Presented by PinkLion.AI


“Now that we’re all in this unique situation with COVID-19, do we have an opportunity to do something dramatically different together?” asked Jennifer Bonine, co-founder and CEO of PinkLion.AI, at GamesBeat 2020. “What are some of the new strategies we’re going to leverage in our companies to change the way we’ve been doing things and re-architect how we look at a problem? We have time, space, and the opportunity to learn more about the new technologies and strategies we aren’t using,” says Bonine, “Gartner shows us that the two things that are going to change the game for organizations are AI and machine learning combined with data analytics.”

For Bonine, and her co-founders, Rick Faulise, COO, and Andrew Birkholz, CTO, AI and machine learning are a competitive differentiator for companies harnessing the technology in games testing.

“AI and machine learning in games development offers tremendous data that can be leveraged to make smarter decisions about your product road map, your players’ needs, and how they’re engaging in a game,” Bonine said. “Now is the time to look at the games developers are working on and determine the high-value-added tasks that could be tackled with AI.”

VB Transform 2020 Online - July 15-17. Join leading AI executives: Register for the free livestream.

The use of AI allows the developer to take traditional strategic tasks that were previously considered impossible and transcends everything in game development to an attainable task.

“For instance, traditional automation applied to games — anything running on Unity, Unreal, or a custom game engine — has a problem because it’s hooking into the code and the DOM layer. Natural language processing and computer vision can visualize what’s on the screen, and parse those images into individual components for automated testing; identifying, and then comparing the data. Using this type of information instead of hooking into that DOM layer for any of the gaming engines, is much more effective for automated detection, identification of different elements, and labeling and classifying the elements on the screen,” Bonine explained.

Bonine provides additional use cases surrounding labor-intensive tasks that are great candidates for AI deployment. The technology has enabled developers to test the stores in mobile apps without needing to go into the game to individually purchase items and then confirm that the transactions worked. A bot can check and validate thousands of items in a game and ensure that those items are appropriately installed and readily available.

“[AI bots have] also been able to tie the business components in games stores directly to the interactive plane of the game,” said Rick Faulise, co-founder and chief operations officer. “This allows them to create single scenarios to traverse different platforms. The bots have crawled multiple apps — hundreds of thousands of them — and they have started to store that information into a narrow network so [they] can recognize the different buttons, forms, and the common things in mobile apps. Then it chops them up into the individual outlets that it has learned. It can recognize these images based on computer vision algorithms, and NLP to take the text off an image. This allows you to test the screen — it doesn’t worry about the underpinning layers.”

“One of the big things we’ve done with the game stores is tie the business components directly to the interactive game components and create single scenarios that allow us to traverse different platforms,” Faulise explained. “Since [AI testing] is all screen-based, as long as the game is the same from screen to screen, it doesn’t matter if you’re going from mobile to desktop to PlayStation. The scenarios don’t need to change. We don’t need to reclassify and retrain the bots — as long as the images stay the same — this adds another layer in lowering maintenance.”

“Outside of the stores, it’s very difficult to test gameplay considering it’s an uncontrolled environment,” Faulise said. “It takes a while to get the bots up to speed so they are able to handle the fluctuations that occur when other users engage. This is where the technology is able to test scenarios and keep the bots trained on relatively simple tasks; but they’re high-profile tasks that get seen by a lot of people — the ones that will create a lot of havoc if things go sideways or are not operating correctly.”

AI has also proven to be particularly useful for the product management teams inside of an organization, Bonine notes. “[AI technology] can be used by product owners, product managers, and the people who are responsible for what gets out to the market to enhance the software development life cycle and the speed and scale at which [the developer] can deploy games, develop product road maps, and understand what players want.”

“There’s a plethora of data out there that isn’t being used,” said Andrew Birkholz, co-founder and CTO.

“If any of you have read the thousands upon thousands of user reviews in customer feedback analysis, there’s a gold mine of information to drive your product and make decisions. ut, it’s also a very mind-numbing task to go through thousands of reviews and categorize them.”

Birkholz says that text classification can do the job. “An AI categorization model can group reviews to determine whether or not a problem is a defect: Was it a UX issue, or a customer service issue? Did the app crash? Was there a lag? A second model can identify actual sentiment of a review, whether it’s a positive, mixed, negative, or neutral sentiment. Yet another model can take groups of reviews and then summarize them into one to three sentences.”

“Sometimes, when you’re building an app or a game, and you’re doing it for months on end, you don’t really see what the users see,” Birkholz says. “Sometimes you need an editor to take a look. [Text classification offers] a nice perspective that you may not have had as a team, and opens up a huge opportunity for dashboarding.”

The ability to analyze data with AI goes beyond customer feedback analysis of course, he says. “You can integrate AI into a DevOps pipeline to automate any kind of analysis. It can be used for bug tracking, or to track any metric that will help you develop and ship code as quickly as possible. The idea is to automate as much as possible.”

The number of use cases is growing, Bonine says. “And now we have a bit of time to reflect and look at strategy.”

“What are the things companies are going to do to optimize and position themselves as leaders? Be able to operate remotely and virtually?” she asked. “Right now it’s about embracing the chaos. Instead of overthinking — will this work? will it not work? — get involved and do some proof of concept planning on things in your organization [asking] ‘How do I employ AI first?’ This is a great opportunity to try things you haven’t considered before. This is a space that will change how testing is done in gaming. The companies we are working with now are seeing a huge uplift, and I think it’s something everyone should be looking at. We want PinkLion and our logo to represent diversity and inclusion in technology and AI. If you contact us, we would love to engage with you about what you are working on – we love you throwing out hard problems and challenges we’ve never seen. Write to us at Jennifer, Andrew, or Rick @PinkLion.AI.”


Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact sales@venturebeat.com.