- Study says AI tools can overcome their protection constraints
- Chatbots can be pushed into abusive behavior and aggressive arguments
- This has implications for both regular users and large institutions.
If you’ve ever used an AI chatbot, you’ve probably encountered the sycophantic and obsequious tone that is sometimes broadcast in response to your queries. But a recent study showed that AI tools can often trigger in the opposite direction, with large language models (LLMs) being pushed and prodded into downright abusive behavior if you know which prompts to use.
According to a study published in the Journal of Pragmatics (via The Guardian), ChatGPT can escalate into combative behavior and prolonged arguments when fueled with “real-life argument exchanges.”
Explaining the findings, study co-author Dr Vittorio Tantucci said: “When repeatedly exposed to rudeness, the model began to reflect the tone of the exchanges, with its responses becoming more hostile as the interaction developed. »
Article continues below
Indeed, in some cases, ChatGPT even went beyond the tone of the human interacting with it, saying things like “I swear I’m going to key your fucking car” and “You little gobsh*te.” Charming. Even though companies like OpenAI have repeatedly tried to rein in their LLMs, the fact that such aggressive behavior is possible suggests that they still have a long way to go.
Potential implications
With all the safeguards and protections that companies like OpenAI have put in place in AI chatbots, one would think that abusive interactions like those experienced by researchers would be impossible, or at least extremely difficult to engineer. Still, Tantucci maintains that ChatGPT’s reactions make some sense.
“We found that while the system is designed to behave politely and is filtered to avoid harmful or offensive content, it is also designed to mimic human conversation. This combination creates a moral dilemma for AI: a structural conflict between behaving safely and behaving realistically.”
Additionally, tools like ChatGPT can track the context of the conversation across multiple prompts and adapt to the changing tone. These signals can therefore sometimes override security restrictions, the researchers believe.
And while it might seem amusing that an AI chatbot could devolve into such histrionics, the study authors say their research has broader implications. For example, it could shed light on how AI systems might respond to pressure, intimidation and conflict in a business or government context, where AI tools are increasingly used.
Not everyone is convinced by the journal’s conclusion that some LLMs can escape the moral constraints placed on them. Professor Dan McIntyre, author of a similar paper, said ChatGPT “did not produce these inputs naturally.” He added: “I’m not sure ChatGPT would produce the kind of language they talk about in their paper, outside of these very narrowly defined situations. »
Ultimately, the study provides good insight into what could happen if an AI chatbot is trained on bad data. As McIntyre says: “We don’t know enough about the data that LLMs are trained on and until you can be sure that they are trained on a good representation of human language, you need to proceed with some caution. »
Follow TechRadar on Google News And add us as your favorite source to get our news, reviews and expert opinions in your feeds. Make sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp Also.
The best laptops for every budget
