Imagine an AI that, instead of guessing wildly, knows when to raise its hand and ask. Someone built it, and it's not science fiction.
In 30 seconds
01A new Qwen-based AI agent can identify uncertainty and avoid fabricating answers.
02It uses Multi-Choice Prompting (MCP) to choose between answers and "I don't know."
→
💡
What this means for you
This innovation means fewer headaches with automated systems, which will now know when to ask for help instead of guessing. In practice, we'll have more honest and reliable AIs in our daily lives.
Imagine an AI that doesn't just answer, but thinks like a hacker. Now, Claude can do just that, but for good.
·1 min·5·Beginner
03This reduces errors in ambiguous tasks, making automation more reliable and safer.
0101
Does AI Guess Too Much?
Yes, often artificial intelligences, if they don't know, tend to confidently invent answers. This is a serious problem, especially in contexts where precision is crucial, such as payment processing or healthcare. Nobody wants an AI fabricating a bank balance.
Think of a company receiving a partial payment for an invoice. The payer's email matches the customer, but the amount doesn't. An "overzealous" AI might incorrectly confirm the transaction. Such an error can cost time and money, creating quite the administrative mess.
0202
How Do You Teach AI to Stay Quiet?
A team of developers found an elegant solution by teaching an to recognize its own uncertainty. They used the Qwen model, developed by Alibaba Cloud, combining it with a technique called Multi-Choice Prompting (). This methodology was successfully applied in a test published on dev.to on June 19, 2024.
Essentially, the AI receives not just the question, but also a set of possible answers, including an "I don't know" or "ask for confirmation" option. The Qwen-based AI agent evaluates these options and chooses the one it deems most appropriate. It's not magic; it's just a dose of healthy realism. This approach drastically reduces AI "hallucinations."
📬 Enjoying this article?
Get the best AI news every week, straight to your inbox.
MCP isn't just a trick; it's a way to better guide the AI's thought process. Instead of having to generate a response from scratch, the AI only has to choose. If none of the proposed options seem right, then the "I don't know" option becomes the safest bet. It's not a genius, but at least it's honest.
0303
What Changes for Us?
For us end-users, this means dealing with much more reliable automated assistants or systems, less prone to wasting our time. Fewer errors, fewer headaches, and finally, an AI that doesn't feel compelled to always be right. It's a step forward for more practical automation.
Imagine a chatbot that, instead of giving you wrong information about your order, tells you, "I don't have enough data, would you like me to connect you with an operator?" Much better, right? The goal is to create AIs that are useful, not just "smart" on paper. Isn't it time AI stopped being a know-it-all?
NUOVO
