Second Draft (Aug 2019),

You may have heard about the competition of big companies such as Amazon, Microsoft, and Google on providing Cloud services. In this regard, A natural question is how can we make smaller businesses that can benefit from monopoly of bigger companies? This is not a new question neither do we provide a new answer to this question. We would explain one existing answer with the terminologies used in the field of artificial intelligence. Through the rest of the post, we talk about the Search Engine providers. Google, Bing, Baidu are big providers of this service for us. How a business can be built to earn profit from the monopoly of these providers?

I explain through an example in the field that I've been doing research in. You may search the term "Recurrent neural network paper" in the mentioned search engines (Google.com, Bing.com, Baidu.com) or any other big search engine that you would prefer. Recurrent neural network is a term for a class of models used in artificial intelligence to model things that happen sequentially such as words in a sentence. Just open the first article in the form of a publication you could find through each of the search results from different websites. If you are familiar with the terms and concepts researchers use in artificial intelligence, then you may find the material and the problem that these papers study totally different (in my own try, e.g., I saw a paper with completely theoretical material, and another one with a focus on the applications). You may be like me in this case: without even thinking, I typically assume that Google is the best search engine ever and search whatever I am looking for in Google.

First of all, if Google was the best search engine ever, there would have not been other big search engines, specifically as big as Baidu and Bing. This is not the actual point we would make here. Let's say Google has a very strong ranking system that provide desirable search results. The point we would make is about the "bias" of these results, i.e. results may be biased toward a specific concept that is intentionally proposed by human or even exist unintentionally (note that these are serious problems and a wide research area in the artificial intelligence and machine learning is specified to it). Assuming Google employees prefer male gender over female (this is just an imaginary example to get a sense about how can human bias be; I am not a Google employee) and they may impose in results a preference over papers written by male gender. This specific bias may not directly seem to be worrying. However, theoretical versus practical preference may be worrying in biasing the people toward specific type of thinking (theoretical thinking, or practical thinking). Another type of bias is the one that nobody knows the existence of it. Let's say Google is using some artificial intelligence models in its search engine that till-now nobody knows what does it exactly do (deep learning models). Again, to provide an imaginary example, assume that these models give a better chance to the publications written in the Europe institutes to be appear earlier in search results, and nobody is aware of the existence of this bias. Even though these biases may not be seen as serious issues, they may have serious negative effects in the future. I would like to mention that in particular, the second kind of bias that we explained in longer term may cause humans to more seriously follow artificial intelligence commands, not artificial intelligence to follow human commands!

One way to diminish these biases is to get the results of our search through different search engines. So, there can be businesses with a focus of diminishing these biases. This indeed has an important value . Not surprisingly, there exist e.g. websites that provide the aggregated result of your search in different search engines (also called metasearch engines). To me, this is analogous to a case that we consult with different people when we are facing with an issue. Consulting with more than one person often results a better decision for us. My personal opinion is that these businesses are not so far taken as seriously as they should be.

One more thing, that the companies that have monopoly in providing search engines won't necessarily gather together to diminish this bias. Thereby, these type of businesses can actually grow in this environment! The same idea may be applied in different monopolistic competitions to build similar businesses. As long as that competition exists among the big players, your business will be alive. Your business will probably be dead in case that they decide to be friends.

Thank you for reading this article.