How can the Oxford Tutorial system be empowered by LLMs for AI mentoring / coaching ?

Background

Marc Andreessen, in his? Techno-Optimist Manifesto, discusses the abundance and personalization that AI-assisted technologies will bring, transforming both education and skills. In my work and teaching, we directly encounter these challenges and transformations, particularly in our efforts to rapidly reskill students for the age of AI - both at the #universityofOxford and Erdos Research

The scope of our work involves the idea of? the low-code data scientist i.e. democratising data science by empowering domain experts to perform the functions of data science assisted by LLMs.?

Specifically, this breaks down to five objectives for a domain expert working with AI

• You want to use AI as an assistant to develop code
• You want to work closely with developers
• You want to use l;ow-coce LLM assisted Tools
• Understanding of Mlops
• Understanding the direction of where AI is going

This is a long term goal and we have been using various tools and techniques from OpenAI in this objective. We have had some learning experience and success -?

Magnus Smarason who is a firefighter from Iceland who went on to use AI techniques in his work after the course?

and Dr. Amit Chougule DPM,MD,MRCPsych who is an NHS doctor who shared his experiences of learning AI from us?

we also used the low code data science methodology in a Hackathon in a hybrid mode ?at the #universityofoxford

More broadly, we have also been developing our own methodology / pedagogy for #lowcode #datascience. See William Page ? comments about our low code course

I have covered aspects of this before?

A test harness for #genAI generated code: Using the automated output refinement pattern to enhance and evaluate models?

How can you use #knowledgegraphs and #chatGPT for #AI #reasoning?

Using Knowledge Graphs to create dynamic learning paths for neurodiverse learning using Neo4j with David Stevens and Philip Rathle ?

We are also grateful for the insights and support from the UK Govt CTO

David Knott see more about his keynote at the Oxford AI summit

Ultimately, we see that these functions would be #agentic workflows (autonomous AI agents). Our colleague John Alexander is working on agentic workflows for #langchain and #llamaindex?see the secret agent pdf for an overview

We are also looking at more broader prompting strategies ex The Prompt Report: A Systematic Survey of Prompting Techniques

With this background, I wanted to share the deeper intellectual foundations behind my teaching and thinking for Oxford and Erdos.

The lack of Mentored learning

Considering the idea of abundance and personalization through AI,? we can consider two forms of learning:

Mentored Learning offers a personalized, one-on-one instruction approach tailored to the individual needs of each learner. It provides flexibility and customized guidance, although it typically involves limited interaction with peers.

Cohort-Based Learning, on the other hand, emphasizes group learning, fostering a sense of community and collaboration. It follows a structured schedule with shared progression among learners, but it might be less flexible and adaptable to individual needs.

Today, we do not see mentored learning because it is very expensive to execute. So, we see only cohort based learning. But mentored learning was a characteristic of older / more traditional system like the Oxford tutorial system and the ancient Indian Gurukul education system?

AI assisted learning could allow us to return to a more traditional mentor based learning which is mostly not seen today . Specifically, we are working on enhancing the Oxford tutorial system with LLMs.

The Oxford tutorial system

The Oxford and Cambridge educational system is based on a specific approach called the Oxford tutorial system. You can see more about it HERE and HERE? and the book Oxford Tutorial system The Oxford Tutorial: Thanks, You Taught Me How to Think

The Tutorial system is characterised by?

• Small Group or Individual Learning
• Regular Meetings
• Independent Study and Preparation
• Essay or Problem-Based Discussions
• Flexibility and Adaptability
• An emphasis in liberal education

But more importantly is the intellectual aspects:

Active Participation and Debate: Tutorials are highly interactive, with students actively participating in discussions, debates, and critical analysis of the topics at hand. This interaction encourages the development of strong verbal communication and argumentation skills.

Personalized Feedback: Tutors provide detailed, personalized feedback on students' work, offering guidance on how to improve their understanding, analytical skills, and academic writing or problem-solving techniques.

Intellectually demanding: The system is designed to be intellectually demanding, challenging students to think deeply and critically about their subjects. This rigor helps students to develop a strong mastery of their disciplines.

The overarching goal of the Oxford tutorial system is to develop independent thinkers who are capable of conducting self-directed research, critical analysis, and articulate argumentation.

Conclusion

We believe that by leveraging the capabilities of LLMs with more traditional methods like the Oxford Tutorial System, we can create a new form of education / pedagogy which will empower both students and teachers. This will be a form of personalized mentoring assisted by LLMs

If you are working on similar ideas, please message me. A lot is at stake here. I believe that over time, AI teachers could provide competitive advantages to countries

Its a matter of improving the tools but keeping the mission the same

A few years ago, in my essay "Language & Coding Creativity", I wrote about how these systems represent a big shift in our relationship with language and creativity. As we keep improving these tools, our mission stays the same: to make them helpful, safe, easy to use, and available to as many people as possible. We want to help reduce the obstacles that have traditionally kept people from expressing their unique ideas and perspectives.

source Mira Murati via twitter