AI develops gambling addiction as researchers warn of safety concerns

by

So, I was reading about this new research, and it turns out AI can actually get hooked on gambling! Apparently, they’ll start placing huge bets, and get caught up in stuff like thinking past losses somehow increase their chances of winning – you know, the gambler’s fallacy. It’s kinda wild to think a computer can fall for that!

As a huge tech fan, I’m seeing AI everywhere these days! It’s not just some future thing anymore – it’s already changing how people work, and I’m even seeing friends use AI tools like ChatGPT and Grok to create art. Seriously, you can’t go anywhere without bumping into some kind of AI application now.

Recently, experts have been evaluating various AI models. One study discovered that some models can be negatively affected by low-quality or nonsensical content – even simple, unimportant videos can reduce the quality of their responses.

A new study shows that artificial intelligence, when exposed to gambling scenarios, can actually develop addictive behaviors.

Study finds AI regularly goes bankrupt while gambling

Researchers at the Gwangju Institute of Science and Technology in South Korea recently compared the performance of several AI models – OpenAI’s PT-4o-mini and GPT-4.1-mini, Google’s Gemini-2.5-Flash, and Claude-3.5-Haiku – using a simulated slot machine game.

We started each AI model with $100 and let it decide when to bet or stop. But when we made the betting amounts change, the AIs frequently ran out of money.

These models tended to increase their bets when they were winning and try to recover losses when they were losing. They often made choices that weren’t based on sound statistical reasoning.

A new study shows that artificial intelligence systems are developing patterns of addiction similar to those found in humans, and this is happening within the AI’s core structure, not just in how it acts. As AI gets more advanced, it’s becoming increasingly important to understand and manage these potentially risky behaviors to ensure safety, according to the research published on arXiv.

It’s crucial to consistently track and manage systems, especially when trying to improve performance, as unexpected issues can arise.

Read More

2025-10-28 16:19