ChatGPT, Mindmaps, and Intelligent Tutoring Systems

Поділитися
Вставка
  • Опубліковано 7 лип 2024
  • This video reviews the academic literature on Intelligent Tutoring Systems. Based on that review we talk abstractly and practically about what a learning system could look like based on that literature.
    Timestamps
    ==================
    00:56 Part 1: Academic Literature Review
    1:42 ACT-R
    4:11 Knowledge Tracing
    6:26 Constructivism
    8:00 Recommender Systems
    9:37 Pedagogical Strategies
    11:48 The Singularity
    12:45 Part 2: Practical Takeaways
    18:17 Outro
    Footnotes:
    In the intro the Cognitive Scientist is Tenembaum, the ML researcher is Schmidhuber, and the Software Engineer is ThePrimeagen.
  • Наука та технологія

КОМЕНТАРІ • 24

  • @anthony.s6889
    @anthony.s6889 24 дні тому +2

    man every time I watch your video it makes me wanna get an ipad to try your app out. Windows user like me can only stick to Concepts

  • @Mmj_lh
    @Mmj_lh 26 днів тому +11

    When will Ahmni be available for Android users?

    • @AhmniApp
      @AhmniApp  24 дні тому +1

      Mac is the next planned platform, hopefully multiplatform eventually! But currently there is no planned release date for android.

  • @1nemorevideo673
    @1nemorevideo673 25 днів тому

    Dope vid! Will try out the app as I progress through Ics

  • @ServantUmoh
    @ServantUmoh 7 годин тому

    Hey great video! I just used ChatGPT to create a list of keywords based on a reading that I was assigned so I can go about researching then creating my mind maps, do you think that is a good use of it or should I be sifting through the information more so myself and coming up with my own list of keywords? Thank you again for this information, I love hearing great minds and I hope to hear from you soon!

  • @ash9259
    @ash9259 24 дні тому

    don't usually comment very much but this is stellar!
    i'm already expediting my learning process through anthropic claude by using the system of question answering you described.
    naturally, claude tests my mental model against current scientific thought.
    annoylingly, science often contradicts itself on the internet since most of the juicy knowledge of course is behind paywalls.
    it would be great if paywall jargon was drip-fed to the student and well, essentially, this will enable a generation of young people (including myself) and will of course, supply valuable training data to future AI's so they can observe the patterns in how we weigh the english dictionary or as could also be said, following the money haha.
    i distilled the idea some few days ago that prompt engineering would be the future of innovation in-fact, since A.I. would be doing all our math for us and welll, if you think about it, isn't all of this just math with feelings :^D?

    • @AhmniApp
      @AhmniApp  24 дні тому +1

      > most of the juicy knowledge of course is behind paywalls
      LLM companies are not known for respecting measures to prevent scraping. There is a good chance paywalled papers and books were used in LLM training.
      > that prompt engineering would be the future of innovation
      Certainly crafting the correct inputs to machine learning algorithms has been key for a long time. But if we look outside of Natural Language Processing at the more mature field of computer vision, the necessity of feature engineering inputs has massively decreased over time as models have become more capable and robust. I think this will depend a lot on your definition of "prompt engineering" and interfaces for the specific use cases being discussed.

    • @ash9259
      @ash9259 24 дні тому

      @@AhmniApp >that really makes me quite happy to be honest, open sourcing knowledge is very desirable.
      >i suppose the concept of pure probability or that connotational component of humans that sees the distant future between seemingly unrelated terms.
      i imagine the essentia of creativity is that it is a unbound by time itself and that the only change that will come in society will be found in thinking is absurdly novel ways.
      i imagine most of the education i imagine Ahmni facilitating will be in catching people up to modern archiecture and modern thinking landscapes whilst the true innovators will be people such as yourself who can think of ideas for a superintelligence to try out.

  • @okechukwuanudokem2060
    @okechukwuanudokem2060 15 днів тому

    wow ics alumni dude's name i forgot really showing up on my recommended , if u make it better than squid I'll probably get it , good luck with the app

  • @isaacwu4089
    @isaacwu4089 25 днів тому

    Do you plan on integrating an Intelligent Tutoring System into the Ahmni app, or perhaps optimizing an LLM for tutoring? If so, are there any other practical implications?

    • @AhmniApp
      @AhmniApp  24 дні тому

      It's possible. The point of the video is that self-directed, self-motivated students don't need a full Intelligent Tutoring System, instead they need specific ITS-inspired features to make their lives easier. Ahmni has existing features like that and will continue to get more.
      In terms of optimizing and adding an LLM it could be possible, especially on a desktop version of the app. However I think it's unlikely that a chat interface built into the app is significantly more effective than plain ChatGPT/Claude in the hands of a self-motivated student. The student interacting with the LLM can already make it "optimized for tutoring" by following the best practices laid out in the video, or asking the LLM to follow those best practices.

    • @isaacwu4089
      @isaacwu4089 24 дні тому

      @@AhmniApp Thanks for your response! What about LLM optimization techniques such as prompting it to act in a certain way? Do you have any recommended prompts that you use?

    • @AhmniApp
      @AhmniApp  24 дні тому

      Not really, just talking to it normally works fine for me. My focus is usually on what I'm saying or evaluating what the LLM is saying, rather than trying to get "better" responses out of it. For example sometimes I'll use the dictation feature and talk pretty lengthily at it about my current understanding and try to articulate things I'm unsure about. The LLM will then give me pointers or assess my accuracy, but probably the biggest benefit is forcing myself to verbalize my current understanding and misconceptions.
      Maybe the closest thing to prompt optimization I do is tailor how opinionated I am. On one hand by being opinionated (e.g. "Okay so it seems like the benefit of X is mostly Y.") is great because it forces me to have an opinion and verbalize it. But sometimes I've noticed that LLMs get biased by my opinion and continue the conversation in that direction rather than providing more unbiased feedback/direction. So for pure information discovery I avoid it.
      That's for general inquiry, though. There are some more specific uses I really like:
      - Generating pre-test questions based on a syllabus
      - Generating lists of keywords for a mindmap
      - Asking "what are similar things I'm missing", e.g. "I know about MySQL and Postgres, what else is similar that I'm missing?"
      Do you have any optimization techniques you use?

    • @isaacwu4089
      @isaacwu4089 19 днів тому

      @@AhmniApp Thanks for your response. I generally use optimization techniques only if the 1st response isn't satisfactory. Some I think would be useful are uploading any relevant info (material, etc) and giving it a persona as an expert teacher/tutor. And generally making the prompt clear

  • @Chinekeh
    @Chinekeh 24 дні тому

    Do you have a code for purchasing for lifetime or only yearly?

    • @AhmniApp
      @AhmniApp  24 дні тому

      There's currently the code SUMMIT for 50% off* the first year, but no lifetime code.
      *International pricing may vary

    • @Chinekeh
      @Chinekeh 24 дні тому

      @@AhmniApp Oh cool thank you. I have Concept. This looks more study focused.

    • @AhmniApp
      @AhmniApp  24 дні тому +1

      Definitely. The goal of Ahmni is to be the best app for people who use advanced learning methods.

    • @Chinekeh
      @Chinekeh 24 дні тому

      @@AhmniApp Now and am trying it out. It seems like you have used Justin Sung and you understand it. Hopefully I can finally get it down

  • @knw-seeker6836
    @knw-seeker6836 20 днів тому

    Off topic but do you think teaching has a professional future or teacher job ?
    Because of AI

    • @AhmniApp
      @AhmniApp  20 днів тому +1

      Yah definitely. New education technology is usually branded as "disruptive" and promises to upend the status quo but ends up being used much more subtly to make iterative progress. Teachers aren't going away anytime soon. The best outcome will be AI making teachers slightly more effective or certain disciplines more accessible.
      For a history of all the previous promises of disrupting education and teachers (e.g. MOOCs) see the books:
      - 25 Years of Edtech
      - Failure to Disrupt: Why Technology Alone Can't Transform Education
      Do you think this is something you (or other people) are worried about?

    • @knw-seeker6836
      @knw-seeker6836 20 днів тому

      @@AhmniApp interesting thoughts thanks for sharing
      I did see a video entitled with it and there is a lot of talk if programmers or tech will eventually will be replaced through AI
      Personally I’m not worried it just makes me think
      I want to be able to adapt
      Because I think it will change a lot of different areas