I knew very little about the field of AI data annotation until I chatted with Iva Gumnishka, co-founder of Humans in the Loop. Humans in the loop is a social enterprise that provides ethical human in the loop workforce solutions to power the AI industry.
Many companies that advertise their products as AI-driven are not entirely AI-driven. Some of them have real people working in the background to answer or do things machines are not capable of.
If you’re interested in AI and want to know how we can use a more human-centric approach so that AI doesn’t displace us, this is the episode for you!
0:00 – Eric speaks Bulgarian
2:10 – Google fires AI engineer
4:05 – stereotypes around AI
5:40 – Foxomation and ghost work
8:00 – Ghost workers in Colombia
10:17 – Why it’s a good idea for companies to keep using humans
11:30 – the opportunity Humans in the Loop provides
13:03 – how and where they recruit people
14:25 – hardest obstacle to overcome
18:03 -what Iva has read lately that has changed her perspective on AI
20:18 – AI in the home
22:16 – best advice Iva received from her dad
For the Innovators Can Laugh newsletter in your inbox every week, subscribe to our newsletter.
Past guests on Innovators Can Laugh include Yannik Veys, Ovi Negrean, Arnaud Belinga, Csaba Zajdó, Dagobert Renouf, Andrei Zinkevich, Viktorija Cijunskyte, Lukas Kaminskis, Pija Indriunaite, Monika Paule, PhD, Vytautas Zabulis, Leon van der Laan, Ieva Vaitkevičiūtė.
Additional episodes you might enjoy:
Hi everyone today, we're chatting with Eva Ishka co-founder of humans in the loop. Humans in the loop is a social enterprise that provides ethical human in the loop workforce solutions to power the AI industry. Well, what does that mean? Well, you know, how all those companies that advertise their products as artificial intelligence, while many of those AI products are not entirely AI driven, some of them have real people working in the background to answer or do things.
Machines are not yet capable of doing. If you're interested in AI and want to know how we can use a more human centric approach so that AI doesn't displace. Then this is the episode for you. Let's dive in. So let's get started is, is Eva.
Oh, wow. Super good job. This was really impressive. I've been practicing you know, B lady yeah. The Bulgaria language is a it's it's very different. In terms of the characters, like, is there another language that it's pretty close to here in Europe somewhere? Oh, yeah, absolutely. A lot of the SL languages also use a similar alphabet.
Some of them like Serbian, for example, like Russian, but yeah, a, a lot of them, even if they use Latin, but they're quite similar as well. Okay. A lot of languages in Eastern Europe. Okay. So that's it because I have yet to inter interview a guess that came from a SL you know, language background. So this is the first time.
So when I saw the the translation, I, I, I was too tempted to not try and say how you doing how are you? Okay. So let's get started, Eva. I recently read that Google. Has fired an engineer over breaching its confidentiality agreement. After he made a claim that the tech giants conversation, artificial intelligence is Senti it because it has feelings, emotions, and subjective experiences.
What are your thoughts on this? Yeah, that was a big scandal. And I think a lot of people were left with impression. The person was fired because of this claim. But yeah, it was mostly because of the, you know, disclosing internal information. I think it's quite interesting because we as humans, when we interact with our social intelligence we tend to ascribe human characteristics to it.
Then we like, for example, especially. And, and companies do take advantage of that. You know, when, when companies create broad wilds, they tend to make them even like very cute or, you know, very humanlike. A lot of them have like sens, female voices and so on. So they do try to transmit this kind of like hu humanity in terms of, you know, the, the tells that they have, but they're.
Statistical systems of, you know, what is a person lets most likely to say in terms of a chat bot that you're developing? So of course, because it's based on their responses and the, the iterations and the its of so many people in the end, it does end up sounding like an actual person. But it's just a Amalga of all of the different authors that you've trained the system on.
So in the sense, it isn't like an actual person. It's just a, a copy or, yeah, a combination of all the thousands of people who've contributed to the system. God, it's, it's becoming so real. Like, you know every year I feel like AI is becoming so much entrenched in, in our lives. What are some misconceptions in stereotypes that people have around AI that you come across?
Well, I think that usually. When you talk about AI, people tend to imagine like the Terminator or like robots and a lot. What, what is known as general artificial intelligence? So let's say an artificial intelligence that can actually do a lot of things and may even be able to create more, let's say system of systems of it on its own.
So that's kind of, you know, the point where we would lose control over the artificial intelligence. And, you know, this is something that a lot of people are speculating about and, and writing about, but we are far away from that. I would. Right now, you know, you're, you're lucky if you are, let's say your vision system and able to distinguish well between different types of pizzas.
For example, if you're trying to, you know, have like a bit of detector or different types of shoes, if you want, if you wanna train like a shoe detector or a brand detector or something. Yeah. Just because, you know, very frequently the actual applications of AI are quite. You know practical and, and they're not that ambitious.
Let's and they're just related to things that we are trying to alternate or make easier for users, or just do it a very, very large skill, which is something that we, we couldn't do previously, especially if some type of human labor or human judgment was involved. Okay. Now, researching getting prepared for this interview and researching you and your.
I came across two works, two words that I was unfamiliar with the first one was Fox donation and the second one was ghost work. And I, I think they're roughly the same thing. Right. Can you describe what Fox donation is? And then also share what an example that a company has been guilty of when, when doing.
We are. Sure. So this is quite relevant to the work that we do, which is preparing the data sets that are used to train AI systems. And this, this term comes from an and let's say a pretense of a company that's automating some service, but there are actually humans behind the system performing a lot of the work.
So let's. You know, I'm a company that tries to do a pizza detector and I'm trying to commercialize it and you have offer users a, a very nice app so that they can, I don't know, like find a pizza that they like and, and immediately find out their its ingredients and it's and the recipe. So maybe if my system is not good enough, I'm actually gonna have human operator.
In the back we immediately have to be like pizza experts and they have to send their, the responsible pizza is, yeah. So even though I'm selling something to users, which is like, oh, you know, I have the super advanced digital type there. And it's like really, you know, top not latest technology. It's actually not fully automated.
And that's why it's called automation. And then goals. Work is definitely a related to this. It basically refers to the fact. The people who are in the backend, they're usually hidden and they become ghosts basically because nobody talks about them. Yeah. Companies try to hide the fact that they're using them.
You know, otherwise they will lose all of their view of like high tech systems. If you know, it's revealed that they actually use humans. Yeah. So this is why a lot of humans that are working in the back end are also suffering from like difficult conditions of work and low wages. And there isn't a lot of visibility on their role and.
It's usually not appreciated enough and so on. So that's why now, you know, there, there's a lot of research on this type of work. And that's why it's called Bill's work. Yeah. And I, I, I was reading somewhere that there was on a university, a robot that was on campus and people thought that it was just a very smart robot, but it turned out that there was actually ghost workers in Columbia that were being paid very little to actually control the robot.
And there was a backlash sort of a backlash against the company because they were trying to to hide that. Right. And that's what I didn't know that I found so fascinating is that now when I see all these advertisements about, you know, this has AI, this has AI now in the back of my head, I'm wondering.
Maybe that's not entirely true. There's gonna be some people that are behind the scenes doing a part of the job that can't be done by the robot or the AI itself. And, and so where do you know, where does your work come in? How are you involved in, in this, this whole, you know, life cycle of AI? How we're trying to corporations are trying to add more AI into their electronics or their robots or their products.
What, what are, where are you involved in this? Yeah, so I would say, you know, my point of view and, and what I'm trying to promote is that it's not shameful to be using humans behind the AI systems. You know, it's not something that it be ashamed of or tried to hide. It's actually. Practice, because these are the people who are helping your robot, let's say navigate through, you know, the streets of, you know, of, or, you know, on campus and are enabling it to actually perform very complicated tasks and to take decisions whenever it's not sure where to go or whether it can cross this type of like, I don't know, grassland or whatever it is.
So. The the real world or the human world is quite messy and a lot of unexpected things can happen. And if you don't have a human to actually guide you in certain decision making points, you know, your AI is, is just not gonna work. It's not gonna know what to do. Let's say if there is, I don't know a celebration on campus.
And so we, you know, everything, you know, the, the streets are busy. Your robot doesn't know what to do. And so on. Then if a human operator can take the lead and actually guide it, that data can also be used afterwards to train the AI system so that it becomes better and better with this additional human input.
So what we are trying to do is to provide this human input so that AI systems have someone to rely on, or, you know, if, if there is some kind of alert, there can be a robot or there can be a human operator to help the. So that the systems can become better with time. Because right now, what is usually done is yeah, companies take a bunch of training data.
Let's say images of campus streets. They train the robot and then they deploy it. But there isn't much of like retraining happening in the standardized way. So what we're trying to do is convince companies, Hey, it's a good idea to keep using humans so they can keep helping your robot. Maybe it's gonna start learning and it's gonna become better and better.
But it's still good to have a human somewhere there being available for the system so that, you know, it can keep improving over time and it can also know what to do in weird situations. Yeah. Can you share a story of a person that you've helped and either, either get them a job or immerse them with a company to assist them with their AI and, and their robots?
Yeah, sure. So our entire model is based on providing works on under represented and under privileged groups. So. Training people to become humans in the loop or, you know, this type of human operators. Even if they don't have a lot of skills, even if they don't speak very good English, you know, it's, it's quite an easy job usually just to mark pizzas based on their ingredients or to help, you know, navigate, okay.
Where is the street? Where is the sidewalk? And so, and then someone anyone can do so we're in, in this we're. A great opportunity. And, you know, I was very proud. For example, there is one bar that we're working with here in Bulgaria, and, you know, she was actually asking a lot of questions. Like what's gonna happen with this AI system, you know, as I'm working with it, is it gonna learn?
Is it gonna need me in the future? Maybe it's gonna learn so well, you know, all the things that I'm showing to it, that it's not gonna need me in this future. I'm gonna lose my job. So, you know, I, I was really impressed by, you know, our workers, even though many of them come from underprivileged backgrounds and may not know a lot about technology or AI and so on, you know, we're trying to raise a lot of awareness among them about.
You know, the importance of their role and, and how important, you know, all of the work that they do is even if it's just like marking pieces, you know, it's actually quite important. So the fact that she was asking me about all of these things for me was very impressive because I felt like she was really, you know, understanding the field and, and thinking critically about her own role and the future of the work that she's doing.
How are you recruiting such people. So we usually partner with NGOs in the different countries where we operate. We work a lot in middle east. So for example, beyond Bulgaria, where we're based and where we, where we have a small workforce, we work in Syria, in Iraq, in Afghanistan, in Lebanon, in Yemen.
So these are, you know, a lot of middle Eastern countries that have faced a lot of challenges in recent years. There's a lack of op economic opportunities. There's a big crisis right now happening in Lebanon. The change of government in Afghanistan, you know, economic crisis as well in, in Syria and especially in the north where we're working.
So it's, you know, quite difficult situations. And now we're actually starting to learn beyond these countries and we're doing some pilots in Ukraine in order to help people who've been affected by the current war and in the, the crime Republic, Congo as well to provide. New opportunities to youth who the only, you know, opportunity for them is to work in mining.
So we're really trying to support people in a lot of these different countries, but we always do that through local organizations that can help us understand the local ecosystem and the local situation and navigate it and help people in the best possible way. Okay. Since launching humans in the loop, what's one of the hardest.
Are one of the toughest obstacles that you've had to overcome? Eva, I would say back in 2019 when we were working with people full time we had a small team of human operators or annotators as we call them. And then once their one year contract was over, all of them wanted to drop out except for one person.
So that. That was a very heartbreaking moment for me, because I was like, okay, what am I doing this for? Is this worth it at all? Do people even like this type of job, because it's, it's, you know, a lot of menial work and can become very boring. And we also didn't have a lot of money so that we could pay good salaries for our people.
Mm-hmm . So at that point I was like, oh, okay. Maybe this is not such a good idea. If people don't enjoy this, why do. So that was, that was the only moment when I've absolutely reconsidered. My decision of was this, even, you know, the right thing to do, maybe I should shift to something else or pivot. But what happened was that we still had our clients, so they still needed someone to do their work.
And when were working with a lot of people, friends full time, but we're just freelancers. So they were still happy to do the work. And they were like, yeah. You know, if I'm working as a freelancer, I can edit around all the time. I can work on evenings or on weekends. I don't have to be coming to the office at working full time.
So that was before the pandemic, but even, you know, before the pandemic, we decided, okay, let's shift to a fully remote and three last type of work. So that. All of our people can enjoy this type of freedom and flexibility. So yeah, this is how it went. And in the end, this has been the model that we've been following ever since.
So happy end you know, the, the, the whole story developed in a positive way. Yeah. But yeah, I think this was one of the most difficult moments for me. Yeah, I can imagine, because you think that you're giving somebody, you know, employment an opportunity, you know, a, a, a sense of, of just being able to contribute to something and that you find out, I don't like this anymore.
you're like, where did I go wrong? Okay. Well, that's cool that you were able to pivot and find a solution for that. Now, what, what do you feel like you've done really well? Since you've launched humans in the loop, I think one of my strengths is partnership building. So, you know, all of the NGOs on the ground that we're partnering with, I wave the foundations for that and kind of the model that we're following.
So now it, whenever we decide to expand to a new location, they have kind of the roadmap and you know, this. Is kind of a good model that we're following and the same thing with also some technical partners that we have from the ecosystem. So we're trying to really collaborate and serve clients together with them instead of competing.
And this is just the model that I really like, you know, to find synergies between you, yourself and other people in the same industry instead of competing with them. Yeah. I. It has definitely helped us a lot especially, you know, for skewing to other countries and supporting, you know, for example, Bargo partners.
We we've done a lot of work to support them in terms of capacity building and also providing. Commissions to that for every project that they help us deliver. So that has helped them financially as well. So I think it has definitely been a win-win situation. And then I really like it when, when it happens like this.
Yeah, no, I'm like you, I love partnerships, whether it's just connecting individuals who I think would be interested in learning more about each other and finding synergies or whether it's its companies itself. So I'm with you on that. What are you reading lately? That has blown your mind or changed your perspective around AI?
If, if, if there's a piece of media. Yeah. So I recently read the Atlas of AI as I was preparing for one of my talks. And it's by this research, I think she's based in the us Kate PROFOR, she's amazing. She's like my favorite researcher in the field of AI, because she doesn't only talk about technology.
Fact, she doesn't talk about technology at all. Almost she mainly talks about society. You know, she states kind of this anthropological approach where she really explores the impact of AI on different, you know, communities in terms of how AI is produced. You know, the people who are behind this, for example, this human animal, but also.
Where is the hardware for it assembled and where all of the elements and mineral neurals are required for these systems extracted from you know, what is the impact of AI on, you know, particular industries and so on. So that, that really blew my mind because I really appreciated her. Wide overview of, of the impact of artificial intelligence systems on our lives up to, you know, the, the debris that we're, that, that we're putting into space.
So for me, you know, the fact that the book was called Atlas of AI was really illustrative of the fact, the, the whole approach that she took. I, in order to explore the role of. Okay. Okay. Now this is kind of a, a question that I didn't think to ask you, but, but since you were talking about her and the impact that it has, I'm reluctant to get an Amazon echo or any other sort of electronic.
In my house. That's like a smart, electronic, I, I, I just don't want for me personally, I just don't want my kids thinking. It's so easy to find an answer to a question by saying, you know, Alexa, can you tell me whatever right. I want them to put a little bit of effort in, even though going to Google is really not that much effort compared to when I was a kid.
Right. do you personally, do you personally have any like AI electronics in your house? How do you feel about that? I mean, are, is it immersed in your world? Not at all. I think it's also because here in Bulgaria, people don't usually buy then, you know, it's not that easily accessible and. Maybe, you know, Amazon doesn't have such a deep penetration in the Arian market yet, but also as a course of a choice, I don't feel like I need graduates like that because usually they're connected to like smart home applications on my husband is a big geek.
So I would say that, you know, if he could, he would do a lot of like the modic applications and like automation and flight lighting and, you know, door opening or window opening or whatever it is. But I'm. I'm not such a big fan also because I agree that, you know, firstly, it's, it's the privacy of your life and your data.
Even though, you know, there are a lot of restrictions and measures around what Alexa or sir, and so on can hear they're always on because they're waiting for you to say, you know, to call them. So there are, there is a lot of data is still going through you towards the AI system. And of course there are humans behind it processing a lot of VR data working to improve AI.
So. You know, the show income is a very new one, but yeah, also the fact that, you know, a lot of things that you share in the privacy of your phone may be used afterwards in order to provide you with recommendations of what you should buy and so on. Which of course, I mean, we're already seeing, even if my phone is next to me and I start talking about pizza, maybe next time I open my, you know, my browser in January for the Smiths.
So it's kind of all around, but yeah, I think it's, it's kind of also my, my attempt to. The penetration of air systems in my personal. Yeah. Now I'm with you. I'm with you. Okay. Now for some fun questions. So the audience can kind of get to know your personality. Okay. First question for you is what is the best advice that your mom or dad ever gave you?
Oh, well, I think it's this quote that I found on the back of like this photo at home, which was like a message from my dad and he had written out. Quality people sooner or later succeed. And for me that was, you know, such a discovery when I found it written behind that, that photo. And, and it was, it was a, a really, you know, a moment of, of meeting very, very, very connected with him.
And, you know, the fact that she, he took the time and, you know, hit that message behind this portal without, you know, knowing whether I would ever find find it. So that was very. Nice. Nice, nice. Second question for you. Does corn belong on pizza? I think so. Yeah. I'm, I'm a, I'm a big fan of weird pizzas. I like Hawaiian pizza.
I would pineapple on it. I like different, you know, weird types of pizzas, so yeah. Yeah, you can really like corn pizza. Okay. All right. Last question for you. What is a favorite TV show that you can watch again and again, I think friends just because it's, you know, such a, such a classic and whenever, you know, I wanna just watch something fun.
I just play it, even though I already know all the, the lines and I know what they're gonna, what, what people are gonna say any tip is owed, but it's. Still like so much fun to watch. All right. That's my wife's favorite show too. okay. Yeah, Eva. Thank you so much for being on innovators. Col laugh, everybody who's listening.
This is Eva from humans in the loop. If you enjoy this, feel free to tell others about it. Feel, feel free to to give us a review. Eva, thank you so much for being on the show. This was a pleasure. Thank you so much.
Thanks for listening to the show. If you enjoyed it, I'd really appreciate it. If you could give us a review and star rating, also, don't forget to sign up for the ICO newsletter at innovators, laugh.com, where you can get the bio and details of each guest. Thanks.