Toxmod is an AI that scans voice chats to detect toxic gamers in multiplayer titles. Now, Modulate, the company behind the AI, has raised $30 million to further develop it. The funding round was led by Lakestar, with Everblue Management and Hyperlane Ventures contributing, among others.
Toxmod utilises machine learning to identify players displaying toxic behaviors in online voice chats. Toxicity toward other players is a long-standing issue in multiplayer games. The potential for such ill behaviors only intensifies with the growth of metaverse experiences like Lava Labs’ upcoming multiplayer Web3 game, Afar.
“This is a problem that everyone in the industry has desperately needed to solve,” Modulate CEO Mike Pappas told GamesBeat. This is such a large-scale market need, and we were waiting to prove that we’ve actually built the product to satisfy this.”
Toxmod aids human moderators
Modulate’s AI can identify various situations and determine which players are instigating and causing the most problems. Currently, Toxmod is detecting problematic players with a 98 per cent accuracy rate. This helps human moderators respond to incidents of player toxicity 25 times faster than they would normally.
“They can start taking immediate action. Occasionally, it can misunderstand the conversation as human language is complicated,” said Pappas. “And each game has different standards.”
While the AI sometimes misidentifies players as toxic, it will continue to learn from human moderators who check its results and identify its errors. The software must consider tone when analysing speech and whether cheating allegations prompt aggressive speech. Toxmod also needs to differentiate between acceptable trash talk common in mature-rated titles and hateful and prejudiced speech.
“No company world has built out this kind of data set specifically designed to focus on real social voice chats online,” Pappas added. “That has allowed us to build accuracy in our models that beats any of the public big company transcription models out there by a pretty handy percentage.”