The human side of AI: It's also wrong

At some point we've all made mistakes, and although it can sometimes be frustrating, they're an important part of learning and development. Mistakes provide us with opportunities to adjust behavior and improve capacity to face challenges in the future.

The same thing happens with Artificial Intelligence: AI systems They train and learn from large amounts of data and despite their ability to process and learn from this data at incredible speeds, they are not infallible.

Here are some notable errors that these technologies have presented in their first applications in some sectors:

1. RUINS THE SOCCER BROADCAST.

The Scottish soccer team Inverness Caledonian Thistle FC announced that it would dispense with human camera operators to broadcast matches and would instead use autonomous cameras thanks to AI.

The first match in which it premiered happened what no one expected; the smart cameras were messed up when, in addition to the ball, a linesman appeared with a bald head.

The AI camera confused the referee's bald head with the ball for much of the game, causing the shot to continuously follow the man instead of broadcasting the real game.

‍

2. RACIST CHATBOT.

Microsoft developed Tay, an AI-based chatbot that fed on conversations of people between 18-24 years old on Twitter.

What happened? Some users decided to use methods to get the bot to incorporate some inappropriate words and phrases into its vocabulary, causing it to respond with racist and even Nazi comments.

Finally, Microsoft had to pause Tay completely. “The more you talk to him, the smarter Tay gets. The problem is that Tay, although he implements filters for bad words, was not ready for something very important: the darker, “troll” side of the Internet.”

‍

3. DAY OF SEXIST SELECTION.

Amazon decided to put much of the weight of its hiring processes in the hands of AI. However, this AI was found to systematically discriminate against women.

The reason for this discrimination is that the AI was trained with the profiles of former candidates who had tried to work in the company for the last 10 years, who showed male dominance within the sector.

Amazon's artificial intelligence thus learned that men were better candidates than women and penalized resumes containing words like “woman”. When engineers became aware of this trend, they reconfigured the logic of artificial intelligence to disregard those terms.

Undoubtedly, our world is immersed in a constant process of technological evolution, and the interaction between human beings and Artificial Intelligence marks a crucial milestone on this path. Both AI systems and people have their limitations, so it is our responsibility to detect possible imperfections in the algorithms.

Living with Artificial Intelligence provides us with opportunities to grow, adapt and evolve with it. The more we teach it and nourish it with knowledge, the more perfect it will be.

‍