What Should We Expect From AI In 2018?

What Should We Expect From AI In 2018?

When computers were introduced, there was a vital acronym that was used to describe how we interacted with machines: GIGO-garbage in, garbage out. Artificial brains are moral and clever just like the people who program them.

Banks Should Keep Up With FinTech
Personal Assistants Can Now Pay Your Bills
Strata Data Conference To Be Hosted in NY, September 26

Artificial intelligence systems are expected to make huge advances in 2018, but we must make sure that they learn to avoid human bias.

When computers were introduced, there was a vital acronym that was used to describe how we interacted with machines: GIGO-garbage in, garbage out. Artificial brains are moral and clever just like the people who program them.

That dynamism will be more important in 2018, as algorithms continue to change everything from commerce to healthcare. Already, automatic programs have put prices on various websites and determined the new sources we watch on Facebook. Algorithms are used to help people interact with bureaucratic systems using a scoring system. Most of the process of voting, college admission and policing is done algorithmically.

But it is only a few people who understand how algorithms work. (Think of Google’s powerful PageRank system how many people know how it operates?). And sometimes, there are several advanced artificial intelligence systems that most people do not understand at all. Their creators know that they work, somehow.

In 2018, AI will make huge advances because of the machine learning feature that will be introduced soon. But AI should be approached with bias and shortsightedness that affect humans.

What will that feature look like? Take Tay, for instance, a chatbot that was launched by Microsoft in 2016 to help customers answer tweets. However, it lasted for only 16 hours before it was closed down.

There is a huge problem with AI as attributed by Tay’s descent. Joy Boulamwimi, the founder of Algorithmic Justice League, developed facial-recognition software that had a problem even to identify her. In fact, it recognized her when she wore a white mask than when she showed her normal face. She concluded that most training sets for facial-recognition software used were light-skinned.

In a TED talk, she noted that most US law-enforcement agencies use such software to identify criminals. She admitted that labeling faces and facial recognition remains a major challenge. She says that they always make fun of mislabeled people on their Facebook photos. If algorithms are to be used to decide on parole conditions or loan applications, then we must be sensitive not to include societal biases in them.

In 2018, algorithms will simplify the way we live, but we will look at their disadvantages as well. In particular sets of data are required to represent and not be Silicon Valley skewed demographics. Our algorithms need to learn and become more efficient than us. Otherwise, we will be required to make use of an old acronym BIBO-which means Bias in, Bias out.

2018 will be a great year for exploring artificial intelligence systems. But they must strive to learn to remove the human bias element in them. Algorithms will continue to be part and parcel of our daily lives. But we need to improve them to remove inefficiencies that may be caused by their creators.

COMMENTS

WORDPRESS: 0