How to keep human bias out of AI | Kriti Sharma

TL;DR

In this thought-provoking talk, the speaker explores the biases and dangers of artificial intelligence, emphasizing the need for diversity and ethical considerations in its development.

Transcript

Translator: Ivana Korom Reviewer: Joanna Pietrulewicz How many decisions have been made about you today, or this week or this year, by artificial intelligence? I build AI for a living so, full disclosure, I'm kind of a nerd. And because I'm kind of a nerd, wherever some new news story comes out about artificial intelligence stealing all our jobs, o... Read More

Key Insights

• 🤖 Algorithms are being used to make decisions about our identity, such as what products we want to buy or which shows we want to watch, but these decisions are often based on biases related to gender, race, or background.
• 💼 AI has the potential to discriminate in hiring processes by learning biases from human hiring managers, resulting in the screening out of certain candidates. This gender discrimination is not acceptable, whether it is done by a human or a machine.
• 👥 It is important to have diverse teams building AI technology in order to avoid bias and to have different perspectives. This diversity includes people who can write and tell stories, solve problems, and face different challenges.
• 📱 The design of AI voice assistants, such as Siri and Alexa, often reinforces stereotypes by having predominantly female voices associated with obedient roles, while high-powered roles are often associated with male voices.
• 🔽 There is a need to address biases in AI technology, by being aware of our own biases and those in the machines around us, by having diverse teams building the technology, and by giving AI diverse experiences to learn from.
• 💡The possibilities for AI are vast and can be used for positive impact, such as providing remote healthcare services or assisting victims of domestic violence in South Africa.
• 💻 Anyone, regardless of their background or appearance, can contribute to the development of AI technology. It is important to encourage diversity in the field and ensure that AI is accessible and beneficial for all.
• ⚠️ We should be aware of bias in technology and take action to address it, as well as advocate for inclusive AI technology that considers and benefits everyone.

Questions & Answers

Q: How is artificial intelligence (AI) being used to make decisions about us?

AI is being used to make decisions about what products we want to buy, which shows we want to watch, whether or not we get a job interview, our car insurance rates, our credit scores, and even our annual performance reviews. However, these decisions are often filtered through AI's assumptions about our identity, including our gender, race, and age.

Q: What are some examples of real decisions made by AI that reflect biases?

AI has recently made decisions such as: a black or Latino person being less likely to pay off a loan on time compared to a white person, a person called John being a better programmer than a person called Mary, and a black man being more likely to be a repeat offender than a white man. These biases stem from the biases that AI has learned from humans.

Q: How does AI reinforce our own biases in decision-making?

AI can reinforce our biases by learning from the data it has been fed. For example, if a hiring manager has mainly hired men for programming positions, AI may learn that men are more likely to be programmers than women. This can lead to AI screening out female candidates and perpetuating gender discrimination.

Q: How can we address the biases in AI decision-making?

There are three steps we can take to address biases in AI decision-making. First, we need to be aware of our own biases and the biases in the machines around us. Second, we need to ensure that diverse teams are involved in building AI technology. Lastly, we should give AI diverse experiences to learn from, so it doesn't just replicate our past mistakes.

Q: What is the potential of AI to make the world a better place?

AI has the potential to greatly enhance our world. For example, it could provide remote diagnosis for pregnant women in rural areas with limited access to healthcare, or offer assistance and support for women facing domestic violence. By building AI with the right values and ethics, and involving people from diverse backgrounds, we can create a more equal and beneficial future with AI technology.

Summary & Key Takeaways

• AI is making decisions about us based on our gender, race, and background, leading to biased outcomes.

• We need to be aware of our own biases and ensure diverse teams are building AI technology.

• AI has the potential to make the world a more equal place, but we must teach machines the right values and ethics to avoid repeating past mistakes.