US EDU

.: All U.S. K-12 educators to get free access to Khanmigo as Microsoft partners with Khan Academy​

Summary ➜ This collaboration also aims to use Microsoft’s Phi-3 models to improve AI-driven mathematics tutoring. The initiative is designed to offer personalised learning experiences and help teachers create more effective educational plans, thereby making teaching more sustainable and enjoyable. Additionally, Khan Academy will integrate more of its content into Microsoft Copilot and Teams for Education to expand access to educational resources.

Why this matters for education ➜ This announcement, from Microsoft’s developer conference, takes my top billing in this week’s issue because the (potential) direct impact on educators (at least in the US).

But after the fanfare, what questions do we have? I’ll go first:

• What do we know about the effectiveness of Khanmigo or any similar agent tutor tools? Has it improved?
• How will educators and students be able to control the experience?
• Is Khanmigo just the next version of Clippy?
• Where’s the space for other tools, and how can students get hands on to build their own?
• Is it me or is there something strange about putting “Teachers are super overworked” at the centre of the rationale for these tools? (more on this below)

Lots to think about here and it will be interesting to hear from Sal Khan in August at this year’s big tech and education conference in Australia.

.: Other News In Brief

​

🧠 Anthropic releases a research paper looking inside the black box of AI

👑 NVIDIA Shows Once Again Who is the Real King of Generative AI

🔍 EU’s ChatGPT taskforce offers first look at detangling the AI chatbot’s privacy compliance

🧑‍🤝‍🧑 Meta’s new AI council is composed entirely of white men

❌ Google Search’s “udm=14” trick lets you kill AI search for good

🎵 Spotify experiments with an AI DJ that speaks Spanish

🚀 NVIDIA says 20,000 GenAI startups are now building on its platform

💰 Elon Musk’s xAI raises $6 billion to fund its race against ChatGPT and all the rest

💵 Amazon is Considering $20 Monthly Subscription for GenAI Enhanced Alexa

📰 OpenAI partners with Wall Street Journal publisher News Corp.

​

What’s on my mind?

.: A Productivity Paradox

​

How is the great fanfare, hype, and clamour around AI tools in education a distraction from the working conditions in our education systems?

I may not be able to answer this question from every angle, but I have a hunch it’s the right query.

My unease stems from the recent surge in big tech announcements and the relentless marketing spin touting “productivity” gains for educators using these AI tools and systems.

Long-time readers will know that I have been highlighting the real workload challenges in our schools for some time. But something seems off-balance when Sal Khan, the founder of Khan Academy, says while introducing Khanmigo,

Is that the foundation stone from which we are meant to launch into a new era of super-productivity and super-creativity?! Super?!

This feels like a distraction from the rotten conditions that created the unreasonable work expectations in the first place.

We all need a lift, and technology like AI gives us a faster and more efficient way to complete our tasks. Don’t misinterpret my sentiment here—go for it and get stuck in! Explore these AI tools, try them out, and see how they work for you.

Take the draft lesson plans, get the report comments, and adapt the mountain of emails you need to send.

Just promise me once you have gleefully hacked and slashed your way through your to-do list and tamed your inbox, you will pause.

Pause and ask this simple question about your productivity:

~ Tom

Prompts

.: Refine your promptcraft

​

Another example of the recommended prompts shared by Anthropic, the research lab behind the Claude family of AI models.

I have been writing lesson plans for a major AI Literacy programme recently and it got me thinking about what it was like to learn how to write lesson plans as a young student teacher.

There is something powerful and clarifying in the process of planning a lesson. Its a complex orchestration of concepts and kids. Learning (or lesson) design is a special type of skill and mindset.

With all that said it is fascinating to see AI systems predict their way to an average lesson plans.

Here is the Lesson Planner prompt from Anthropic.

Edit the subject information to anything you want and adapt the task prompt too. Note the promotion of active learning and critical thinking. Here are some examples of how different AI systems complete the task.

Part of the response from OpenAI’s GPT-4o

Here we have a Think Pair Share activity as an engager and a transition into direct instruction from OpenAI’s GPT-4o.

Part of the response from Gemini via Workspace

Gemini needs to work on what a Hook is intended to do! I don’t think this is enough for Year 7s!

Worth noting here that these models have no understanding of the concept of designing a hook or engagement activity for a lesson. It just predicts the next most likely word.

Anthropic’s older Claude-2-100k model suggested a 3-2-1 exit ticket strategy for finishing the lesson. Most of the models seemed to like the exit ticket idea.

Part of the response from Claude-2-100k

My recommendation with any of these AI systems for learning design is

1. Take responsibility.
2. Refine the prompts to suit your context, community and class.
3. Amplify the pedagogical approach you want.
4. Generate LOADS of examples and push the tools to do weird things. Filter for ideas you can build around.
5. Break down the learning design process (Plan) into smaller chunks.
6. Build a bot which can do all of these things, so you don’t have to keep prompting.

:. .:

Remember to mak​e this your own, try different language models and evaluate the completions.

Learning

.: Boost your AI Literacy

​

INTRO

.: Explained: Generative AI​

The MIT News article provides a comprehensive introduction to generative AI, explaining its definition, mechanisms like Markov chains, GANs, diffusion models, and transformers, and diverse applications such as text and image generation. It also addresses challenges like bias, plagiarism, and worker displacement, offering insights into both the technology and its implications.

​

ASSESSMENT ​
​.: Do Students Want AI Writing Feedback?​

In a recent experiment, Mike Kentz tested AI-generated feedback on student essays alongside his own feedback. Students found AI feedback clear and useful but preferred personalised feedback from their teacher. The AI tool, while efficient, often provided generic suggestions and struggled with nuanced critique. The study highlights the balance needed between AI’s efficiency and the irreplaceable human touch in education, suggesting that grading should focus more on the writing process than just the final output.

​

COURSE ​
​.: TAFE NSW | Introduction to Artificial Intelligence​

A free beginner-friendly course from CSIRO and TAFE NSW in Australia, that covers real-world applications and terminology without needing prior programming knowledge. The course offers insights from industry experts and covers topics like machine learning and natural language processing in a 2.5-hour online, self-paced format.

Ethics

.: Provocations for Balance
​

Scenario 1: A New Hope

In a dystopian future, an entire generation has learned math solely from tools like Khanmigo. Though they test well, they are mathematical zombies, unable to innovate or problem-solve without an AI feeding them steps. Society stagnates as complex challenges like climate change and disease outbreaks fester, untackled by minds trained in rigid algorithmic thinking. Once hailed as a boon to education, AI systems are now seen as an intellectual prison. An underground resistance of human maths teachers arises, working in hidden analogue classrooms to cultivate the creative mathematical spark the world desperately needs. But they are hunted by the Algorithm Enforcement Agency, which seeks to stamp out any challenge to the AI teaching regime…

​

Scenario 2: Voice Theft

In a world where AI can clone voices with near-perfect accuracy, personal privacy becomes a relic of the past. Celebrities and ordinary citizens find their voices used in unauthorised ways—advertisements, political speeches, and even criminal activities. The boundaries of consent are blurred, as anyone’s voice can be synthesised and manipulated without their knowledge or approval. This leads to widespread paranoia and identity crises, as people can no longer trust what they hear. The lines between reality and AI-generated deception become indistinguishable, causing a breakdown in social trust and personal security.

​

Scenario 3: The Lonely Generation

Fast-forward a decade. Universities without professors are now the standard, with ‘mega-courses’ of thousands of students ‘taught’ by AI. Overworked faculty have been reduced to ‘AI wranglers,’ providing prompts to lifeless algorithms. The college experience has transformed into a bleak imitation, with students never interacting with a human instructor. But the most disheartening are those oblivious to the difference, the Zoomers who’ve never experienced learning as anything other than staring at a screen. The ‘Lonely Generation,’ raised by algorithms, grapple with forming human connections. A sombre new academic department emerges: ‘Crisis Counselling and AI Addiction Recovery’”

.: :.

Inspired by some of the topics this week and dialled up.

How would you rate this issue of Promptcraft?

(Click on your choice below)
​

​⭐️⭐️⭐️⭐️⭐️ Loved it!
​​
​⭐️⭐️⭐️ Good
​​
​⭐️ Poor​

​

If you have any other kind, specific and helpful feedback, please reply to this email or contact me at tom@dialogiclearning.com​

​
.: Tom Barrett