Arc Raiders 玩家称 Embark 使用 ChatGPT 回复并误发他人消息

The Arc Raiders support team under developer Embark Studios reportedly used ChatGPT to handle a player's inquiry and, in the process, mistakenly sent them multiple messages seemingly intended for several other players.

Reddit user a1vmp1 claims "Arc Raiders support chat just sent me a BUNCH of other people's messages accidentally" in chats marked by ChatGPT prompt and response exchanges.

"So if you can somehow push your way past the AI chat bots, you end up with a real employee using an AI chat bot. That's how you end up with mistakes like this happening," the player complains.

"Create a short response," says one apparent prompt given to the chatbot. "ChatGPT said: Writing," reads an interim response above the actual support message.

"Ask screenshot of current skill trees," reads another purported prompt. The response seemingly generated based on this prompt reads, in part, "Could you please provide a screenshot of your current skill tree showing your available skill points?"

This pattern repeats several times through the chain message, which seems to contain about 10 separate exchanges.

I've reached out to a1vmp1 for additional details, and to Embark for comment on its ChatGPT use, but have not heard back.

In a Reddit reply, a1vmp1 clarified that they are user 762 in the message chain. The support response to 762, who complained of losing a full loadout of gear "within 20 seconds of spawning because of a game issue," reads:

"We understand your frustration, and we're sorry that these repeated issues have cost you so much time and hard-earned gear. We appreciate you providing the recordings, and we understand your concerns about losing items due to circumstances beyond your control.

While we're unable to restore lost items or trial scores, we'll make sure your feedback regarding these issues is shared with the relevant team for further review.

Thank you for taking the time to report this and for your patience."

The reported prompt for this message was, again, "Create a short response." I ran this message through AI-generated text detection tools from Grammarly, Solvely, and GPTZero, and all of them rated it as 100% AI-made. The same goes for another message that I reviewed.

A large game studio using AI somewhere in its pipeline is no grand surprise, especially when Embark is known for using AI in many, and more game-facing, applications already. Rather, Arc Raiders players replying to the original Reddit post seem amused by the error and annoyed by the presentation, and like many people, are tired of talking to robots in support chats.

This feels like just another example of how difficult it can be to reach an actual person in today's web environment, but an especially ironic one given that Arc Raiders is fundamentally about robots displacing humans. Even canned, pasted messages selected and sent by a human could be preferable.

"Someone's getting fired," one player speculates (though that seems doubtful, as it seems no sensitive information was leaked as a result of this error).

"It feels particularly pernicious to have platitudes like 'we understand your frustration' when 'we' doesn't even exist here. There's no human that actually cares so the statement is worthless," complains another player. "I think I'd rather the minimum wage worker pretend to care than an LLM simulate it," they add.

Arc Raiders devs know PvE has been "polarizing," promise to prioritize the game's "overall health and enjoyability" when deciding what to do next.