I’ve grown increasingly interested in artificial intelligence and have been trying to increase my understanding of the topic. Like I mentioned in my last entry, it falls pretty far outside the realm of traditional computer science. It is a huge field which is moving quickly, so I can understand why most practitioners tend to have specialized Ph D’s in the field.
I stumbled on this talk from 2017 by the AI researcher, Andrew Ng. At the time we was the Chief Scientist at the Chinese search engine Baidu. It is the best layman’s discussion I’ve heard on topic, and there are a couple key take aways.
- Modern AI techniques require a ton of data for training models. One example he mentioned was using 10 years of audio including transcripts to train a speech recognition AI
- Companies like Baidu (and certainly Facebook, Google, Microsoft, and Amazon) create applications for the explicit purpose of collecting data and not for the value in the product in and of itself
- Data is considered a strategic asset for these companies
- Smaller companies will have a difficult time competing due to lack of data (a probably also a lack of computing power)
- The most profitable application of AI to date is ad targeting
For many years AI seemed like a fringe area of research in computer science. As Ng says, the field went through many winters, but he now believes we have entered an “eternal spring.” From my perspective it seems like there were a couple breakthroughs about 5 years ago, and then the field exploded. When discussing the negative consequences of AI he mentioned multiple times the loss of jobs, and even mentioned radiology as a field that may no longer exist.
I have been an advocate of the internet and web technology throughout my career, but I’m worried that in light of issues like the Cambridge Analytica scandal, we might be creating a dystopian future controlled by the few who control the data.