The artificial intelligence market is expected to grow at a compound annual growth rate of 36.1 percent by 2024 and will reach a value of $ 3,061.35 billion on the same period. We can only expect this forecast to grow as more big name companies such as Microsoft, Google, and IBM are heavily investing in the space as they aim to develop solutions that will benefit mankind.
All of this growth begs the question about whether everything about AI good or should Stephen Hawking’s advice be heeded in that we must take a careful approach with it?
AI has shown biases
Sure, there have been concerns about how AI will be replacing certain jobs that belong to people, but that’s not the only problem people should worry about as these systems are already showing bias in interesting ways.
For example, The Next Web writer Bryan Clark discovered how a simple Google search can show how biased an intelligent system is. By searching for ‘women’s professional hairstyles’, Google Search delivered photos of white female sporting various types of hairstyles from short or long haircuts, braided hairs, in a bun or a ponytail, curled, and even one with cornrows on it. Searching for ‘unprofessional women’s hairstyles’ delivered images of black women sporting hairstyle almost the same as the ones sported by the white females. Looking at the ‘unprofessional hairstyles,’ you could easily wonder ‘What’s wrong with these hairstyles?’ In most cases, the answer is nothing at all.
In many of these results, the subjects have gorgeous hair that are in no way ‘unprofessional’ looking. The problem, or at least the one consistent factor, is the color of their skin.
Flaws that are surprisingly, human
So should Google be to blame for this seemingly racist search result? Not entirely. You see, much like any other AI system, Google’s Search is trained on people’s perception and queries. So even if the slighted subjects in that test are all professional looking, if the people that set that intelligence in motion don’t see it that way, then the system will also see them as unprofessional. That is a case of replicated bias.
Definitely not a one-off, this is not the first case AI has unknowingly replicated bias. If you recall Microsoft’s AI chatbot Tay, the Twitterverse was able to turn it into a racist chatbot in just a couple of days. Can you imagine how ruthless it would be now if Microsoft decided not to take it down?
AI is a great development in the human-computer experience, but it’s not a stretch to imagine some pretty ugly scenarios. Imagine the day when an AI system will be a determining level in the application process of getting a job or a loan. What if your name, the color of your skin, or some opinion that is totally biased gives you a declined status? Will you still be taking things lightly or will it be too late by then to take any actions?
Artificial intelligence marks the first time in history that computing ceases to be a simple (yet powerful) tool. AI makes things autonomous, free of its passive past, and the potential arbiter between events that affect people. I am personally not a Hawking fan (or fanboy), but his warning is valid. Governance, ethics, and awareness must be a formal part of the developing AI story.
Image source: Pixabay