OpenAI has just launched the Socratic "Learning Mode"! Specifically designed for college students, it helps you easily master complex topics like game theory, Bayes' theorem, and discrete mathematics through conversation. As soon as it was launched, users went crazy. Some predict: Another wave of educational AI is about to be disrupted.
Just now, OpenAI welcomed a new feature update, but it's not GPT-5.
ChatGPT has launched a new mode - Study and Learn Mode!
Using this mode, ChatGPT not only provides answers but also guides problem-solving step by step.
Starting today, free, Plus, Professional, and Team versions can all use the learning mode!
In the next few weeks, ChatGPT Edu version will also launch this feature.
As soon as this feature was launched, it was immediately popular. Clearly, traditional educational models are about to be disrupted by AI!
A university professor voluntarily commented, hoping to open a professor link directly to student accounts to make teaching smoother.
Some predict: Another wave of educational AI is about to be disrupted by OpenAI.
OpenAI's "Learning Mode" is here!
This new learning mode created by ChatGPT is a Socratic step-by-step guidance.
Why develop this mode? OpenAI states that ChatGPT is becoming one of the most widely used learning tools globally.
But how can we ensure it truly helps users learn, rather than just feeding solutions?
For this reason, OpenAI's team specifically constructed this learning mode.
The main functions can be summarized as follows.
Interactive Prompts: Socratic questioning to guide active learning.
Step-by-Step Solutions: Information organized into easily understandable parts, highlighting key connections between topics, providing appropriate context, and reducing complexity.
Personalized Support: Adjusting to your level based on assessment and previous memory.
Knowledge Check: Quizzes and open-ended questions, with personalized feedback to track progress.
Flexibility: Learning mode can be turned on or off at any time.
When students enter learning mode, ChatGPT provides guiding questions that adjust answers based on their goals and skill levels to help deepen understanding.
It can help you with homework.
And help you prepare for exams.
The original intention of designing this learning mode was to help college students.
For example, if you want to learn game theory, you can request from ChatGPT: I want to learn game theory, mainly for you to teach me, develop a high-level plan. I'll ask questions when curious, but please ensure I quickly grasp this knowledge.
Then, ChatGPT will quickly provide a teaching outline, including what game theory is, classic game types, prisoner's dilemma classic cases, Bayesian games, etc.
Then, it will guide you step by step, letting you answer questions one by one.
For instance, explain game theory in your own words? Afterward, ChatGPT will lead you step by step into the core, through guided dialogue, helping you quickly master related knowledge.
For example, it can also help you understand Bayes' theorem.
Help CS senior students relearn discrete mathematics.
Such an incisive and patient teaching mode naturally made the ChatGPT learning mode popular among many college students upon launch.
Here are some student feedbacks.
This is a real-time, 24/7 all-knowing Q&A time.
The learning mode effectively breaks down dense materials into clear, well-paced explanations.
I tested the learning mode to tutor a concept I had tried many times before: sine position encoding.
It's like a tutor who never tires of my questions. After 3 hours of learning, I finally understood thoroughly and felt confident.
Prompt Secret Discovered by Experts
The prompt behind this learning mode was quickly discovered by Simon Willison, co-creator of Django.
He found it's a carefully designed system prompt, and OpenAI hasn't yet taken measures to protect these prompts, so he easily extracted them.
The Chinese version looks like this.
Thanks to OpenAI's generosity, reproducing the AI Socrates now seems quite easy.
ChatGPT Transforms into a Teacher
After testing this feature, for the first time, I felt AI is so "warm"!
In "learning mode", ChatGPT won't directly answer questions but first ask about our educational background.
To understand the questioner's basic level.
When given a specific "question", ChatGPT will progressively inquire about our subject background and how we understand the topic.
For example, when the question is "What is quantum mechanics", ChatGPT will ask if you want to understand "entry-level" knowledge or "underlying principles".
Very much like a teacher-student dialogue.
Moreover, in "learning mode", ChatGPT truly becomes "Socrates", conducting progressive teaching.
Even for the same question, ChatGPT will make targeted responses based on different educational backgrounds chosen.
In this way, ChatGPT becomes "highly personalized," even with a sense of "teaching without discrimination".
At this point in the large model competition, simply answering questions seems to have reached a "bottleneck".
Making AI respond in a "tailored teaching" manner appears to be the new competitive direction.
Not only OpenAI, but Google and Anthropic have also recently begun testing "Socratic" guided questioning functions.
In the face of cold AI ranking data, this warm educational function seems to be the AI we hope for.
References:
https://openai.com/index/chatgpt-study-mode/
https://x.com/OpenAI/status/1950240351547248941
This article is from the WeChat public account "New Intelligence", author: New Intelligence, published with authorization by 36kr.
