End of last year, Benedict Evans, working at the well-known venture capital fund Andreessen Horowitz located in Silicon Valley, talked about if AI makes strong tech companies even stronger. He argues that it is only partly true that data-rich companies will get even stronger using machine learning (ML). ML is facing much diffusion of capabilities – “there may be as much decentralization as centralization“.
Evans states that the starting point of the discussion is data. Even though machine learning might be already working with small data sets in the future, having much data is still the target to go for today. Saying that ML is working better the more data available, will there be a winner-takes-all effect?
The venture capitalist states that data used for ML must be very specific to the problem. Meaning that the application of ML will be widely spread to everyone owning specific data to a problem. Google, for example, will not have all the data, but all the data of Google. Thus, having the opportunity of getting better at being Google but not at anything else.
He then looks into the situation of building a company around ML to solve real-world problems. Where do they get their first data from and how much data do they need? Clearly, some data is unique to a business and insights will keep hidden for the ones not owning the data. Other data is useful across companies or even industries, which means a network effect will occur, the more data one has. But there will be the point no incremental data is needed any more since the product is already working.
Thus, Evans state that the diffusion of ML does not mean that Google will get stronger, but rather that the floor is open to everybody building new business ideas with the cutting edge much quicker than before.
He finally compares ML to Structured Query Language (SQL). ML is opening up new opportunities and will be all around. So not making use of it will let you fall behind. But, SQL was all around as well, but then disappeared. Evans states that the same will happen to ML.
Due to my opinion, this is a very interesting perspective on the future of ML. I agree with Evans that opportunities to make use of ML and getting better at what you are doing are open to everyone. Still, I expect industry leaders, specifically tech companies, to have an advantage over competitors facing their resources. This is supported by a study of Google (2017), claiming that the success of deep learning is dependent on “(a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labelled data”. Thus it is still a higher effort for small firms to compete against the tech leaders facing their resources. Specifically, the argument of a large data set is a very serious one in my opinion: Facing the high degree of competition, a small enhancement can result in an advantage, at least in the short term.
Evans, B. (2018). Does Ai make strong tech companies stronger? [online] Available at: https://www.ben-evans.com/benedictevans/2018/12/19/does-ai-make-strong-tech-companies-stronger
Sun, C., Shrivastava, A., Singh, S. and Gupta, A., (2017). Revisiting unreasonable effectiveness of data in deep learning era. In Proceedings of the IEEE international conference on computer vision, pp. 843-852. Available at: http://openaccess.thecvf.com/content_iccv_2017/html/Sun_Revisiting_Unreasonable_Effectiveness_ICCV_2017_paper.html
