SciTech Now: Episode 615

TRANSCRIPT

♪♪

Coming up, the built-in bias in artificial intelligence technology...

We don't necessarily think about how they're programmed and what those programs are sort of reflecting back to us.

...bringing back a vital resource...

When the tap runs dry, you're wondering how can you get water.

You know, you're panicking.

You're in a state of shock.

...the automotive tech industry...

Again, today's vehicles, it's more complex than it was 20 years ago.

...the science of seasons...

Phenology is the study of seasons and how animals and plants respond to those seasonal changes.

It's all ahead.

Funding for this program is made possible by...

Hello, I'm Hari Sreenivasan.

Welcome to 'SciTech Now,' our weekly program bringing you the latest breakthroughs in science, technology, and innovation.

Let's get started.

Have you ever wondered why Siri, Alexa, and many other personal assistant devices speak with female voices?

Curious about whether systems using artificial intelligence may treat people differently based on gender or race, joining me now is Sarah Myers-West, a postdoctoral researcher at the A.I. Now Institute at New York University who is studying bias in this new technology.

So, a lot of us have digital assistants on our phones now, and some people have them in their homes.

You know, most people don't think about the fact that they're Siri and Alexa.

What does that do to us over time?

Sure.

So, we're constantly using these devices and issuing commands, asking them for help, and we don't necessarily think about how their programmed and what those programs are sort of reflecting back to us.

When these phones are defaulting to a female voice, we're issuing commands and sort of placing as these female-gendered devices in positions of servitude essentially.

So they're like the maids.

You were just saying that basically anybody who's going to be in this subservient position is a woman by default.

Exactly, and these programs are also sort of reflecting back cultural biases and stereotypes.

You'll find when you interact with Siri that often if you are getting angry or things, she'll just kind of deflect and try and de-escalate the situation.

Give me an example of that.

One of them is if you ask Siri to marry you, you can go, 'Hey, Siri, will you marry me?'

So, she'll also sometimes say, like, 'Let's just be friends.'

There's a lot sort of more abusive Easter eggs in there as well, but essentially what she tries to do is present, you know, a neutral, happy, cheerful response to any kind of thing you confront her with.

Is that something that ends up changing how subconsciously we perceive things?

Yes, so as scholars like Safiya Noble have noted, if we're across these many different interactions and contexts, if this is the kind of engagement that we're having, it is going to affect our implicit understandings of how we relate to women out in the world.

And artificial intelligence is essentially a reflection of the people who programmed it, right?

And if meaning the people who wrote the codes, all their biases are coming out in the app.

Well, not just their own biases, but biases that exist out in the wider world.

So it's certainly an issue of the design of the system, but often it's just as much, you know, how these systems are trained on data sets that are culled from, you know, our many interactions and historical data that's been produced.

And what that's going to do is reflect back and amplify existing biases but often in ways that are much less visible to us, so it's harder to acknowledge and harder to grapple with.

So for example, a lot of different résumé filters use artificial intelligence today, right?

Some human being is not going through all 30,000 applicants.

They're going to try to figure out how to get it down to a manageable number that a human can actually screen.

But what have been some of the problems that have happened in that?

Right.

One of those ways is in scanning through résumés and looking for keywords and trying to essentially profile match to things that a company might think are a good fit for the job.

But one study by a company called Upturn found that one such résumé scanner matched people named Josh who had played high school lacrosse as being a great fit for a job one that had absolutely nothing to do with the requirements.

It's more a reflection of past hiring biases -- basically taking discrimination from the past and bringing it forward into the future.

And if you have staff that's full of people from a certain college and not from other colleges, then your résumé filter might end up wanting to just bring more of those people in.

Precisely, and that's not something that's necessarily intentional on the part of the designers.

Another example would be Amazon had created its own internal recruiting tool.

Now, this was never actually put into use, but they were testing out if they could use the same kinds of optimization logics that they use on the buying platform, you know, making a recommendation if you liked this book, you might also like this book.

Well, if you hired this employee, maybe you also want to hire this employee.

But what they found was that the résumé screener was systematically demoting résumés by people who went to all-women's colleges or who even had the word 'women's' in their résumés.

Now, this was unintentional.

This is part of trying to replicate Amazon's past hiring practices, which are not particularly diverse.

But the real concern is that these are some of the best engineers in the world and they applied the leading techniques on how to de-bias or reduce the discrimination in the system, and essentially it broke down.

It would no longer work unless it had that discrimination built in.

And what about facial recognition?

That's something that artificial intelligence is being used with or on a lot these days.

What are some of the inherent biases in that system?

Yes.

So, studies have shown that facial recognition systems are -- have much lower accuracy rates on women and particularly women with darker skin pigmentation.

So there was a really excellent study by Joy Buolamwini and Timnit Gebru called the gender shade study that looked at three of the leading facial recognition platforms, and systematically they found that these systems just simply don't work as well if you have darker skin or if you're a woman.

Is that partly because the data said that there was the standard for what a face should look like maybe didn't use enough women or darker-skinned women?

Exactly, and that's one of their proposals was to create a more diverse dataset on which these systems can be trained.

But I think that the larger question that can be raised is what does it take to produce a more diverse dataset.

It can result in surveilling those same communities that are disproportionately impacted by these technologies.

So it's important to balance the remedies with the harm.

And, you know, we've had this kind of problem with mugshot books before when, after flipping through page after page after page, somebody who is an eye witness might just start to say, 'You know what?

That person did look like the people that I just saw,' even though maybe that's not what they saw in the first place.

Right.

And we do see that showing up in the technologies as well.

So the ACLU conducted a study where they took the database of images from Congress people, their profile images, and they ran it against Amazon's recognition database looking at a collection of mug shot images.

And they found that the system match 28 Congress people's faces to the images of people who had been arrested.

And so we we see exactly those same kinds of practices emerging in A.I. systems today.

So, is it fixable?

Now, that's really the question.

I think that the core thing to think through is what will it take to fix it.

Now, certainly diversifying data sets is a start.

Certainly we can look at the system design, but we also need to think about the social implications of the use.

Where are we using these systems, and are we defaulting to technology in places where it's not necessarily appropriate to use?

We're all prone to something called automation bias, which is the notion that an algorithmic system is going to be fairer or more accurate than we ourselves are.

And that's not to say that we ourselves are aren't prone to flaws, but it's to say that these systems are not necessarily going to be an improvement on that.

They're going to be a reflection of our existing prejudices.

All right.

Sarah Myers-West from NYU, thanks so much.

Thank you very much.

♪♪

This, in a sense, is kind of a Holy Grail for us.

We're in the outback of Australia because this is some of the oldest convincing evidence for life on Earth, and the Mars 2020 mission and ExoMars mission are going to be looking for signs of life in the ancient past on Mars.

And it was a great opportunity to bring the mission teams here to really see for themselves what we're talking about when we're talking about ancient biosignatures.

This part of Western Australia called the Pilbara specifically, it's really absolutely a mecca for understanding the record of life on Earth.

This is one of the most important places on the planet geologically speaking.

Members of the science team came out here to look at some of the oldest rocks that are on Earth.

These rocks are anywhere from two and a half to three and a half billion years old, about the same age as the rocks that we're going to find on Mars.

And what's very special about them is they have evidence of the earliest life on Earth.

The rocks you see right in front of me here, these wrinkly layered structures that we call stromatolites, structures like this actually represent fossilized microbial mats.

A microbial mat is actually a structure preserved in the rock made by communities of millions and millions of microorganisms.

Basically fossilized pond scum, in the sense.

Microbes, bacteria, living in a shallow water environment.

And we believe that if life ever existed on Mars, it would have been purely microbial.

And those are left behind in the record that are very distinct.

And so we've been showing the NASA and European Space Agency scientists the details of what those textures look like.

When we say with Mars 2020 we're seeking the signs of ancient life on Mars, this is precisely the kinds of signs of life that we'll be seeking.

I learned to be kind of a Martian, to be in a harsh environment.

My comfort zone is the laboratory, but, you know, here I can see that this is the real stuff.

It's not just simulating stuff in the laboratory.

This is a real good training for us.

It was really important to get the science team out here, and speaking for myself, I've seen pictures of rocks like this, and they didn't really convince me that they were the product of life.

And when you see them up close and personal, it really tells a story -- that this was once life.

And that's something you just can't get if you don't go out and look at rocks like this.

You can study it, you can read about it, but there's nothing like the practical experience of trekking around in the desert and really trying to think about as we land our Mars rovers there and we look at the images provided by our sensors, how do we interpret that and follow the clues to try to find the kind of evidence that has been on Earth here in Australia.

Could Mars ever have supported life?

And then to take the next step, did Mars ever host life?

We're just smart enough now about Mars to ask the really hard questions.

♪♪

In Detroit, Michigan, we follow a community advocate who's working to help residents get their water turned back on.

This segment is part of the Great Lakes Now reporting initiative, and here's the story.

♪♪

Demeeko Williams calls himself a community advocate.

He runs a group called Hydrate Detroit, which helps people when their water has been shut off.

You have 45,000 people that are on payment plans, 9,000 people shut off, and there is no room for amnesty.

There is no room for even modification of some of these outrageous outstanding bills.

Earlier this week, he was on the road across Detroit meeting his clients -- people trying to get their water turned back on.

Everything is done by a system of giving.

We reach out to social media.

We reach out to our friends, our family, our neighbors to see what they can do to help.

When you work off $200 a month.

what can you do?

Dushawn Adams lives on the east side.

She lost her water a week ago.

My new bill just came in.

It's over $900 at this point.

We'll go to do...and see what happens.

Yeah, that's what they told me.

Right, 'cause I had already called about you and it was the market in computer.

When I called the lady today, and I wasn't able to get in touch with her.

The bill just built up, built up, built up.

I'm searching for employment, but I'm limited to what I can do because of my physical limitations.

I have multiple health problems.

For me, the main one is having to sleep with a CPAP machine that does not operate without water.

This water shut off could make or break a person.

When the tap runs dry, you're wondering how can you get water.

You know, you're panicking.

You're in a state of shock.

You don't know what the hell to do.

And basically Miss Adams stayed as calm as anybody I've seen in this condition.

Charlotte Woods bought her home on the far west side, on the other side of Telegraph Road a few years ago.

She works full time at an auto parts manufacturer.

She's also a single mother with four kids.

Her water was shut off two months ago.

She says she can probably handle reasonable water bill payments but not a bill adding up to thousands.

Wow.

Yes, wow. [ Chuckles ]

What in the world?

What happened here?

There is a spout that's on the other side of the house.

Okay.

And I was unaware that was there.

I didn't know it was there.

And I've been here about, what, maybe ever since -- I bought the house in March of '14, and I didn't know that there was a spigot on the outside.

The only way I found out was that the next door neighbor was like, 'Did you know that thing was a water main?'

This water right here will probably last me for about four days, about -- well, this part of it will last me no more than about like a day and a half.

But this because they get very good about drinking water.

They just drink one or two a day.

I'm just worried, first of all, just about the kids.

That's my biggest concern is the kids.

You know, brushing your teeth.

You know, the good water to brush their teeth with, wash their face, to take towel baths with.

Because of my water situation, I keep plenty of Clorox on hand for the bathroom, for the kitchen to clean other things until we get the water back on.

Yeah, I have to talk with my team about this, see what we could do, because $9,000 is -- and we could get this knocked down.

We can get this knocked down.

Demeeko Williams and Hydrate Detroit can only help a few clients at a time, but Williams wants more people to know there are some outside agencies like his that are here to help.

It could happen to anybody, but we're not talking.

And when you don't talk, people don't know.

And if people don't know, how could people help you?

♪♪

My name is Leah Atai.

I am 17 years old, and I invented the Syllaboat.

The Syllaboat is an aquatic robot that is used to teach middle school students how to code.

In middle school, I became introduced to robotics, and that's the entire reason I'm interested in computer science today.

In order to get more students involved in computer science, a big part of it is making it accessible.

So I wanted to recreate a tool that middle schoolers could use and be passionate about just not at a $400 price tag.

When I was younger, I tried, like, imagining different inventions, and I remember having a journal where I'd write down a different invention each day.

It's just about seeing a problem and thinking of different ways to solve it through building and creating.

I don't necessarily want every kid that uses this toy to go into computer science.

I just want them to be a little more confident and know that they have the capability to try new things.

This production is part of American Graduate: Getting to Work -- a public media initiative made possible by the Corporation for Public Broadcasting.

♪♪ ♪♪

The national 'Race to 2026' program partners with local technical training schools to offer opportunities for students looking to get into the automotive industry.

In this segment, we take a look at the program.

We don't have enough students to supply the workforce.

That's a fact.

Lincoln Technical Institute Campus President Robert Paganini says the demand for qualified technicians is growing in the automotive industry.

Charlie Smith is with Icahn Automotive Group, a company that owns and operates businesses that provide automotive parts and services.

He points out the culture of working on a car in your backyard as a kid with your parent is long gone.

We used to have a lot of kids when they were just out of high school or just out of a vocational school, right?

So a lot of the vocational schools have dropped the automotive program throughout the country.

So that's a big problem also when we talk about the shortage of it, right?

So the interest is not there.

Pop a hood.

There's so many covers and electronics.

With the advance in technology and the intimidation factor of today's automobiles, we're seeing students start to steer to a different career path because of that.

You really need a degree of electrical engineering, mechanical.

So there's a lot in today's vehicles that's more complex than it was 20 years ago.

Smith says his company, like many others across the country, is already finding it difficult to find qualified technicians, and that challenge is only expected to grow.

The U.S. Bureau of Labor Statistics predicts by 2026 there will be 46,000 positions that need to be filled in this industry.

As these schools get less and less students, and we have more and more workers retiring from the field, it's creating a large shortage throughout the whole industry.

That's why Icahn Automotive Group created its national 'Race to 2026' program.

The company is teaming up with technical training schools like Lincoln Technical Institute by offering scholarships, classrooms, and internship opportunities, among other things.

We're here to help promote and give exposure to these schools and help get students understanding that the skilled trade is really a great trade to go into because it can have a huge need.

This generation has been dealing with technology their whole lives, and they start to understand that they can do this and that they can excel at this because they are in such demand.

Lincoln Technical Institute's Campus President says they work in close contact with industry players to make sure the school's curriculum advances with technology.

♪♪ ♪♪

Students in upstate New York have created the Finger Lakes phenology trail as a tool to help hikers and citizen scientists better understand the seasons.

PBS NewsHour Student Reporting Labs has the story.

I'm Carol Klepack, and I've lived in Ithaca for the last 30 years.

I love the Finger Lakes area.

I love the lakes.

I love the rolling agricultural lands, the forests, the spring wildflowers, and I've supported the Finger Lakes Land Trust, So when I heard that they were starting a phenology project, I became interested in learning more about it and then participating in it.

My name is Hannah George, and I'm the conservation easement steward here at the Finger Lakes Land Trust.

Finger Lakes Land Trust has a phenology trail at the Roy H. Park Preserve in Dryden, New York, a place where people can go on a hike and recorded their observations of six different trees that are part of a national database.

I wasn't familiar with the word phenology until the Land Trust began its project last spring.

Phenology is the study of seasons and how animals and plants respond to the seasonal changes.

For example, in plants, flowers opening for the first time that spring, falling leaves, a breaking leaf buds -- those are all phenological events, which we call phenophases.

There is an app on my phone which records the data.

It's all set up with the stations that Hannah has put in front of certain trees that they want to document week after week.

That data actually gets entered into this national database that people such as scientists or other people interested in plants, they can look at that data and use it for their research.

[ Bird singing ] I started the phenology trail with my fellow intern Nick Dietschler, and he's actually also using citizen science to help with combating this invasive pest that's threatening our forests.

The hemlock woolly adelgid, or HWA for short, is an aphid-like insect that was introduced to the east coast from Japan in early 1900s.

Hemlock woolly adelgid was first detected in the Finger Lakes region in 2008.

It spread pretty rapidly throughout the Finger Lakes.

This pest has moved in and has been causing the decline of hemlock trees, and these trees make up more than 25% of our forests in this region.

We have a steep slope behind our little lake cottage, and there's hemlock that have been badly affected by woolly adelgid.

And you see kind of hemlock stands that -- it's so sad.

These magnificent trees that just look like they're sick.

Here in the lab, we raise predators that feed on hemlock woolly adelgid.

Currently, we're working with a beetle, Laricobius nigrinus, or Little Larry, and a species of fly called silver flies.

These predators are native to where hemlock woolly adelgid comes from, and we bring them here to New York State to research them and release them into the wild to help control hemlock woolly adelgid across its introduced range and protect hemlock trees for the future.

By enlisting citizen science observers around the state, they give us the information that we need so we can release the predators at the appropriate time of year, and this is crucial in the success of the biological control program.

So we release the predators when they have the right life stage of the adelgid to feed on.

Citizen science observations like these are really important for many different types of research, and the phenology trail at the Park Preserve is another place where these types of crucial observations can be gathered for researchers to use.

A lot of the people that visit the trail don't know what phenology is.

But when they get there, they find out that there is actually a way for them to get involved in citizen science.

What's interesting is that a lot of people don't know what these trees are and they walk past them every single day.

Red maple, sugar maple, white pine -- people just look at those, and they're just another leafy plant.

I think to go out to the same area week after week and just see the changes and, you know, turning from brown to vibrant green and now there's birdsong and now there's insects buzzing around and it's just this blossoming of this living planet, that's so exciting.

And that wraps it up for this time.

For more on science, technology, and innovation, visit our website, check us out on Facebook and Instagram, and join the conversation on Twitter.

You can also subscribe to our YouTube channel.

Until then, I'm Hari Sreenivasan.

Thanks for watching.

Funding for this program is made possible by... ♪♪ ♪♪ ♪♪ ♪♪ ♪♪