AI & Society: How can technology make the world a better place?
Rosana de Oliveira Gomes18:372021
The AI that filters your spam can also find tumors years earlier
The same machine learning that recommends purchases and flags credit card fraud can, as the speaker cites from a Google Health study, detect a breast tumor four years before it developed. Rosana de Oliveira Gomes, an astrophysics PhD and Lead Machine Learning Engineer at Omdena, argues that people in tech should point it at society's most urgent problems. Which real problems, from endangered species to children at risk online, can AI already help solve?
See how computer vision, NLP and deep learning already serve health, nature, safety and art
Rosana de Oliveira Gomes, Lead Machine Learning Engineer at Omdena, argues that the AI used for recommendations, spam filters and fraud detection should also be aimed at the Sustainable Development Goals. She maps the challenges society faces in the 21st century and walks through examples of computer vision, natural language processing and deep learning in healthcare, conservation, online safety and art, ending with a call for more women and minorities in AI for good.
“And at the same time, funnily enough, we are getting to the era of misinformation.”
“And of course, if you can predict this faster, then we can treat the patient faster. And this can be crucial whether someone will survive or not.”
What you will learn
Connect AI techniques such as computer vision, natural language processing and deep learning to specific Sustainable Development Goals.
Recognize how automating image review and drug discovery frees doctors to focus on work that needs their expertise.
Identify environmental uses of AI, from counting endangered animals with drones to mapping rooftops for solar panels and sorting trash.
Understand how text analysis can surface hidden harms such as domestic abuse and predatory conversations with children.
Find a place in AI for good communities, whether you are new to the field or experienced.
Key takeaways from the speaker
AI for good
Details
Level
Beginner
Format
Talk
Technical depth
Introductory technical
For
All career stages
Length
18:37
Published
June 2021
Questions this talk answers
Why should technologists care about the Sustainable Development Goals?
Around 193 nations agreed to help reach the goals by 2030 as a blueprint for a better, more sustainable future. She argues that AI can save crucial time on urgent problems such as health, poverty and climate.
How can AI help in healthcare?
AI can speed up drug discovery and use computer vision to diagnose faster, for example detecting cancer earlier. Her team won an AI healthcare hackathon with an engine that detects fractures in X-rays and generates the report.
Will AI replace doctors?
She says AI cannot replace doctors. It can take over reviewing large numbers of images so doctors spend their time on work that needs their knowledge and can prioritize patients.
How can AI protect people from harm online and at home?
Text analysis of social media and search trends helped show a worldwide rise in domestic abuse during the pandemic. NLP can also assess whether a conversation is predatory and power warnings or chatbots that alert children.
Who is needed in AI for good work?
Everybody, from newcomers to very experienced people. She calls especially for more women and minorities in AI for good initiatives, whether through NGOs or industry.
About this talk
AI is commonly know for its great potential in industry applications such as costumer behavior, marketing, among others. However, these same technologies and methods can also be used for the better good of society and not only for profit.
In talk, I'll provide an overview about the many challenges that society faces in the 21st century and how AI and other technologies can help us to achieve the Sustainable Development Goals. I will present different AI and Machine Learning techniques and discuss their potential and several sectors such as energies and environment, vulnerability and even art.
How text analysis exposed hidden domestic abuse (75 s)
We need everybody in AI for good (66 s)
Speaker 1: Hi, everybody. Thank you so much for voting to make this talk public to everybody and to being here today. I'm very pleased to join you today to talk about the topic that I am so passionate about, that is how to make the world a better place through technology. And I am a data scientist and machine learning engineer, and I've been dedicating my career in the past years to AI applied to social impact.
So maybe like me, some of you also grew up in the 90s. And as I was a kid, I was super excited about the future, about robots and futuristic cities and flying cars and all the cool stuff. Speaker 1: And well, we know that we are for over two decades in the 21st century, and this is not necessarily the case. However, of course, we managed to do quite cool stuff with science, such as like putting robots on the surface of another planet just to explore it and so on.
Speaker 1: But of course, we have technology everywhere. And we use technology on a daily basis for the most mundane things like taking pictures, move around the city, Speaker 1: and among other technologies. There's also AI that can help us to, for example, recommend items that you want to buy or to predict whether email that we have as a spam or not, or if there is some fraud in our credit cards.
Speaker 1: And that's all very nice. But of course, that we can still ask ourselves, what if we could actually use this? So such powerful tool like AI, Speaker 1: to actually address other social problems in our society. So, as you may know, almost all nations in the world, like around 193 nations in the world agreed to help to reach the sustainable development goals by 2030. So, Speaker 1: Those sustainable development goals are a collection of designed and interconnected goals to reach what is called like a blueprint to achieve a better and more sustainable future for all.
Speaker 1: And those goals go from healthcare, education, addressing poverty, among many other things. So, these are like the most concerning topics of our time. And we can think like, Speaker 1: how we can actually use technology to address them. But maybe the first question we need to ask is what is society like? What do we understand as society? And when you think about it, Speaker 1: it has some ingredients like here, I put a couple of them. And then this doesn't mean that, of course, they exist in all societies and so on. But like, we all have like, when we had like our normal lives, and even nowadays, during the COVID times, Speaker 1: we have, you know, like people to help us to keep us safe as police and firefighters and we also have like places that provide us with services such as supermarkets or like the post office and town hall and so on. And we have, of course, like infrastructure such as like to provide us a supply of energy or gas or water. And we have like places where we live and so on. And
And when we think of all of this society, we can still think that we can cluster all of these elements of society into a more general topic, like into a more general topic, sorry. And here I am showing just a couple of them, like just as an example of how we can society in terms of all those clusters.
And there is this one ingredient that is still missing, like when we think about society, and this is the challenges that we have in it. So in the 21st century, we came to the 21st century with a lot of problems that we had before, and there are also like piling up, I would say. And for example, we have the topic of overpopulation, and it is estimated that by 2050, we're going to have 9.7 billion people on our, and that brings a lot of limitations to our resources and a lot of, for example, conflict.
There are conflicts that we already have, actually, like man-made conflicts that are created through like a society or political differences. And that can, for example, lead to the refugee crisis that, like I'm trying to show here, down the bottom to the right. And this refugee crisis is also something that can become even worse with the problem of climate change as well.
So we have nowadays the climate change challenge, one of the biggest ones together with the environmental crisis that we have at the moment. And this is putting our planet to the limit. So we have like CO2 emissions, we have plastic pollution, and we also have, of course, like risk of extinction of many species that are crucial to keep the biodiversity in our planet.
And at the same time, we still have a lot of, nowadays, we've never been in such an era of our world in which we have access to so much information and technology. And at the same time, funnily enough, we are getting to the era of misinformation. So nowadays, you're reaching the point that everything that is on the news is more, like, there is more fake information on the news than actually real information.
So we have like all those challenges present in our society, apart from all the ingredients that we need society to work. And, but there's another ingredient to consider, and this ingredient is technology. And this is what I'm trying to convince you all of this talk today. And my idea with this talk is to ask you the question, how we, as people from the technology fields, from my face, like the AI, but of you in other fields as well, as well in technology, how we can tackle all those problems within those parts of society with technology.
So, there are many applications of AI for, you know, like companies to just improve production or just make companies earn more money, but the idea is that like, how can we also use technologies to make the world a better place through, for example, automation of, or even like help with all those problems that we have that need to be solved urgently.
So, many of the problems that I mentioned need to be solved, like just for example, the climate crisis, because we don't really have time and that's the power, at least that I see the potential there is re automate many of those problems that we have. And, um, just try to us to like save time that this crucial thing to solve these problems.
So, in the next minute, we teasing you a little bit with just a couple of those examples of how we can use AI to actually make the world a better place. Of course, there are many other applications and, uh, if you're interested, you're more than welcome to, you know, reach out and talk to me and so on.
And also like many of my other friends that are also in this kind of link in this, um, in this challenge together. So, um, I will start with healthcare, uh, healthcare is associated with sustainable development goal. Number three, there is about, uh, wellbeing and health and, uh, and in this topic of healthcare, I can play a massive role, such as for example, um, automating and speeding up processes for, for example, drug discovery.
So drug discovery is a matter of trying to understand which components, uh, of a drug can be, uh, useful to create a new one that might be more efficient and so on. And this is exactly the, the kind of like thing that AI is very, uh, efficient at and data analysis and just like the automating processes, just analyzing several things at the same time.
And so, so for this data analysis and machine learning as well can be very helpful. And when it comes to diagnosis, uh, computer vision, there was one technique of AI that is just like the ability of actually, uh, process images and extract information from images, uh, together with deep learning that it's, um, and machine learning, there are also other AI techniques, uh, can be crucial.
Understand the diagnosis faster. So they're extremely powerful techniques. And so just to mention about examples, like, uh, for cancer, uh, AI can detect, uh, cancer, like, uh, with even the image here is like, uh, AI can detect, uh, breast tumor four years before it was developed. And, um, um, this is a study from the Google health and this is, uh, the publication.
And, um, it's like, now we know that AI can, uh, detect a cancer more accurately than doctors at this point in time. And this can help reducing false positives and also predict the chances of a patient to live. And of course, if you can predict this faster, then we can treat the patient faster. And this can be crucial whether someone will survive or not.
And, um, extremely important, um, to, uh, help, uh, doctors to just make an automation of their tasks. So for example, instead of, they have to look into several images, they can actually spend their time, um, like they can have the AI doing this for them. And then they can really spend their time, um, focusing on, on researchers and things that are more, um, they need more the intellect and their knowledge.
So in this sense, AI cannot replace doctors, of course, but it can be an AI, let's hope to say. And, uh, just to mention in an example, this is a work that me and my colleagues here did, uh, we won the AI healthcare hackathon. And, uh, by creating an engine that, uh, not only detects, uh, fractures in X-ray images, but it also generates a report.
So this is an entire automation of the medical process that would, uh, that is still in many places still done, like very manually. And that it requires someone to take an image and look and, and write a report by hand sometimes even. So this is an entire automation of the process and this is crucial. So for example, to make doctors prioritize patients and just be faster in the entire process, which for example, now in COVID times is extremely important.
So that's it for help. Um, um, oops, sorry, this shouldn't be here. Uh, we can also talk about, uh, how AI can help to make the world a cleaner and greener, uh, place. Uh, by having just more sustainable solutions. So just to mention a couple of them, you can think of computer vision again, and a geospatial analysis to, uh, for example, uh, help detecting animals and counting them.
And this is important because as I mentioned, we have like this, uh, several animal animals in our species, sorry, in the risk of extinction. Um, by, for example, being able to track animals, like a tracking device and those, uh, endangered species. And then we can see whether they are, um, safe, for example, whether there are other, uh, predators around and so on this way, we can protect them.
And, uh, so for example, we can have drones that come and just, uh, see like, uh, that doesn't disturb them so much. Like they can just see whether they are there and so on. And by that, they can recognize the images of the animals and, um, and just count, uh, how many of them are there. If they reproduce and so on.
And, uh, we also have computer vision that can also be used to automate installations of solar panels. For example, uh, you can just like, uh, see images of rooftops from the top or from satellite images and use this to count. And, uh, of course, um, it's also, we can also have like, uh, AI for helping us to sort our trash.
For example, you can take a picture of, of your trash and like, for example, have a bottle of water or something. And then the AI can help you to like with machine learning, for example, can use techniques to classify the kind of trash as that, and then help you to, to tell next with your, with your, uh, trash.
So just, these are just like above examples, things that I worked with before. And, uh, and like, uh, AI can also be used for safety. So nowadays we have at home, a lot of people. Uh, online and, um, we, uh, the beginning of the pandemic became clear, uh, that, uh, there is a lot of, uh, domestic abuse happening nowadays, uh, because everybody's locked.
And, uh, thanks to the, uh, data analysis or, uh, natural language processing, like text analysis, we can see, for example, what kind of, uh, topics are being posted on social media. What kind of topics are being researched on Google, for example, and, or such, sorry, on, on Google, for example. And then this way we can know, uh, we managed to track the correlation between the rise of, um, searches for domestic abuse, uh, keys and so on.
And, uh, and this way we managed to identify that this is actually something that is happening in the world nowadays, like a worldwide, as a matter of fact. And, uh, now we actually, we have like a, probably, uh, as you remember, like last year, we created the symbol, the symbols that we could use to, if you're in a call and you want to say that you are suffering domestic abuse and so on.
These are like, uh, things that we get to know through text analysis, using machine learning as another example. And of course we have the, the unfortunate case of like, uh, kids also being exposed to predators online. So this is something that, uh, we can also use NLP. So for example, to create, to understand the risk of those conversations, to understand what kind of like a language the predators use and, uh, analyze the risk of a conversation to be a predatory conversation.
And this way also, for example, create a warning system or even create a chat bot that can help kids to just say like, Hey, this is a dangerous conversation. We actually have this in some, uh, platforms and there may be kids as well. And, uh, just to mention the last example, I just want to mention that we also have art and, uh, art is not necessarily something that is like an obvious application of AI for good.
But it's also like a way that understand, uh, ourselves better, like a artist part of the human condition. So to say, and in the past art was done mostly by human beings. And nowadays we start to have this combination of machine that learns from the patterns of famous artists or basically whoever artists and can help artists. And, uh, even people that are not necessarily artists to actually understand art and participate on the.
And just to mention here a couple of examples, we have the portrait. Let me. That is something that was done, uh, with, uh, against. So this is like, uh, this was not painted by anyone. And we also have the chapter of Harry Potter. There was written with a recurrent neural networks, just like by understanding the sequences and so on.
And we also have like this, uh, idea of a style transfer that you can, it's a technique from deep learning that you can, uh, understand the patterns of some paintings or something, and then apply it to other kinds of techniques. Uh, sorry, other parts of images. So this way you can have the same techniques, um, and kind of make a combination of art and so on.
And, um, I just want to say like, to finalize saying that of course, we still have a lot of, um, challenges in society, but we still have a lot of, uh, technologies such as AI and many others that can have the potential to help. And of course we have to still, uh, integrate our AI and technologies to our daily lives and, uh, address topics such as bias and ethics and data literacy as well.
But we have a powerful technology and we should want to think that we are in the 21st century for real. We have to really start to use these technologies. And this is our, our examples that are being used. And there are many others into, um, but also addressing all those topics that are. That are the come also the challenges that come with the technology as well.
So this way we can finally, uh, um, sustainable development, better society. And, uh, I'll end, uh, with a call of action that, uh, we need a lot of diversity in this, uh, uh, field and like, and also to change the world. And, um, the more women that we have involved in the more minorities that we have involved in all these initiatives available.
Uh, initiatives of AI for good, uh, be it through NGOs or to industry. This is extremely important to, to us to, to become a better society. And then in my opinion, also like better humans. And, uh, it doesn't really matter if you're starting in the field or if you already like a very experienced person. Uh, we need, uh, everybody, we need the help of everybody.
So here I leave with, um, um, um, just showing you some communities that you can join that are doing a lot of good work in terms of AI for good. And, uh, also try to connect with like-minded people. I, uh, took me a while, but I managed to find all my friends that are like people that are like me are also very passionate about this topic.
And I'm pretty sure that many of you are as well. So, um, shout to me, like, here's my contact and, uh, I'll be happy to discuss more with you. Thank you.
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Article based on this talk
Can the AI behind spam filters help solve society's biggest problems?
Rosana de Oliveira Gomes shows how computer vision, language processing and deep learning are already working on health, nature, safety and art.
In short
Rosana de Oliveira Gomes argues that the AI already used to recommend products, filter spam and catch credit card fraud can be pointed at humanity's hardest problems. Framing her talk around the Sustainable Development Goals, she shows how computer vision, natural language processing and deep learning are helping in healthcare, conservation, online safety and even art. She closes with a call for more women and minorities to join AI for good, at any stage of their careers.
Key takeaways
AI for good.
Connect AI techniques such as computer vision, natural language processing and deep learning to specific Sustainable Development Goals.
Recognize how automating image review and drug discovery frees doctors to focus on work that needs their expertise.
Identify environmental uses of AI, from counting endangered animals with drones to mapping rooftops for solar panels and sorting trash.
Understand how text analysis can surface hidden harms such as domestic abuse and predatory conversations with children.
From flying cars to the Sustainable Development Goals
Gomes opens with a memory many viewers will share. Growing up in the 1990s, she pictured a future of robots, futuristic cities and flying cars. "And as I was a kid, I was super excited about the future, about robots and futuristic cities and flying cars and all the cool stuff." (0:43)
Two decades into the 21st century, that future has not quite arrived, yet technology is everywhere, from photos and navigation to AI that recommends purchases or spots fraud. Her question is what would happen if this powerful tool were aimed at social problems instead. She points to the Sustainable Development Goals, which, as she cites, around 193 nations agreed to help reach by 2030. The goals span healthcare, education, poverty and more, and she presents them as a blueprint that technologists can work toward.
A society full of challenges
Before turning to solutions, Gomes describes what society is made of: people who keep us safe, services such as supermarkets and post offices, infrastructure for energy and water, and the places we live. The missing ingredient is the set of challenges piling up in this century. She cites an estimate of 9.7 billion people by 2050, which strains resources and can drive conflict. Man-made conflicts feed the refugee crisis, and climate change may make it worse. CO2 emissions, plastic pollution and species extinction push the planet to its limit.
She also highlights a paradox of the information age: "And at the same time, funnily enough, we are getting to the era of misinformation." (5:31)
For Gomes, technology is the final ingredient. Many of these problems, the climate crisis above all, cannot wait, and she sees AI's ability to automate work as a way to save the time that solving them requires.
Healthcare: faster answers, not fewer doctors
Healthcare, linked to the third Sustainable Development Goal, gets the most attention. AI can speed up drug discovery by analyzing many components at once, and computer vision combined with deep learning can support faster diagnosis. As the speaker cited from a Google Health study, AI detected a breast tumor four years before it developed. She stresses why speed matters: "And of course, if you can predict this faster, then we can treat the patient faster. And this can be crucial whether someone will survive or not." (9:30)
She is clear that AI does not replace doctors. Instead, it can take over reviewing large numbers of images so clinicians can focus on work that needs their knowledge. Her own team won an AI healthcare hackathon with an engine that detects fractures in X-ray images and also generates the report, a process that in many places is still written by hand. "So this is an entire automation of the process and this is crucial." (10:43)
A greener planet and safer online spaces
Gomes then turns to the environment. "Uh, we can also talk about, uh, how AI can help to make the world a cleaner and greener, uh, place." (11:08) Computer vision and geospatial analysis can count endangered animals, with drones observing them without causing much disturbance, so conservationists can see whether a population is safe or reproducing. Images of rooftops, from above or from satellites, can help plan solar panel installations, and a photo of your trash can let a classifier tell you how to sort it.
Safety is her third area. At the start of the pandemic, with people locked at home, text analysis of social media posts and search trends helped reveal a worldwide rise in domestic abuse, knowledge that led to signals people could use on calls to show they were suffering abuse. NLP can also learn the language predators use, assess the risk that a conversation with a child is predatory, and power warning systems or chatbots that alert children to danger.
Art, ethics and a call for everyone
Her final example is less obvious: art, which she describes as part of the human condition. Machines can learn patterns from artists and help both artists and non-artists take part. She mentions a portrait that was not painted by anyone, a Harry Potter chapter written by recurrent neural networks, and style transfer, a deep learning technique that applies the patterns of one painting to other images.
Gomes acknowledges that integrating AI into daily life means addressing bias, ethics and data literacy. Still, her conclusion is urgent: "But we have a powerful technology and we should want to think that we are in the 21st century for real. We have to really start to use these technologies." (16:52)
She ends with a call for diversity, asking more women and minorities to join AI for good, whether through NGOs or industry. "And, uh, it doesn't really matter if you're starting in the field or if you already like a very experienced person. Uh, we need, uh, everybody, we need the help of everybody." (17:38) She encourages viewers to join AI for good communities and to find like-minded people, as she eventually did.
Watch the talk
"AI & Society: How can technology make the world a better place?" by Rosana de Oliveira Gomes, 19 minutes: https://www.womentech.net/video/rosana-de-oliveira-gomes-ai-society-how-can-technology-make-world-better-placeapply-speak
About the speaker
Rosana de Oliveira Gomes is Lead Machine Learning Engineer at Omdena and holds a PhD in Astrophysics, dedicating her career to technology for social impact, with experience as a Lead Machine Learning Engineer and Project Manager on projects in disaster management, renewable energies and sustainability. She is passionate about empowering women and young students into AI, and offers advice on getting into AI or changing careers into AI through calls and her Recommendation of the Week series on LinkedIn.
Questions this talk answers
Why should technologists care about the Sustainable Development Goals?
How can AI help in healthcare?
Will AI replace doctors?
How can AI protect people from harm online and at home?
Who is needed in AI for good work?
Drafted with AI from the talk transcript. Editor review pending.
Frameworks and ideas mentioned
Sustainable Development Goals
Cite this talk
Rosana de Oliveira Gomes. "AI & Society: How can technology make the world a better place?." WomenTech Network, 2021. https://www.womentech.net/video/rosana-de-oliveira-gomes-ai-society-how-can-technology-make-world-better-placeapply-speak