Artificial intelligence, or AI, is one of the greatest technological prodigies of our era. That is undeniable.
Still, proof that AI is gender biased keeps piling up: online translators, virtual assistants, emojis and almost every ‘intelligent’ piece of technology is glitching against women. Let’s see a few examples.
A couple months ago, Amazon had to manually correct a recruiting algorithm that was down-ranking women. The iPhone suggests a male emoji when you type ‘CEO’. When translating a sentence in a gender neutral language, Google Translate will always consider a doctor a he and a nurse a she. Bing’s translator takes masculine nouns and assigns them masculine properties:
‘When Bing translates “the table is soft” into German, it offers the feminine die Tabelle, which refers to a table of figures.’ (Kate Mccurdy, computational linguist).’
Virtual assistants take the issue to the next level. Siri, Alexa and Cortana are our modern Rosie the Robot Maid. The Western-world woman is absolutely stereotyped as the homemaker, the caregiver or the attendant. Alexa used to take insults with a graceful ‘thanks for the feedback’, but not anymore: Amazon heard the users’ complains, and now Alexa will retort ‘I’m not going to respond to that’. She also claims to be a feminist.
Amazon, Google and other virtual assistants manufacturers explain that there are reasons for their gadgets not to be ‘male’: female voices are more calming and children respond better to them. Hey, no one wants another HAL 9000! They refuse that the devices are embodying the stereotype that connects female figure to service-oriented job.
Lisa Simpson wants her female plushie to be a professional – here’s what she gets.
Now wait – is this whole mess ‘accidental’? What are the sources of bias?
John McCarthy, the father of AI, described the field as ‘the science and engineering of making intelligent machines”. It makes it possible for machines to learn from experience, adjust to new inputs, even perform human-like tasks. An artificial intelligence learns my mechanisms such as machine learning, deep learning, Word embedding and natural language processing. Machines are fed information in large amounts, process the data and start to recognize patterns.
Deep learning is the particular branch of machine learning that teaches computers to recognize speech, make predictions or identify images. These are what we call human-like tasks, and the main principles at work in virtual assistans. This branch is the one that has been showing alarming evidence of gender bias.
The information used to train Amazon’s recuit AI consisted ten years of resumes submitted to the company…all from males. Google searches the web and counts the time the word nurse appears next to he versus she…and we know how that story goes. The algorithm learns gender stereotypes because it is fed partial, preconcieved information. Virtual assistants get female voices because women are overrepresented in the service sector, a market section on the rise that men still avoid because ‘it’s women’s work’.
The vast majority of the data AI feeds on is biased, not only in gender but also in race.
There is something very curious about classic sci-fi and retrofuturism. Back in the 1950’s and 60’s, this is how television and books would portray the future: flying cars, machines that produce food, servant robots. Thing is, the cars were driven by men, the machines were operated by women and the robo-servants were called Rosie. They imagined the future merely in terms of technological evolution, not social progress. I would have never expected a transgender couple as main characters in The Jetsons, but even the most progressive and imaginative pieces of pop culture could not conceive something as simple as a woman boss.
Technology reproduces digitally what human beings are analogically, because technology is man-made – and yes, I chose that term on purpose. When top tech companies are mostly directed by white men, how can we expect things to be different? The presence of women in the industry is growing, but not in the decision-making positions.
Technology should be a paradigm shift, not a mere recreation of the status quo. Our societies have developed towards (if not fair) less unfair scenarios, but when technology systematically raises this kind of errors…we start to worry again. Have we really come that far? What is the true state of gender equality? Artificial intelligence is confronting us with the flaws of our very own human nature, and that makes it less artificial. Still, we cannot allow inequality to persist in any field. Diversity is a political decision, and companies have to decide: how are they going to shape their algorithms? What kind of information is to train and reinforce them?
AI has the potential to be revolutionary, but just the potential. As of today, it is as conservative as The Jetsons.