomping sour grapes, "You shouldn't have wanted the job anyway".
What would happen if a bucket of popcorn reported the following nutritional data:
1200 Calories
1500 mg sodium
60 grams saturated fat (more than 2 Big Macs, from the coconut oil)
Do you think theater owners might object? Do you think parents might object?
THAT's the downside of posting the nutrients.
Popcorn Calorie Bomb
Hmm... Here's the missing link:
IEEE Mosquito Laser
Not so different from a May 2010 pioece in IEEE Spectrum where they built a laser targeting system for backyard mosquitos.
What we really need is the marriage of both products -- a laser canon that automatically tracks and vaporizes squirrels. And starlings. And grackles. And cowbirds...
If the controversy evolution is fair game, then Tennessee's students should question all kinds of controversies, like:
- Does God exist? Do literal interpretations of the Bible make sense?
- Are christianity and capitalism incompatble? Isn't love of money a sin? Was Jesus a socialist?
- Was Robert E Lee a traitor? So shouldn't the statues of him be torn down and spat on?
I dare you. Open Pandora's Box.
So if the camera fails to see your license plate you get no gas? Clever. I'm sure that another car or truck will never obstruct the camera's view, that snow will never obscure the plate, that fog will never blur the plate letters, that the plate will always be adequately illuminated, that the cameras will never break down, that the license database will always be up-to-date and on-line. No flies in THAT ointment, no sir.
All this fal-de-ral just to make sure that a few people pay their vehicle tax? Why not simply require everyone to pay their tax annually when they register their vehicle? Put a sticker on the windshield showing that the tax was paid, LIKE THEY DO EVERYWHERE ELSE.
Or if you must monitor everyone's tax status minute-by-minute, have everyone carry a tax-paid UPC fob that is scannable by a credit card swiper (or an attendant) when you pay for your gas? Would that cost, oh perhaps, a BILLION pounds less than buying and wiring up multiple spy cameras for every service station in the UK?
Who comes up with ideas this overcomplex, ineffective, and brain damaged? Newt Gingrich's british cousin?
Not John Wayne. It's Dean Wormer (played by John Vernon) in Animal House. He's admonishing Flounder (Stephen Furst).
http://www.imdb.com/title/tt0077975/quotes
What you see on TV has no basis in reality.
Actually, breaking into a strong safe exactly like breaking strong encryption. Both will require greater-than-average expenditures of time and money to crack, but given enough resources, both are possible.
If the state chooses not to expend the resources, that doesn't change the role of the defendant. It only makes it more likely that the prosecutors will whine to the judge about the unfairness of it all, and the judge will then throw the defendant's ass in jail until s/he capitulates and opens up the safe or removes the encryption.
The relevance of the 5th amendment to encryption seems to be, what evidence is necessary to justify the search of your computer? If 1) contraband or your acquisition of contraband has been observed, and 2) containers that *might* contain contraband are observed in your possession, then you're screwed. If 1 but not 2, this ruling says you're OK. If 2 but not 1, then you're OK as long as you wait out the imprisonment for contempt because you should eventually win your appeal of the search warrant since it lacked probable cause.
So the moral of the story is: make sure you hide your encrypted containers.
CNN has a nice article that puts "The Fair Labor Association" in proper perspective:
http://edition.cnn.com/2012/02/17/opinion/nova-apple-foxconn/index.html/
No. Large or academically motivated tech employers strongly prefer degrees from bigger name schools. There is no comparing the employment prospects of someone from MIT vs some low brow state school. Huge difference, fair or not.
Engineering master's are essential for advancement to technical project management. If the degree follows a few years of commercial experience, they're worth more on the floor than a PhD, since the latter targets research and not production.
An MBA who lacks engineering chops is useless toward leading a tech team to deliver a product that works and arrives on time. And useless for pretty much everything else too.
> 'I think it will actually be a long time, maybe never, when actual Stanford degrees would be given for fully online work...
Lots of fine universities already offer degrees entirely on-line (especially MS in engineering). Unlike the dean, I am certain Stanford will do this. But it'll cost... full tuition.
I doubt Thrun intends to offer a few courses and stop there. I think he'll offer an entire CS curriculum within maybe 3 years, and offer some soft of CS degree program soon thereafter.
It seems like you could offer other degrees using this same technology -- probably all engineerings, physics, probably math and statistics, maybe biology (but without labs).
Not only would the degrees be FREE (a huge thing for the poor in the third world and BRIC countries), but they'd be FAST. By excluding all the non-essentials, the equivalent of a BS in CS could be completed three times faster, in no more than 1.5 years.
Based on what I've seen from Thrun so far, I bet the degree will be widely respected, and frankly, better than 3/4 of today's CS degrees.
Universities beware. You're about to run smack into The Innovator''s Dilemma. And in my humble opinion, it's about damned time.