How can tech companies achieve responsible AI? Trust, ethics, and responsibility in data management at scale

At the Data Summit conference held by VentureBeat, Saiph Savage, one of the leading experts in AI ethics, and Olga Megorskaya, CEO of Toloka AI and the woman "spearheading the company's mission to advance the AI industry through empowering people" according to Tech Times, sat together with VB's senior staff writer, Kyle Wiggers, to go over a set of best practices for how data-driven businesses can achieve Responsible AI.

Combating bias in AI

One of the biggest contributing factors to biases in AI is the lack of representation among data labelers, who are the people who initially train AI models.

"One of the critical things to think about is, on the one hand, being able to get different types of workforces to conduct the data labeling for your company," Saiph Savage, an assistant professor and director of the Civic AI Lab at the Khoury College of Computer Sciences at Northeastern University, said during VentureBeat's Data Summit conference. "Let's say that you only recruit workers from New York. It's very likely that the workers from New York might even have different ways of labeling information compared to a worker from a rural region, based on their different types of experiences and even different types of biases that workers can have."

Toloka has been making really great progress in being inclusive and in representing all types of workers, as well as designing interfaces that are tailored for their different needs to create a more heterogeneous workforce, and through this create wider representation.

— Saiph Savage

"An effective model cannot be trained on data collected by a narrow group of pre-selected people who only engage in data annotation. The broader the audience, as well as their collective expertise and range of experiences, the richer the signal for the models," commented Olga Megorskaya. "If you're a business, applying AI responsibly means constantly monitoring the quality of the models that you have deployed in production at every moment of time, and understanding where the decisions made by AI come from. You have to understand the data which these models were trained on and constantly update the training models to the current context which the model is operating in".

The human touch in AI

According to Olga, Responsible AI also means responsible treatment of people who are actually acting behind the scenes in training AI models. She explained that this is where Toloka's close cooperation with many researchers and universities comes from.

Recently, Tech Times included Toloka in its list of data labeling companies making a positive social impact, outlining how the company is "empowering workers, academics, and industry actors to be able to better cooperate with each other and ensure that each actor can reach their goals". It also mentioned that Toloka's "users have praised the company for evangelizing crowdsourcing, AI, and ethical treatment of workers"

Toloka has come a long way since being founded in 2014; today, the platform provides extra income opportunities to upwards of 200,000 users from more than 100 countries around the world. The company has also recently launched in Israel, and is excited to join its vibrant AI community in tackling pressing AI and ML challenges and supporting research in Responsible AI.

In collaboration with researchers at laboratories like the Civic AI Lab at the Khoury College of Computer Sciences at Northeastern University, where Saiph Savage works, the company also develops tools to assist in the detection of biases in datasets, as well as tools that provide rapid feedback about issues related to labeling projects that could have an impact on a client company's interfaces, project or tools.

One such collaborative project has recently been awarded UNESCO's 100 Most Impactful in AI designation.

You can watch the recording of Saiph, Kyle, and Olga's conversation at the VentureBeat Data Summit website here.