NEW: RESOURCE PAGE

Searching for tutorials and software about Deep Learning and Neural Nets? Be sure to look at my Resource Page!
Looking for Octave? Go to my Easy Octave on Mac page!

Friday, February 27, 2015

The AI application dilemma: Buy it or Build it?


Artificial Intelligence is supposed to be "smart", its elaboration is a very specific competence - when a firm is going to use AI, should they code, hire or outsource? It's a complex question, but I think a clear and useful answer can be given; and there is also a warning at the very end of this article.
  • If you have a core competence, eg. loans in a consumer bank, and you use AI to automate it, you need to keep full control of a system that embodies your expertise, do a custom build and make sure some of the participants in the build process remain in house, although using an outsourced AI engine makes sense.

 A good example is automating loan applications: the bank has the experts on loans, the system needs to be well understood by employees, it may need tuning and fixing over time, and the best way to do ensure initial system acceptance, updating and user training is to make sure that the loan experts who tune the loan system are in house and available. Also, in a competitive industry, it is likely that some functionalities of the software will be proprietary, and a drop in solution will not be sufficient. However the software itself can be running on an external system and using a ready-made engine eg. Watson or Azure, and it can —and probably should— be implemented with the help of external AI consultants.

  •  If you are looking to do an add-on module or a user interface, eg. simple voice recognition, outsourcing a generic module is probably the best way to go. In fact, the more standard the module the better, because then it will be easy to get the technology updated as the state of the art evolves. 

An example of an add-on situation might be a company which wishes to add voice synthesis and simple voice recognition features to its products eg. to ATMs. It makes sense to buy the speech AI technology from a third party, who will curate and update this module over time; and in fact it might make sense to have several providers eg. for different languages or geographic regions. 

Last, not least a warning: AI is a specialty, like brain-surgery, and it is often viewed as one of the hardest specialties in computer science. So it is usual to call in or hire a specialist for an AI job, just like a hospital might call in a specialist surgeon. If you are going to use AI, make sure your team will have outside help for the whole duration of the project, without trained specialists on call an AI project will probably fail

Finally, here is an old book on this ageless topic. Artificial Intelligence Applications for Business

Edmund



No comments:

Post a Comment

Hey, let me know what you think of my blog, and what material I should add!