4 ms·

Buy/get a dataset, such as images or audio or biological data, and spend the money on computing time to do machine learning/genetic programming to create a clas

Buy/get a dataset, such as images or audio or biological data, and spend the money on computing time to do machine learning/genetic programming to create a classifier from it. Sell/rent out the classifier for big bucks. For example, label smiling/unsmiling faces on 10,000 images using mechanical turk, run an algorithm (neural net,svm,genetic programming) on it using your own computer cluster or amazon ec2. Once you have the classifier, allow people to use for a small payment each time. There are almost infinite number things to do in this way, some examples accent identification in audio ; telling the gender/race of person from photo ; labeling person attractive/not attractive

Tichy 17y ago
I'd be really interested in working in machine learning, so if you know an easy way to make money from it, I would be interested to learn more. You think such a classifier would be an easy sell?
- marshallp 17y ago
  It's use is already sold (rather expensively) in inspection/robot systems in factories and in automatic license plate recognition. I imagine it would be easy to sell. Create a classifier that rates a user's attractiveness and many dating sites or individial users might interested in paying for it. Call centers might be interested in paying for accent identification, they can route a call based on accent to a employee with similar accent. Microsoft will release project natal later this year and probably make billions from it. As far i can see the opportunity is absolutely huge (trillions and trillions of dollars) since it is so commercially useful (unlike the web industry which has been based around selling eyeballs).
Aron 17y ago
Do you have a link(s) to someone who appears to be doing this? I'd like to see that this idea has been implemented, and how.
- marshallp 17y ago
  This area is generally referred to as machine learning and is used extensively. Many inspection systems and vision guided robots in factories use machine learning. The netflix prize was one famous example of this. The darpa grand challenge was another famous example of this. Machine learning is used in many different places hidden from view. Probably your car automatic braking system has a machine learned controller in it. Search engines, recommeder systems. Star detection in astronomical observatories. Cheque reading in atm's and mail sorting systems. This web page about computer vision companies http://people.cs.ubc.ca/~lowe/vision.html http://people.cs.ubc.ca/~lowe/vision.html Many automatic license recognition, speeding detection cameras might have machine learned algorithms Google voice is based on machine learning. Face detector in many different places (e.g. picasa) uses haar features and boosted decision trees. However, because it is so commonly implemented and was first made in research community (at microsoft research) no one charges for it. Field Guide to Genetic Programming (google it) contains many examples of gp application. Type in machine learning applications, or "something" classifier, e.g. face classifier, accent classifier, on google and in many cases somebody has already done it as a student project.

Tichy 17y ago

I'd be really interested in working in machine learning, so if you know an easy way to make money from it, I would be interested to learn more. You think such a classifier would be an easy sell?

marshallp 17y ago
It's use is already sold (rather expensively) in inspection/robot systems in factories and in automatic license plate recognition. I imagine it would be easy to sell. Create a classifier that rates a user's attractiveness and many dating sites or individial users might interested in paying for it. Call centers might be interested in paying for accent identification, they can route a call based on accent to a employee with similar accent. Microsoft will release project natal later this year and probably make billions from it. As far i can see the opportunity is absolutely huge (trillions and trillions of dollars) since it is so commercially useful (unlike the web industry which has been based around selling eyeballs).

Aron 17y ago

Do you have a link(s) to someone who appears to be doing this? I'd like to see that this idea has been implemented, and how.

marshallp 17y ago
This area is generally referred to as machine learning and is used extensively. Many inspection systems and vision guided robots in factories use machine learning. The netflix prize was one famous example of this. The darpa grand challenge was another famous example of this. Machine learning is used in many different places hidden from view. Probably your car automatic braking system has a machine learned controller in it. Search engines, recommeder systems. Star detection in astronomical observatories. Cheque reading in atm's and mail sorting systems. This web page about computer vision companies http://people.cs.ubc.ca/~lowe/vision.html http://people.cs.ubc.ca/~lowe/vision.html Many automatic license recognition, speeding detection cameras might have machine learned algorithms Google voice is based on machine learning. Face detector in many different places (e.g. picasa) uses haar features and boosted decision trees. However, because it is so commonly implemented and was first made in research community (at microsoft research) no one charges for it. Field Guide to Genetic Programming (google it) contains many examples of gp application. Type in machine learning applications, or "something" classifier, e.g. face classifier, accent classifier, on google and in many cases somebody has already done it as a student project.