AI4Nomads

There are 139 million people above 60 in India. 4 to 15 million informal and 140 million migrant workers in India of all ages. What do they all have in common? Poverty, computer illiteracy, no upward mobility, no social safety net – and a mobile phone.

88% of Indians own a mobile phone and 40% of Indian households will have internet access by 2023. 82% now use Whatsapp.

Data-labeling jobs pay 4 dollars an hour minimum in India, yet most Indians live on less than 4 dollars per day. Digital illiteracy is the only thing between them and that 4 dollars an hour.

What if the platform to label data for AI is as easy to operate as a social messenger app? If they can navigate WhatsApp, why can’t they annotate data? Our solution is to use technology to make technical jobs easier for millions to complete.

Low skilled, unskilled, informal, middle aged workers are losing every day, to the advances of technology. What should lift them up is instead dragging them deep into poverty.

The pandemic has only exacerbated their condition, expediating their economic decline. Welders, farm workers, factory workers, vendors, guards, tailors, bus drivers, housekeepers, gardeners, and many menial job holders are now asked to stay home because of their age and their comorbidity conditions to Covid-19 risk. Thrusting many into immediate destitution. They are not allowed to board trains, buses, or any public transport. They cannot commute to their workplaces and they are prohibited from working for their own safety.  Many contract and freelancers are also seeing their jobs disappear. Re-skilling is almost improbable, given their lack of formal education, and will not help them given the current restrictions and the continued elimination of their jobs due to technical advances.

Without any social safety net to catch them, they are doomed to starvation and neglect. More than 400 million in India alone are impacted. Out of which 140 million are above 40.

The global data annotation tool market is projected to reach 2.57 billion U.S dollars by 2027 and job creation in this area is booming. Out of this, manual and image/video annotation is expected to boom in coming years. The young, computer literate, high school educated population can be upskilled to benefit from this, while those 40 and up, digitally illiterate are left to wither in poverty. The complex computer interface which requires digital literacy and location specific workstations are the two major hurdles in the way of their economic empowerment. 

Most own a mobile phone with access to Wi-Fi and can navigate a messaging app. Our goal is to utilize this singular skill, to create friendly UI for the worker to complete an annotation task without learning any of the complex computer literacy skills. In short, our platform will enable anyone that can operate WhatsApp to partake in AI/ML training data annotation. Our platform’s API will pull tasks from local partnering companies, sort and display it on simplest navigation interface possible with a powerful backend processing, to our workers, to label accordingly, and get paid directly.  

We downskill digital mobile platforms to match competencies instead of upskilling workforce.

Our solution will currently target the informal, low skilled, middle aged population greatly affected by the pandemic. Our initial focus groups are limited to demographics who were employed prior to Covid-19 and recently laid off. Small factory workers, street vendors, domestic help and gig workers like Uber, Ola drivers. We are documenting their daily struggles, accessing their skill levels. Many of them are on the brink of despair as support from family is dwindling as the pandemic has impacted everyone’s bottom line. Most of them survive on less than 4 dollars per day. Less than 150 dollars a month total covers food, medicine including their rent. Many have lost their homes. Along with the migrant crisis, local populations like this are also facing displacement.

Data and image annotation at the lowest level pays 28 dollars more they earn per day! At the higher end, as much as 400 dollars a month; the kind of money that this population has never seen even on their best days.

Our solution will not just sustain them during crises but create a reliable income source for them in the future. A remote working environment will protect them from crises like the pandemic.

HI for technology-based solutions for good jobs and inclusive entrepreneurship.
AI is transforming the world in every sector of life. The advances of AI are mostly positive but, in its wake, it is also decimating many livelihoods. It is the most segregated and homogenous sector of our modern workforce. It requires a very narrow skillset; higher level of education and it favors the young. Our solution breaks down all barriers by allowing the most underutilized, underappreciated humans and their intelligence to label the training data for AI. Our solution makes technology work for the people, instead people working for technology.

iMerit.net, Playment.io, MTURK etc., are all webbased platforms for human data annotation for ML training. They combine both automation and human labelers to train data for AI/ML algorithms. Most of them have their own proprietary platforms and computer vision tools or they partner with established sources, and offer workers training in labeling, quality control and workflow feedbacks. Once established, they source from partners like AWS Sagemaker and Google Cloud for the pipeline of tasks.

Many recruit workers from underserved communities, but none of them provide a workaround for anyone beyond a certain age, education level to step into the technology sector. College or high school education, younger, with minimum computer literacy skill is required.

The tools and the platform are user friendly only to those who meet that standard.

The job by itself is not as complex as the platform. Humans of any age or education level can identify, label and provide contextual information on their surroundings. It is the raw human intelligence that is needed for image and data labeling to produce accuracy in AI algorithms and not a sophisticated knowledge of technology!

Human vision, categorization, understanding of speech and language nuances and human ability to learn from experiences – these skills are innate to all. In fact, those hardened by struggles in life, have heightened awareness in this.

They know how to navigate on a dime, as well as live on one.

Our platform accommodates and accentuates that ability.

We augment HI to train AI.

Our solution separates the labeling platform & its tools from jargon - that is visual and functionality jargon.We will source tasks and push them to workers on our mobile interface.

We start with the following data labels and will expand further.

• Image Annotation with Bounding
• Image Annotation Cuboid
• Image Annotation Polygon
• Image Annotation Semantic Segmentation
• Video Frame Annotation

All of them require simply drawing the boundaries around an object to classify its edges and identify it on a sliding confidence scale. Just draw & rate your confidence! Our app’s UI will allow anyone to do that. No complex instructions, no jargon.

A demo will auto play to guide the workers. And there will direct feedback from the app to guide workers to label accurately. This feedback will be powered by a ML model which will act as the QA.

Our ML /QA backend will double check it, and then push the labeled data to the client. We believe strongly in human intelligence to power this program. We don’t think labeling data has to be made so complex. Humans have been labeling, tagging data from birth! Kids can identify a dog from a cat after having seen only one of each, and yet machines still need to see hundreds in each category before it can do the same.

The problem is the interface. AI4Nomads believes in simplification of technology to enable everyone to participate in the digital economy. HI is the engine behind AI, and it should be inclusive.

Image labeling for AI training is an established field, and the technology for that labeling is widely used and accepted technology. Data annotation is outsourced to India at extremely low costs to companies here. Instead of exploiting the market for the benefit of corporate profit, we hope to utilize its potential to create life sustaining income for millions.

The article below sets the background for our innovative idea.

In the article below, they mention the age and minimum skillset required to do this data labeling job. What we are trying to do is to eliminate those entry level barriers by creating a simpler interface and a robust backend platform. Data annotation is already web based but not yet mobile.  Our AI annotation platform is mobile, remote and has no entry level barriers.

Anyone who gets laid off, or has limited education or has gone thru a major life event requiring them to stay indoors will have access to this market. Our platform simply opens the door wider and creates a path to financial security. Young Indians are already working on MTURK and iMerit, Samasource. We are just taking it mobile, fully remote and inclusive to all.

Same model but easier to use.

https://timesofindia.indiatimes.com/india/how-artificial-intelligence-is-creating-jobs-in-india-not-just-stealing-them/articleshow/71030863.cms

The rise of data labeling tech companies in India:

https://analyticsindiamag.com/will-we-see-more-data-labelling-jobs-in-india/

The fact that most data labeling can be moved to remote during a pandemic.

https://infolks.info/stay-home

The fact that this is a booming industry with potential for job creation in underserved and impoverished areas of India.

https://www.grandviewresearch.com/press-release/global-data-annotation-tools-market

Our theory of change posits that immediate economic relief using existing skills empowers humans to seek learning and be open to new ways of earning an income. It opens their mind to see their own power and to see technology in a welcoming light instead of seeing it as the enemy of their livelihood.

We believe technology should be the bridge to economical and social equality, and not the obstacle.

Yet it is not possible to go back to school when you have mouths to feed, a roof to keep, and bills to pay. It is also not possible to upskill yourself without knowing what sector to focus in, especially when jobs in most industries are being automated away. And it is almost impossible to re-skill or research new opportunities during a pandemic. When curfews and lockdowns decimate local markets, small businesses and small traders, it creates the perfect storm for lasting financial despair. And creates psychological despair which can immobilize many and drive them to desperation.

Our solution is two fold: 1) Using what they have on hand, a mobile phone & their innate human intelligence, we focus on providing them an income source to make ends meet. 2) Without the immediate fear of economic despair, we hope to open the door wide for exploring new avenues for income and gaining more technical skills.

Throughout the world, those who moved successfully to remote models were able to survive the pandemic or at least managed to scrape by. Providing that kind of model to AI data annotation jobs and also making the platform easier to use will insulate millions more to have that kind of financial security that right now only the IT sector has.

Once they master annotation on a mobile platform, moving them to laptops would not be so challenging. Formal education may be out of reach for them, but using their existing life skills in new ways to reach economic stability is still within reach.

We plan to launch our app AI4Nomads to android phones in three specific locales. 50 people in each locale. Total of 150 to 200 people for starters. In one year, we will scale up to 1000 in each locale, and expand to 100 more new locales.

In five years, we hope to be scaling up to a workforce of 20 to 40 thousand strong. Our hope is to go beyond those thousands and reach the millions of people as early as possible with partners. Since our platform is a mobile and online application, we believe we can scale up more rapidly than our expectations.

But we don't believe in releasing just the app without coordinating a support system to guide them. This is why we are releasing it within small groups. We plan to nominate a local guide for each cohort who will be with them until they master the end to end process. We will be supplementing them short video tutorials and teaching them the larger scope of AI, and how they can continue to play a role in training it to benefit them eventually. Opening new avenues of income for all skill levels is our long term goal. That nobody should be left behind in the Fourth Industrial Revolution. Creating an inclusive digital work place is not that difficult.

Impact goals:

Our primary goal is sustainable income. With local coordinators, we will orient the worker to the app’s primary functions within minutes, instead of months of training. Instead of commuting to a tech center, this will be in their homes.

Our mobile demos will be interactive and will issue a small payment at the end, if the user correctly tags the images in the demo. This is to show them how the system works and also to ensure their banking info is set up correctly on the app. There is no need to wait for monthly wages or for a check in the mail.

We know the workers will be encouraged to continue with tasks after seeing how easy it is to complete tasks and receive payment. Initially the tasks will be limited to image tagging, but progressively move on to boundary mapping, edge defining and semantic, contextual mapping, paced with the worker’s ability to ramp up. For those who have laptops, there will be a web application similar to the mobile one.

1YR: Many will be able to live in their homes, pay the rent, buy their medicines and groceries without depending on charity and government hand-outs.

5YR: A robust subscription model to source tasks from vendors, and a growing community of remote annotators with laptops & mobiles. Skills that can be transformed to other remote digital tasks as well, opening a wider economic window for millions.

Technical & Time Barriers:

This is biggest barrier for us now and it is not because we lack technically savvy team members, we have them, but they don’t have time. This project was started in March with a group of friends in the technology sector but many of them had to work from home, with their kids also at home – and spare time was non-existent. We barely managed to scrape out a demo prototype but it still lacked a lot of backend and frontend functionality. And we were unable to iterate for the reasons given below.

Market barriers:

India went through a migrant crisis that was beyond anyone’s imagination. Millions of people who were temporary workers left the cities to go back to villages. Starvation in the cities was a real possibility for them, so they decided to take their chances in their hometowns. Since transportation was restricted, many left on foot and succumbed to their death on the way. India was curfewed into zones and travel from city to city was prohibited. Interstate travel was also prohibited.

The situation on the ground was too dire for an intervention like this.

Our local team members could not travel to our desired locales to collect data such as worker’s needs, existing language skills, type of mobile phones and access to internet.

Financial Barriers:

Although the project was started with donated time and talents, it cannot be sustained in that manner. We need to hire dedicated engineers and app developers.

We hope conditions on the ground will improve within the next few weeks as India comes out of the lockdowns and curfews.

We decided to pivot and focus on people with whom we have established contact, our friends’ housekeepers, domestic helpers, tailors, drivers, and their immediate families. Since all of them are also affected and have lost their jobs, they can be our focus group for launching the pilot testing.

As for our own barriers on time and technical support, we plan to hire a dedicated app developer to augment our existing team members and complete the app. 

To fund it, we are submitting our project here and also planning to start a website and launch crowdfunding measures very soon.

Launch the project as a non-profit organization. Raise funds through crowd funding orgs like Kickstarter.

Partner with local companies for image annotation task orders or partner with existing crowd sourced entities like MTURK to pull tasks for our workers. Partner with digital pay platforms to ensure instant payments. Assess security of data and workers privacy at every stage.

We don’t plan on launching an app in the Apple or Google app store and hope it works!

We plan on launching a sustainable community-based movement towards economic mobility with technology. We will have local, community members guiding them until they are confident, to figure out the best way to integrate them into the digital economy and data labeling industry. To create a welcoming place for them in technology.

We currently have 4 volunteers including myself.

Susanna R.

Yogesh K.

Sweety B.

Leena P.

All of us work in the Technology Sector. Our educational and experience backgrounds are: A computer engineer, an android app developer, a computer applications business analyst and a cognitive science researcher who has worked in the AI sector. We understand the scope of AI and the cultural, economic makeup of India, since it is our native place of birth.

We currently do not have any partners.

Our business model is sourcing tasks from third party vendors and pushing them through our user friendly mobile platform to our workers. We plan to offer a monthly subscription service for data annotation to our task suppliers, to sustain our platform. Our workers can access and complete tasks for instant payment without any cost incurred.

Our impact is sustainable income for informal and unskilled workers. We will provide data labeling services to third party vendors, suppliers, and companies by utilizing a diverse workforce that is fully remote. Our labeling will be quality control checked and will have cultural diversity and inclusiveness.

We plan to generate our revenue by offering a monthly subscription model of service to data labeling projects. Current data labeling subscriptions charge by the hour for each data labeling task, but we plan to charge a monthly fee for continued influx of projects from partnering companies and vendors. Our business model will be non-profit. The revenue collected through subscription will be to maintain the platform and a lean team of technical and admin support.

We want mentorship and advice. We want to be part of a community of academic and professional experts who can guide us on this journey.

We want feedback on feasibility of the project and any unknown variables to execution or development that we may have missed.

We want funding and ongoing expert technical support. We need funding to establish the ground work for a project like this. Solve seems like the perfect place to seek advice, and monetary help.

We want advice on business model and solution technology in the area of AI/ML training data labeling on a mobile platform.

We also want partners in making this project a success story by connecting us to networks of experts and companies that will want to do business with us.

We need ML and QA tools on the backend of our apps, and we are looking for partners in that area.

We are very open to ideas and anyone who shares our common value is welcome to join us. Our goal is not to make money for us but to help millions of people garnish a livable income. Economic mobility is our purpose and mission. Local community support, or providing us with product support like mobile phones and laptops to be distributed to marginalized communities, are all welcome. Technical support as we develop and distribute the app is also welcome.

Expert advice, mentorship, and guidance is also welcome.

The continued support of MIT Solve community members would be ideal.

Our mobile app can be used by anyone who wants to find a way to make sustainable income. The only requirement is a mobile phone or a laptop with internet access. Refugees need work placement and until they find one, this can provide them with income. Income lifts people out of despair. Will enable them to pay for a temporary rental home. The only barrier I see is how will they be paid on the mobile, if they don't have an account established but that can be solved with a community account or a digital banking solution. We are willing to work with community members or local agencies to support Refugees in finding sustainable income with our technology.

I am a woman, and 3 out of the 4 team members on this project are women. And this project is focused on being inclusive to all ages, gender and socioeconomic conditions. Women are included in this.

Especially domestic and household informal workers in India who are mostly female. And right now they are all out of jobs. They don't have any other skills except doing household work. Our hope is to bring thousands of women like that into the tech sector and slowly build them a sustainable, stable, independent life style.

Our project is focused on creating good jobs and inclusive entrepreneurship. We hope to eliminate digital illiteracy with technology. Fighting fire with fire. Use technology to create platforms that can be accessed by all, and enable everyone to participate in the digital economy. We see human intelligence as innate, and it can be used to build our future technologies, but barriers like socioeconomic conditions and education stand in the way. Our project wants to disrupt that model.  Everyone can use a mobile. Everyone can use Whatsapp and requires no training. Why not use the same model to train AI? Why not include all in that process? Our project aims to answer yes to all that! Our project will use this money to create not just good jobs but better jobs for all.

Yes, our project believes in augmenting HI to train AI! We are advancing inclusive economic growth through upskilling mobile UI and creating greater digital literacy! We will use this award to create digital literacy videos that are simple and easy to understand, focusing on how AI is trained and how anyone can play a role in it! We will use it to enable community based support for digital jobs and also for providing laptops, internet access to rural communities.

AI is now deployed in every aspect of our lives. Yet AI is not trained by a diverse set of humans. Humans who come from the same cultural, social backgrounds, the same educational backgrounds, with the same degrees, from the same communities are now training our AI.

The result?

It is full of bias. Bias against women. Bias against minorities like Asians and African Americans. Bias in data collection, bias in data labeling, and bias in data interpreting and analyzing. Bias, bias everywhere!

We are now opening our eyes to the mess we created. It is huge and solutions have to be multi-pronged. But our project eliminates bias in one area - bias in data labeling. Our project is now limited to India but it can easily be extrapolated to other countries and regions. That is actually my hope.

Data should be labeled by the young, the old, the middle-aged, the married, the unmarried, the educated, the uneducated and people from all stations of life. How will a artist label a set of data? A musician? A farmer? Woman vs man? We all label data differently. That diversity is lacking in AI now.

Our app and web solutions will open the world of diversity to AI data labeling. Our future hopes are to train our workers to become full time data label agents and develop technical skills that will benefit their everyday lives. Access to healthcare and other resources are now online and many cannot access them because of digital illiteracy. We hope to eliminate that as well. We hope to bring diversity in AI data labeling to healthcare as well. That will require a remodeling of platform or resources to re-train workers in that area. With the money, we can afford to pay for both.