We need to take ‘black box’ out of AI to increase diversity, says panel

Автор: | 09.07.2018

People developing artificial intelligence (AI) technologies should consider the front end of these systems as well as the back end, says a panel.

At the Women of Silicon Roundabout event in London, it was pointed out that there is a lot of discussion around the importance of making sure AI algorithms aren’t biased, but teams should also consider the front end and who will be using such services.

By helping people to better understand what AI does and how embedded it is in our lives, it will also become easier to encourage a more diverse range of people into AI development roles.

“You shouldn’t look at AI in terms of just the algorithm and what goes on in the back end,” said Sybil Wong, COO of Sparrho.

Much like mobile technology, Wong said AI would eventually become part of use cases that developers often don’t think about, so not only should those making AI products consider the gender diversity of their teams, they should also take into account users from outside of their ecosystem.

“[We need] to make sure that when we are developing AI [technologies], the people at the cutting edge are aware of the biases that exist,” said Wong. “People in under-represented societies, and people who aren’t as technologically savvy, are being subjected to these systems, and to them it’s just a black box.”

According to Wong, experts are concerned that while gaining access to AI systems could be a gateway to equality, many only think of gender when taking into account segments of the population who are overlooked in AI development.

“That’s why women and people in other disciplines are so important in AI,” she said.

Developer diversity for diverse applications

A number of negative stereotypes about the technology industry still exist, including misconceptions about the types of people the industry is open to and what jobs in tech involve.

As well as building the algorithms that power AI, Wong said it was important to emphasise and take into account other roles that contribute to artificially intelligent systems, such as copywriters and designers, especially when trying to attract under-represented groups to the space.

“I’ve had women who have come in for marketing roles, because they thought that was the type of role they should be going for, who were then surprised when I told them they would be good for a product role,” said Wong. “They start to understand the types of roles available to them. A lot of roles are invisible to graduates.”

With different kinds of people working on AI-based applications, more use cases are considered, which is important because AI is an input-based learning system which grows based on what data is fed into it.

Since the younger generation has grown up using technology, Wong is hopeful about what new use cases the next generation will throw into the mix. “The hope is that the AI-native generation come up with completely new ideas. They can do whatever they want,” she said.

Fixing AI bias

But Deepa Mann-Kler, CEO of Neon, said there are already biases in these systems that need to be addressed before that next generation filters into the workforce.

“My deepest concern is the perp of bias – we live in a deeply flawed and biased society, and all we’ll do is replicate that [in AI],” she said.

Mann-Kler cited facial recognition technology as an example of this, claiming the error rate for the technology is significantly higher for women of colour than it is for white men.

“These facial recognition systems are being used by the military, by the police. [The technology] will become more adopted and become more mainstream,” she said.

As a starting point for making the AI industry more inclusive, Mann-Kler said more women should be encouraged into science, technology, engineering and maths (Stem) roles, and to help make this a reality she suggested those already in the industry should make themselves visible to “anybody passionate and interested coming in your wake”.

Since in the near future AI will be in people’s homes and personalised to them – including information such as their age, how much they sleep, what allergies they have, what music they like or when they’re at home – Mann-Kler said it was important for these systems to understand that individuals are “so much more than just male or female”.

Tech needs change of image

But it is difficult to encourage minorities into tech roles because of the stereotypes surrounding what counts as a technology job.

Avanade executive Laura Malcolm admitted there was a “branding challenge” in the industry, but said AI was an interesting topic that required a broad skillset, including creative skills, which could serve to make the area more attractive to minority groups and women.

Paired with the increased need for creativity in tech candidates as a result of automation, Malcolm said firms should focus on finding “early access” to talent through universities, scholarships or internships to expand the range of diversity and skills they are bringing into the industry.  

One way Avanade is trying to attract diverse candidates is by making sure job descriptions properly reflect the role. It is also making sure that different types of roles – the technical and the creative – work together.

“One of the big challenges when you enter this industry is that you’re still in the minority as a woman and as a creative,” said Malcolm. “My biggest hope is that tech and digital, or AI-like tech and digital, stops being this separate discipline and topic that it has been in the past.”

The fact that AI and other technologies are now “mainstream” will help to encourage more people into the space, according to Malcolm. She predicted that as tech becomes increasingly embedded in everyday life, it will also become a bigger part of the education system, being taught alongside maths or languages.

Tech’s expanding reach

But many believe the education system is in need of a shake-up to cater to the future, and Elaine Safier, chief commercial and operating officer at HSBC, said there was a “disjoint between education, where it’s at and this fourth industrial revolution that we’re in right now”.

Safier highlighted that in the future it was very likely that people would have several more jobs than they do now, and creativity would be a “key” skill, as well as flexibility.

By engaging children through talking about the technology they are using, such as applications like Pokémon Go or Snapchat, Safier said a more diverse group of children would become interested in tech careers.

In the future, Safier said we might not even call it tech, as it will be used widely in every industry.

“We need more diversity, not just women, in terms of the tech roles, and we need more types of roles looking at the diversity and bias piece when [developing] AI [technologies],” she said.

 

Source

Добавить комментарий