How many times is 'r' in strawberry?

👋 human,

Happy Tuesday. There have been some great blogs that have attempted to synthesise a lot of the research out there already on AI and student learning. There is also a new study around the use of AI Tutors versus lectures that look interesting too. Let’s dig in

📚 Knowledge builders

• ChatGPT on student engagement → Bradley Busch from InnerDrive has done a literature review on what is currently out there with regards to research on AI and student learning. Naturally, this is all mixed. Personally, I am sceptical of letting students loose on AI, particularly primary aged. That said…

• AI Tutoring Outperforms Active Learning → In this randomised controlled trial, those students who were assigned to an AI Tutor outperformed those with a lecturer, even after controlling for prior knowledge. Not only that but those students with the Chatbot reported improved levels in the following areas:

  • Engagement

  • Motivation

  • Enjoyment

  • Growth mindset

  The crux of this is that those students were Harvard undergraduates. I would be sceptical as to whether younger students would reap these same benefits.

🤖 Industry updates

• OpenAI releases o1 → Codenamed ‘Strawberry’ due to its ability to answer the question that has tricked AI chatbots, it is an update to the ‘omni’ model. This release promises greater reasoning abilities. Currently, it is not available to free users, but there are plans for a gradual release of this later on in across the year. The link above will let you see what some people with early access are able to do with it. What’s interesting is that they have released some prompt guidance that is quite different to prompting other models.

• Keep prompts simple and direct: The models excel at understanding and responding to brief, clear instructions without the need for extensive guidance.

• Avoid chain-of-thought prompts: Since these models perform reasoning internally, prompting them to "think step by step" or "explain your reasoning" is unnecessary.

• Use delimiters for clarity: Use delimiters like triple quotation marks, XML tags, or section titles to clearly indicate distinct parts of the input, helping the model interpret different sections appropriately.

• Limit additional context in retrieval-augmented generation (RAG): When providing additional context or documents, include only the most relevant information to prevent the model from overcomplicating its response.

✨ Fresh prompts

1. Reference this. → Born of my own frustration of having to Harvard reference articles that I have read for work and presentations, this prompt will help you automate that process. Once you have all your references, you can ask them to put them in the correct order.

I need you to format the following media sources into Harvard-style references. Make sure to include the correct format for each type (e.g., books, journal articles, websites, videos). Follow these instructions:

For books: Author(s) (Year) Title of the Book, Edition (if applicable), Publisher, Place of publication.
For journal articles: Author(s) (Year) 'Title of the article', Journal Name, Volume number (Issue number), Page range.
For websites: Author(s) or Organization (Year) Title of webpage, Available at: URL (Accessed: Date).
For videos: Title (Year) Video type (e.g., YouTube, Online video), Available at: URL (Accessed: Date).

Here is the [type of media] source.

Here is an example of the output:

As ever, thanks for reading and keep on prompting! Mr A 🦾