My daughter is about to get her driver's license. Her children won't need one as their cars will drive themselves. One issue I have as SIGACT chair is gathering information together from various old proceedings and newsletters. In a few years computers will do this for me as quickly as I can ask for them. Right now we have great tools for finding information on the web. In the future our tools will make sense of that information. In the near future there will be no such thing as unstructured data. Where will all this lead us? I wish I knew--I could be a wealthy man.
Machine Learning has become a very mathematical and statistical-based research area yet the theoretical computer science community hasn't played the role in this area that we could have.
The New York Times have been running a series of articles about AI and its implications to society. A recent article talked about how legal firms save considerable money by using computer software for document discovery. Less money means less white-collar employees needed to sift through documents, a point Krugman pointed out in his column. Krugman also says
Conversely, jobs that can’t be carried out by following explicit rules — a category that includes many kinds of manual labor, from truck drivers to janitors — will tend to grow even in the face of technological progress.I don't expect truck drivers will exist in 10-20 years either. Technology has so far tended to create more jobs than it destroys but will there be any safe jobs in the future?
Why do you think cars will drive themselves, and why do you think that soon?ReplyDelete
Prior to AI cars, lets have driverless trains and metro rails.ReplyDelete
There are already many driverless trains and subway lines: see here:Delete
Assertion: Technology has so far tended to create more jobs than it destroysReplyDelete
Having spent some time living with my son in one of the earth's few remaining paleolithic societies, I can testify that this assertion rests upon dubious foundations.
As it turns out, even paleolithic societies are remarkably technological and skill-intensive ... just try to sail one of these canoes, for example.
Unfortunately, Theoretical Computer Science is a very unfriendly area for Machine Learning.ReplyDelete
"In the near future there will be no such thing as unstructured data."ReplyDelete
You really think so? Then consider Goedel incompleteness and randomness in Kolmogorov theory
The technology for automated cars is here and testing has already begun. Of course, it could be another decade or so before automated cars will be allowed for the public, mostly due to safety and legal issues. If you had a racetrack and an automated car, you could as well ride it this very instant.ReplyDelete
To adamo: In this sense, the future is hereReplyDelete
Keep your sensors on the road and your actuators upon the wheelReplyDelete
Keep your sensors on the road and your actuators upon the wheel
Yeah. we're goin' to the Singularity
We're gonna have a real
LOL ... yeah ... for sure ... Country & Western trucker songs are gonna need a rewrite ...ReplyDelete
"Chorus: We got a big 'ol network, travelin' down the road."
Q: "What do you hear when you play a trucker song backward?"
A: "Your operating system rebooted, your cron backup worked, your system update succeeded, you remembered your master encryption password ..."
"Machine Learning has become a very mathematical and statistical-based research area yet the theoretical computer science community hasn't played the role in this area that we could have."ReplyDelete
I don't understand this statement. Researchers in machine learning who prove theorems are working in theoretical computer science, aren't they? Do you mean, "The subset of theoretical computer scientists that I am in does not work in ML"?
We were saying stuff like this 20 years ago and even crazier stuff 50 years ago.ReplyDelete
Our children's children will still be driving on four wheels and should still be using a pencil and paper to learn math.
It's fun to think about, but the reality is that either things aren't moving as fast as we think they are or that we're too much of a control freak to stop driving.
The vast majority of the researchers in machine learning who prove theorems are NOT working in theoretical computer science.ReplyDelete
Bon Crowder asserts: We were saying stuff like this 20 years ago and even crazier stuff 50 years ago.ReplyDelete
Speaking as roadmap aficionado, I can testify that the 1950s were witness to some outstandingly successful STEM roadmaps ... these roadmaps are well-worthy of in-depth study.
It has often occurred to me that the success of 1950s roadmaps may reflect, in part, to the military training that so many of that era's scientists and engineers received. As Eisenhower famously said: "Plans are useless, but planning is essential" ... the same is broadly true of STEM roadmapping.
Highly recommended are (and available on-line) the Time Magazine cover story for Monday, Apr. 29, 1957, titled "The New Age". There's a terrific book by Neil Sheehan too, titled A Hot Peace in a Cold War, that vividly describes the mathematical foundations that John von Neumann (among others) provided for (what amounted to) a Cambrian explosion of systems engineering.
With respect to TCS, I have posted on Gödel's Lost Letter and P=NP about Jules Harmanis' 1981 roadmap for TCS. To my mind, Harmanis ranks high among authors whose seminal works provide foundations for a similar Cambrian explosion of quantum systems engineering (as broadly associated to regenerative medicine) in the 21st century.
I hope so, anyway. :)
the first autonomous vehicle I saw was at CMU, over 20 years ago, and it drove on a city streetReplyDelete
Come to think of it, the Isaac Asimov story Sally (1953) vividly describes the union of strong AI with autonomous vehicles, as does the Cordwainer Smith story Mark Elf (1957).ReplyDelete
Both stories have a gratifying informatic depth to them.
Self-driving cars: That's a terrible idea. Makes public transport even less attractive. Who is going to save the planet?ReplyDelete
My opinion: No, no job will be safe. And though I think it's very difficult to make predictions about the next 20-50 years, it does look like autonomous cars will be here soon. People often dismiss such claims because we have a long history of our imagination outpacing technological progress, but suddenly we have computers that can pass the CA state driving exam, recognize human faces more reliably than humans, and the degree to which they "understand" us has grown rapidly.ReplyDelete
The question I get hung up on is: how do we transition from our current social-economic models to one in which all of the work is done by machines? I don't know when such machines will exist, but I'd be surprised if it takes more than 100 years. And I see no way to avoid this--capitalism always has and always will drive technology to replace humans.
Lately I've been geeking-out about the future of brain-computer interactions, mostly due to this TED talk presenting a 'cheap' EEG headset that can read your facial expressions and emotional state out of the box.
Also, to the last Anonymous comment, concerning autonomous cars and their impact on the environment: it could be hugely beneficial. First, cars could be shared rather than owned, and more efficiently utilized, reducing the number of cars needed and more importantly, parking spaces. (It is estimated there are ~800 million parking spaces in the US, taking up valuable resources in land usage, paving, maintaning, etc.) But also autonomous cars could potentially be more reliable and safer, allowing safety systems to be lighter which in turn increases fuel efficiency. Additionally, reduction in the number of accidents would save money, lives, infrastructure/manufacturing, and so on.
The real challenge again is, how does a society make such a transition?
Also, Google is currently testing autonomous vehicles in California, and they have concerns over the legal implications, but currently think they are covered by existing legislation concerning cruise control.ReplyDelete