Friday, August 15, 2025

Answer: Best ways to do just-in-time learning. Part 1.

  What's a great way to... 

A woman studying a complex anatomy topic. P/C Pexels.com by RF._.studio _

... learn something relatively quickly?    

As I said in the Challenge last week, in my line of work, I often have to be a fast learner and become an instant expert (or as close as possible within the time I've got).  

This motivates the Challenge: What are some new (largely AI-driven) ways to rapidly learn something?  

So here was the Challenge: 

1. What are the best AI-inspired (or AI-driven) new Micro- and Meso-learning that you've found?  Have you actually tried this method?  How well has it worked out for you? 

There are SO many ways to answer this Challenge that I'm going to give it in parts.  Here's Part 1.  I'll give some more ideas next week (instead of another Challenge).

 I AM mindful that real learning takes work.  As Andrej Karpathy wrote (Feb 10, 2024): 

Learning is not supposed to be fun. It doesn't have to be actively not fun either, but the primary feeling should be that of effort. It should look a lot less like that "10 minute full body" workout from your local digital media creator and a lot more like a serious session at the gym. You want the mental equivalent of sweating. It's not that the quickie doesn't do anything, it's just that it is wildly suboptimal if you actually care to learn.

That's all true.  But there are times when we need to learn something rapidly.  How can we do that?  I'm happy to sweat and do focused work, but sometimes I need a micro-learning or meso-learning experience.  How can we do that? 

Here's my approach.  (Next week we'll talk about other approaches.)  

1. Identify what you're trying to learn. Does it have a name?  Do you have a fairly clear learning goal? Just flailing around watching short videos isn't the same as having a plan.  This is often the hardest part of the learning task--especially if you don't know what you're really trying to learn.  As a way to clarify the task... 

2. Write down your goal. What I do is to write down (usually in pen on paper!) what I'm trying to learn about.  

Example: Last week, after meeting a kelp researcher on the beach in northern California, I wanted to learn about the different kinds of kelp that grow locally. I literally wrote out, "Learn: kinds of kelp in N. California. How many varieties? Ecology. Threats. Current condition."  

(That's Russell-ese for "Learning Goal: To learn about the different kinds of kelp that grow off-shore in California. Learn about the different species of kelp, what kinds of ecosystems do they grow in, what are the current threats to their growth, what's the current condition of kelp beds in 2025?")

In essence, I'm outlining a topic and a few relevant sub-topics. In the past, I would have started classical searching, and pulled together a few articles to read, then spent a happy couple of hours reading them, digging deeper as interest led me.  

Here's my first search.  Note that I've opened the AI Overview as a way to get a first-glance at the topic:   


This quick search tells me a lot. There seem to be only 2 species of kelp along California, NOAA is doing research, there is an effort in kelp forest restoration because of multiple threats.  

3. Do a quick search for pre-existing edu materials.  It's worth doing this first--you never know when someone out there will have already made exactly the right tutorial for you.  No matter how obscure the topic, someone might have already put together a package for you.  As my friend Leigh says, "Dan, it's the internet... there's always someone who is interested in that."  Time has proven him right. 

Point is: Check for pre-existing tutorials (or lectures, or lessons, or lectures) first. 


Looking at the videos, it seems there are a couple of potentially useful ones already.  

To scan a video like this, I'll typically open the video and look at the transcript. (Sometimes the link to the transcript is a little buried. See below for a useful tip.)  


Then, once that panel is open, you can click to open the transcript. 


Then, once that panel is open, you can click to open the transcript. Click on the magnifying glass to search in the transcript.  (Or use Control-F / CMD-F.) 

I then do a few searches inside of the text to make sure that the video really is on-topic and discusses what I want.  (For instance, does the word "ecosystem" appear in the video?)   



Alternatively, you can ask Gemini to summarize the video (again, to see if it's really what you want to learn.. keep in mind that you want to watch the whole video, not just read the summary).  Looking at the summary below, it seems like a great video to watch--it's nearly an hour long, and probably worth the investment of time.  


When you've scrolled down a bit, you might well find the "People Also Ask" section... which is another way to see what relevant topics you might learn.    




Of course educational materials can be more than just YouTube videos.  Using the term "lesson" is often useful, though I've found that they're often oriented for 6-12 grades.  (Which might be what you're looking for.)  

For more advanced learning units, I'd use the terms "university" and "class."  That will often find entire semester-long classes, with a syllabus and everything.  

4. Create a quiz to help test your knowledge. As we know, self-tests are a very useful way to ensure that your studying has paid off.  In a great tip from Ben Gomes at Google, he suggested using Gemini to make a quiz to check.  You can do this with either a video OR a longer text document.  

One of the documents about kelp I read was this one from California State Fish and Game - you can see what I asked Gemini to make a quiz for me (look on the left hand side... I said "Here's a document... <URL> ... create a study quiz from this file." 


You can do the same thing with videos.  Here, I dropped in a link to a YouTube video about kelp.  





SearchResearch Lessons


This is Part 1 of our "how to learn rapidly" series.  We learned that: 

1. Identify your task. Use this as a way to delimit what you'd like to learn.  

2. Write down your learning goals. A great way to do this is to literally write down what you'd like to learn.  As you write, you'll learn what you don't know--identifying knowledge gaps is a great method to figure out how to proceed. 

3. Search for already existing learning materials on your topic. Don't skip this step!  

4. Create self-test quizzes to ensure that you've learned the material. 


Keep searching!  

8 comments:

  1. Thank you, Dr. Russell.

    In the AI Overview, you verify or do something or is just a starting point. I ask because this week the Oreo case (still today with bad answer) was very popular.

    Also I was thinking; LLMs should search and find information from all the sources right? Or they just take some parts?

    This week I was reading about which movie was watching a President of the United States in an specific date. The person that mentioned that says LLMs gave wrong answers. Only Grok was right because a tweet from the past of that person.

    Search Lesson 3 could be like your famous [ list of...] search tip right?

    I mean: "Point is: Check for pre-existing tutorials (or lectures, or lessons) first." Should work. I like it. Is there a more specific moment to try each of the mentioned words?

    ReplyDelete

  2. trying to have some fun while quick learning... I lost my pen & pad...
    https://tinyurl.com/33s5kvh5
    https://farallones.noaa.gov/eco/kelp/decline.html
    https://sanctuaries.noaa.gov/news/jan20/kelp-forest-restoration-greater-farallones-national-marine-sanctuary.html
    https://en.wikipedia.org/wiki/Strongylocentrotus_purpuratus
    similar to Northern CA, but these kelp forests are near Tibet... ;^P
    https://i.imgur.com/SJFVLL0.jpeg
    also stressed in the Himalayas - starting to petrify.
    https://i.imgur.com/RsRHNGg.jpeg
    advanced die-off, not sure if the Purple Sea Urchin did that...
    https://i.imgur.com/lArsbfu.jpeg
    Cali Purple -
    https://i.imgur.com/hlOnFGl.jpeg

    ReplyDelete
  3. need more otters?
    https://i.imgur.com/teBVUdv.jpeg
    ... more AI-Otter-nese?
    https://i.imgur.com/1Imp0NN.jpeg
    "Urchin barrens"
    "Purple sea urchins have several natural predators, including sea otters, sunflower stars, and California sheephead. Sea otter predation on purple sea urchins helps to maintain healthy kelp forests by preventing overgrazing by the urchins, according to the Monterey Bay Aquarium. Humans also consume purple sea urchins, and in some areas, they are harvested for food or scientific research. "
    https://royalsocietypublishing.org/doi/10.1098/rspb.2024.1539
    https://farallones.noaa.gov/eco/kelp/decline.html
    https://www.nature.org/en-us/magazine/magazine-articles/kelp-forest/
    Kelp forests are ''carbon vaults'...
    ". Through ongoing surveys using satellite, drone and piloted aircraft imagery, TNC has found that in the past 10 years, 96% of kelp forests in the region have disappeared. It’s an ecological disaster: From San Francisco to Oregon, nearly all that remains of one of the most productive ecosystems on Earth is an underwater wasteland" (written in 2023).
    ... not to mention the Himalaya Kelp Forests - down by 99.87%.

    https://www.ourbreathingplanet.com/farallon-islands/amp/?gad_source=1&gad_campaignid=1602607328&gbraid=0AAAAADL0VgBE9Ps8pW8rIDnk1lE7YY-Ju&gclid=Cj0KCQjw2IDFBhDCARIsABDKOJ6JX-t9C51oVVcPoYDCAH7Zjw3zK-FJKVv92NoHvbwAQFnxZ8pbsNoaAhObEALw_wcB

    ReplyDelete
  4. Kelp on Everest & K2...
    https://en.wikipedia.org/wiki/K2#:~:text=13%20Bibliography-,Name,its%20widespread%20use%20is%20scant.
    "Yes, the Himalayas were indeed once underwater. The area where the Himalayas now stand was once the seafloor of the Tethys Ocean. The collision of the Indian and Eurasian tectonic plates caused the ocean floor to uplift, eventually forming the mountain range. "
    https://www.youtube.com/shorts/o_xYbEeCKK8
    https://en.wikipedia.org/wiki/Tethys_Ocean
    AI overview...
    "The Tethys Ocean was a large, prehistoric ocean that existed for millions of years between the continents of Laurasia and Gondwana. It's known for its role in shaping the Earth's geography and climate, and for the diverse marine life it once harbored. The Tethys Ocean no longer exists as a continuous body of water, but its remnants can be found in the mountain ranges of Europe and Asia. "

    ReplyDelete
  5. ...as I was noodling about this, it occurred to me that now that "traditional search engines (google, bing, etc.)" have added AI, they are a good place to start still - they offer more options - news, images, videos, books... as quick source references and they are more "real time". So unless you already anticipate a specific task; e.g., image generation, deep research or summaries - Search Engines are a quicker doorway to quick learning?
    https://tinyurl.com/5n6sh7xa
    https://scholar.google.com/scholar?q=quick+learning&hl=en&as_sdt=0,6
    video with search strategies:
    SERP results now include 'key moments' for videos- speedy
    https://www.youtube.com/watch?v=dsBzBvbtGO4
    app:
    https://chromewebstore.google.com/detail/google-scholar-pdf-reader/dahenjhkoodjbpjheillcadbppiidmhp
    https://youtu.be/AshWAiYuhAM?si=2LUUvi15t9fKuMDE
    ...noodling...
    https://i.imgur.com/VRjnIeG.jpeg

    ReplyDelete
  6. a side note - another just published author concerned with "distilled learning".
    https://tinyurl.com/3yfr6ncu

    ReplyDelete
  7. elsewhere (searchResearchAIDevelopment)
    https://plparchitecture.com/wla-artificial-intelligence-lab-shanghai-china/
    looks "chip-like" from overhead...
    https://www.lingang.gov.cn/html/website/lg/English/about/multimedia/1790312146357219329.html
    Lin-gang (a googleplex on steroids - Lingang New Area global R&D district)
    https://en.lingang.gov.cn/html/website/lg/English/about/overview/index.html
    one application:
    https://plparchitecture.com/plp-labs/
    some non-existent views:
    https://i.imgur.com/TDc3zHl.jpeg
    https://i.imgur.com/N696hF0.jpeg

    https://hai.stanford.edu/ai-index/2025-ai-index-report
    https://www.forbes.com/lists/ai50/
    https://www.usnews.com/education/best-global-universities/switzerland/artificial-intelligence

    Gemini AI Overview (do you concur? - you are my 'ground truth source')
    "Switzerland boasts a vibrant AI research landscape with several leading researchers making significant contributions to the field. Here are some prominent names and the areas they specialize in:
    Jürgen Schmidhuber (Dalle Molle Institute for Artificial Intelligence Research): Widely recognized for his work on artificial neural networks, deep learning, and particularly Long Short-Term Memory (LSTM) networks, which have been foundational for breakthroughs in areas like natural language processing.
    Prof. Andreas Krause (ETH Zurich): Focuses on machine learning, reinforcement learning, Bayesian optimization, and robustness and safety in AI systems.
    Prof. Luc Van Gool (ETH Zurich): A leading researcher in computer vision, his work encompasses areas like image enhancement, object recognition, and 3D reconstruction.
    Prof. Luca Benini (ETH Zurich): Specializes in hardware architectures for machine learning, design of efficient AI acceleration engines, and edge AI.
    Prof. Martin Vechev (ETH Zurich): His research centers on safe and reliable artificial intelligence, robust and fair deep learning, and the certification of machine learning models.
    Prof. Boi Faltings (EPFL): His research interests include various aspects of artificial intelligence, and he is affiliated with the EPFL Artificial Intelligence Laboratory.
    Prof. Wulfram Gerstner (EPFL): Focuses on theoretical neuroscience and brain-inspired artificial intelligence, exploring the computational mechanisms of learning and memory.
    Prof. Thomas Hofmann (ETH Zurich): Leads research in areas related to data analytics and machine learning.
    Prof. Andreas Krause (ETH Zurich): His work focuses on machine learning, reinforcement learning, Bayesian optimization, and robustness and safety in AI.
    Leading Swiss Institutions for AI Research
    ETH Zurich - Swiss Federal Institute of Technology: A globally recognized institution with a large and active AI research center, involved in a wide array of AI areas.
    EPFL (École Polytechnique Fédérale de Lausanne): Another top-tier university in Switzerland, known for its cutting-edge AI research and strong industry connections.
    Università della Svizzera italiana (USI): Hosts the Dalle Molle Institute for Artificial Intelligence Research (IDSIA), a leading Swiss AI lab.
    University of Zurich: Offers strong programs and research in AI, including a focus on machine learning and generative AI. "

    checked one...
    https://en.wikipedia.org/wiki/J%C3%BCrgen_Schmidhuber
    https://i.imgur.com/MHDVwAe.jpeg

    ReplyDelete
    Replies

    1. AI, Juergen and coral...
      https://people.idsia.ch/~juergen/
      https://people.idsia.ch/~juergen/blog.html
      https://www.kaust.edu.sa/en/news/adoption-of-ai-to-accelerate-world's-largest-coral-restoration-project
      https://www.instagram.com/p/DLSZayBt3NR/?utm_source=ig_web_copy_link&igsh=MWYzdnk4MTAwcnZ2aA==

      Delete