Summary
- Sony is exploring the use of machine learning to combat spammers targeting video game chat systems, as indicated by recent patents.
- The patented system would analyze individual messages and user behavior to generate a composite score, which would then be passed to another system to determine if punishments like warnings, chat suspensions, or bans are necessary.
- These patents demonstrate Sony's continued experimentation with machine learning in gaming communications, although their real-world implementation is not guaranteed.
Sony is exploring ways of combating spammers using machine learning, some newly emerged intellectual property filings reveal. The Japanese gaming giant has been experimenting with neural networks for years by now, with its research most recently yielding a patent for an AI system that would help PlayStation gamers manage their time.
As is the case with most widespread avenues of communication, video game chat systems are a popular target for spammers. Even some of the world's biggest developers continue to struggle with cracking down on such behavior, as previously illustrated by the influx of player complaints about World of Warcraft spam harassment, among many other examples.
Sony now appears to be developing technology that would combat barrages of bad-faith messaging with the help of AI, as indicated by the fact that the company just secured two patents describing such solutions. One of those details a messaging risk management system using gamer behavior to gradually learn to identify and penalize spam, while the other envisions an underlying mechanism for annotating large data sets in order to define models that would assist in training neural networks to combat spammy-looking communications.
On a fundamental level, the patented system would feed each individual message to its chat model, which would then produce a spam rating based on its actual contents. The mechanism would then also rate the overall account behavior with a separate user model, which would generate a higher-level outlook of a given player's spamming tendencies, or lack thereof. The two figures would then be joined into a composite score and sent for final evaluation. Should Sony's system determine that the examined message has a high spam risk rating, it would invoke a separate "punishment engine" that would decide whether to issue a warning, temporary chat suspension, or an outright ban.
These newly emerged IP filings are not the first sign of the PlayStation maker experimenting with machine learning applications in gaming communications. Back in spring, Sony secured a patent for modifying voice messages to filter out strong emotions using what's essentially a real-time AI censor. As was the case with that solution, this anti-spam system was seemingly envisioned as an application of supervised machine learning that would allow it to gradually become better at accomplishing its task as human researchers keep grading its performance.
The existence of these Sony patents does not necessarily guarantee either will ever see real-world use cases. After all, the PlayStation maker is one of the most prolific inventors in the world, with a recent Statista study revealing it filed for nearly 1,400 patents in 2022 alone, the vast majority of which don't appear to be verging on commercialization.
Source: Statista