UPDATED 09:03 EDT / NOVEMBER 06 2019

AI

Applause launches reinvented AI solution to make algorithms act more human

Crowdsourced automated testing company Applause today announced the launch of a new artificial intelligence training solution that will help machine learning algorithms better process human input data.

The new scalable, enterprise-ready solution trains machine learning algorithms to learn quickly and tests them to detect and correct bias, thus ensuring that they are processing and responding appropriately.

To do this, Applause is using its large community of vetted human testers in order to produce the widest variety of potential training inputs in an environment that best matches real-life conditions for the products and services being tested. That data is then fed back through the system and tested across every possible device, location and circumstance in order to help identify issues that might arise and provide actionable user feedback in real time.

“Users want AI to be more natural, more human. Applause’s crowdsourced approach delivers what AI has been missing: a diverse and large collection of real human interactions prior to release,” said Kristin Simonini, vice president of product at Applause.

By design, the data Applause gathers is designed to help avoid the bias of small, isolated groups, which would be a poor representation of any population. This is especially an issue with talented testers, whose population could easily introduce biases simply because people who train to test products could easily come from the same backgrounds.

As a result, the machine learning training data collected by the company comes from people across numerous countries, ages, genders, races, cultures political affiliations, ideologies, socioeconomic backgrounds, education levels and more. That broad base of data samples results in a fashion designed to better model a less biased output than if the data had been sourced from a smaller group.

“Not only will this improve AI experiences for consumers everywhere, the breadth of the community also has the potential to mitigate bias concerns and make AI more representative of the real world,” Simonini said.

All types of AI currently on the market have suffered an extremely similar problem: sourcing enough data to teach the ML algorithms how to interpret and respond. That challenge has hampered the production of many AI solutions from nutrition apps designed to identify food to virtual assistants learning how different users ask for the same thing.

Applause’s new AI solution operates across five different types of AI engagement: voice, optical character recognition, image recognition, biometrics and chatbots.

For voice, the data pool uses source utterances to train voice-enabled devices. For OCR, it contains read documents and corresponding text visually to build context. With image recognition, it provides machine learning algorithms the ability to detect and categorized predefined objects and locations. For biometrics, it sources biometric inputs such as faces and fingerprints. And for chatbots, the data sources sample questions and varying intents to better understand user needs and respond in a more human way.

When teams training AI systems fail to take into account the large variety of potential human inputs, it can result in more than simply bad customer service or products. Some extremely large machine learning failures have made the news because they could not properly identify faces such as when in 2015 a Google LLC photo identification algorithm misidentified black people as “gorillas” (the company corrected this by removing gorillas from its training stock). And in 2016 a New Zealand passport verification algorithm misidentified Richard Lee, a 22-year-old DJ with Asian features, as having his eyes closed.

Although these failures were well-publicized and the underlying problem seems obvious in hindsight, they also reveal a deep, cultural problem with how AI systems are being trained for interaction with diverse human populations. The proliferation and implementation of machine learning as it interacts with people is a human-oriented problem and requires solutions that take that into account.

Using its crowdsourced pool of human data, Applause hopes to help take the first step towards building a better AI training system that will reduce the possibility of these errors in the future and potentially detect and warn about bias before it becomes a human problem.

Further details are available on Applause’s blog about this service update.

Photo: Pixabay

A message from John Furrier, co-founder of SiliconANGLE:

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU