AI, Big data, Automation,
|
Last November Microsoft and the University of Cambridge published a paper about how its AI was writing code. The process includes scanning bits of code and considering lots of possibilities for how bits of code could fit together to create the right solution. This was falsely reported as Microsoft AI plagiarizing existing programs. That says something interesting about us and about what learning really is.
We all accept that technological change destroys some jobs and creates new ones. We remain excited about the process because, in general, the newly created jobs tend to be good ones and the innovation lead to positive things for society. As bad as factories were, the small-scale farming people left behind had its issues too. The personal computer revolution created entirely new categories of white collar jobs in the tens of millions. Because of advances in tools and training, it became easier to become a professional computer programmer even without a college degree in computer science. My grandfather turned wrenches, my father was the first in our family to work in an office and I work with computers. One worry about the AI revolution is what if this one doesn’t create new jobs like the past revolutions did? When you conduct media interviews you get a peak into the process of creating stories. I have seen how journalists find new ideas for new stories in academic and other original sources of news and report on them in industry press. If they are interesting enough then reporters in mainstream press will pick it up. It is often a lot of steps from medical journal to the table in your dentist’s waiting room. It is a complicated process whereby journalists learn about a topic and then simplify it and make it accessible and interesting to their audience. Some reporters do it very well but others get parts wrong. Much like a game of telegraph (or telephone?) reporters will find story ideas from other reporters, often without reading the original source materials. When a reporter misinterprets some coverage and drives an engaging conclusion that can be picked up by another journalist in another story, and so on and so on. The process of how stories are generated reflects a lot about how humans learn and how original works are created. Even the most novel innovation contains some derivative parts. Solving new problems means building upon what is done before. This is natural and the line between this and plagiarism boils down to now much originality has been added to what already existed. One of the current fears of AI is that it will replace us in our jobs. One intellectual life preserver for many people in tech is that programming is a creative job that reflects how differently we think than AI ever possibly could. This argument boils down to the view of AI as a tool for automation and not one for actual creation. So if AI steals that means it isn’t creating and programming jobs along with whole categories of jobs are safe. The particularly scary scenario is that AI creates itself and destroys some jobs but does not create new types of jobs. But, AI is already starting to create AI such as this experiment at Google Brain. It is also predicting cancer better than doctors, fighting parking tickets, creating art and stories and yes, writing code. So one of the first things we need to do to prepare for the full impact of AI is to rethink how we think about creativity and learning.
1 Comment
MarkHNYC
7/19/2017 12:32:35 am
Interesting article. Comparisons between AI learning and human learning are useful to help us develop better machine learning techniques
Reply
Leave a Reply. |
AuthorMichael Zammuto is the CEO of Completed.com and Cloud Commerce and a strategic adviser to several startups. Mike's background is in SaaS services, B2C sites and B2B firms and has worked extensively in online reputation, digital marketing and branding. Archives
October 2019
Categories |