Machine Diversity: a Non-Fatal Way of Doing A.I.
As we know, the definition of intelligence in its simplest form is “the ability to acquire and apply knowledge or skills”. By definition, we need diversity in the content we consume in order to broaden our knowledge, make the right decisions, and then take action by applying those decisions. Intelligence involves the ability to modify state or action in response to varying situations and past experience. But what about Artificial Intelligence? Today’s A.I. is based on technology called Machine Learning. The philosophy behind Machine Learning is very “divisive” in the sense that it can tell a signal from noise, a cat from a dog, and positive sentiments from negative. In its essence, Machine Learning is a branch of statistics that deals with extracting and predicting patterns from large amounts of data acquired from sensors such as cameras, microphones, and keyboards. Therefore, today’s A.I. is lacking the ability to diversify. There is a great need to integrate diversity in A.I. and to shift away from building Artificial Intelligence based only on Machine Learning: we call this “Machine Diversity”.
One of the earliest problems Machine Learning tackled was clustering, which involved segregating different portions of the data into groups based on their commonalities. When applied to content, instead of making us more intelligent, A.I. is actually doing the exact opposite. It tends to separate what we like from what we don’t like, so for example, when we log in to our Netflix account, we are given automated prompts or suggestions based on shows and films we have watched before. Similarly, when we purchase something on Amazon, we are suggested to purchase products we will most likely be interested in based on what we have already purchased from the site. Our Facebook News Feeds are saturated with content shaped by what we have “liked” in the past. So A.I. today automatically chooses a specific category of content that we have been interested in at one particular time and keeps developing on that specific line of content because it is what we have enjoyed previously.
But why do we call it “Artificial Intelligence” if it’s not helping us to become more intelligent? We are constantly being fed with the same genre of content that we aren’t expanding our knowledge and are actually becoming less intelligent. It cements us into a specific moment in space and time, building only on that line of content. It is discriminatory and lacks diversity, skewing our knowledge by limiting the variety of content.
So what does this mean for marketers and businesses? It could lead to potential market saturations and market implosions, eventually resulting in a substantial decrease in sales and a difficulty in releasing new products. This is due to the fact that the customers’ interests are now extremely specific to just one narrow line of content with no room for any diversification. But technology can once again save the day - we need to start integrating “Machine Diversity”. This is a higher form of intelligence, which takes into consideration the importance of diversity and provides us with new content and new products we initially thought we wouldn’t be interested in. So it is important to start investing in the promotion of diversity in order to avoid the otherwise inevitable occurrence of market saturation and implosion.
More recent issues with Artificial Intelligence have been in the realm of discrimination, specifically in relation to gender, race, and age. Scientists at the Biogerontology Research Foundation developed Aging A.I., which utilized deep neural networks trained on the basic features from blood tests of reasonably healthy patients from Central and Eastern Europe. Many age-related diseases are causally, phenotypically and symptomatically related, meaning that very few of the patients were above the age of 60. During testing, it became clear that the deep learned biomarker was population-specific and that “the error rates in older age are higher because older patients are under-represented in the training sets.”
A.I. is also being used in the beauty industry and racial bias was found to be prominent there. Anastasia Georgievskaya and the company she co-founded, Youth Laboratories, found that machines can actually learn prejudice. The company used machine vision and A.I. to judge a beauty contest and almost all of the winners chosen were white. It would appear that A.I. systems have an inherent lack of diversification abilities, but machines can only learn from the data they are given.
As the CEO of LogoGrab, I believe that “there is no doubt that A.I. will have an important impact on the future of society, but we want this impact to be a positive one. At LogoGrab, we believe that Machine Diversity plays a significant role towards this vision.”
As we are moving towards the future of A.I., LogoGrab is working hard to make all of this a reality. We believe that Machine Diversity is a fundamental component for the future advancement of A.I. and we want everyone to freely prosper from this vision so that civilization as a whole can benefit from it. As LogoGrab is working on the development of Machine Diversity, it is our hope that the future of A.I. will produce a more diverse and inclusive array of content thanks to this ingenuity.