V2V2 · Centre signalCentre signal

Find a session, its slides and recording, or choose a course.

Search sessions by title, and courses by topic. Try “agents” or “MondAI”.

Recently added

Sessions with their materials

    Training

    Three from the Michaelmas 2026 catalogue

    Consulting that starts with your work

    Explore where AI can help, what is feasible and how to assess the result.

    Discuss your project
    AI Competency Centre, University of Oxford. Design mock-up built from the round-1 content pack; the Resources link returns to the top of this page.
    October 2026 · MondAI roundups

    MondAI Roundup, October 2026

    Presenters: not in the content pack yet

    Scripts blocked (simulated). The slide reader below uses only its HTML and stylesheet: contents links and scrolling still work. The slide counter bar, previous/next buttons and current-slide highlight need scripts, so they are hidden. To see the real thing, disable JavaScript in your browser.
    Slides · first 16 of 155 slides · illustrative sample: BETT 2023 talk

    How to Teach and Learn with ChatGPT: AI and new horizons for assistive technologies

    Dominik Lukes, EduTools.fyi

    Slide 1 of 16
    Contents (16 slides)
    1. 1How to Teach and Learn with ChatGPT: AI and new horizons for assistive technologies
    2. 2Lots of developments since we scheduled the session
    3. 3What has changed
    4. 4bit.ly/academic-ai-newsletter
    5. 5Poll: What is your experience with ChatGPT?
    6. 6Poll: Experience?
    7. 7Comparison from a recent talk
    8. 8Poll: What do you think AI might be best at?
    9. 9What do you think AI might be best at?
    10. 10Comparison with a recent talk
    11. 11What it is actually good at?
    12. 12Nobody expected AI to be any good at this
    13. 13Or this
    14. 14We thought AI could do this
    15. 15But it is also unexpectedly good at this
    16. 16How do we know what modern AI will be good or bad at?

    How to Teach and Learn with ChatGPT: AI and new horizons for assistive technologies

    Dominik Lukes, EduTools.fyi

    View original slide 1

    Lots of developments since we scheduled the session

    A rocket icon beside the slide’s statement.

    View original slide 2

    What has changed

    • Bing Chat
    • GPT-4
    • Google Bard
    • Office ‘Co-Pilot’s
    • GPT-4 Plugins

    View original slide 3

    bit.ly/academic-ai-newsletter

    https://bit.ly/academic-ai-newsletter

    Screenshot of the LinkedIn newsletter “AI in Academic Practice” with a cartoon computer with a puzzled face and question marks, beside a QR code.

    View original slide 4

    Poll: What is your experience with ChatGPT?

    A raised-hand icon beside the poll question.

    View original slide 5

    Poll: Experience?

    • Use it every day
    • Played around with it
    • Heard a LOT about it
    • I heard something about it

    View original slide 6

    Comparison from a recent talk

    A poll result from an earlier talk: a pie chart beside the answers Use ChatGPT in my work (5), Have used once or twice (32), Have heard a lot and seen examples (21), Have heard something but not sure (23) and Other (share in chat) (1).

    View original slide 7

    Poll: What do you think AI might be best at?

    A raised-hand icon beside the poll question.

    View original slide 8

    What do you think AI might be best at?

    • Looking up facts
    • Performing complex calculations
    • Writing computer code based on text
    • Making complex decisions from data
    • Writing poetry
    • Making beautiful art

    View original slide 9

    Comparison with a recent talk

    A horizontal bar chart of poll results for what AI might be best at, listing Looking up facts, Performing complex calculations, Writing computer code based on text, Making complex decisions based on data, Writing poetry and Creating beautiful photos, with a pop-up for “Creating beautiful photos”: 1st choice 2 (2%), 2nd choice 3 (4%), 3rd choice 6 (7%), 4th choice 7 (8%), 5th choice 18 (21%), 6th choice 48 (57%).

    View original slide 10

    What it is actually good at?

    A superhero-flying icon beside the question.

    View original slide 11

    Nobody expected AI to be any good at this

    Two ChatGPT conversations side by side. Left: the prompt asks for a poem to remember the bones in the hand and the answer is a four-stanza rhyming poem. Right: the prompt asks for a mnemonic for the carpals and the answer is the sentence “She Left The Party To Take Cathy Home” with each carpal bone matched to a word.

    Prompt (left): Write a poem to help me remember the bones in the hand.

    In the realm of hands, a tale unfolds,
    Of bones and joints, a story told,
    Listen close, my friend, and hear,
    The names of bones, so you'll endear.

    Twenty-seven bones, a framework grand,
    Form the structure of our hand,
    Three groups of bones to memorize,
    Carpals, metacarpals, phalanges arise.

    Begin with carpals, eight in sum,
    In the wrist, they're not cumbersome,
    Proximal row, four in row,
    From lateral to medial, here's how they go:

    Scaphoid, lunate, triquetrum align,
    Pisiform, a small pea, so benign,
    Distal row, to continue the tour,
    Trapezium, trapezoid, capitate, hamate ensure.

    Prompt (right): Make a mnemonic for the carpals

    Recall the carpals, eight in all,
    With this mnemonic, you'll never stall:

    She Left The Party
    To Take Cathy Home

    • Scaphoid - She
    • Lunate - Left
    • Triquetrum - The
    • Pisiform - Party
    • Trapezium - To
    • Trapezoid - Take
    • Capitate - Cathy
    • Hamate - Home

    View original slide 12

    Or this

    A collage of AI-generated pictures: a man sitting beside a hole in a wall, colourful retro robot and computer characters, a smiling person with a tree of leaves for hair, and a robot reading in a hollow.

    View original slide 13

    We thought AI could do this

    A ChatGPT conversation about the arithmetic “4^2 - 234 * (-23) * 329870134987”: it works through the steps and its answer reads 10825535722618; beside it a calculator shows 1775361066500050.

    View original slide 14

    But it is also unexpectedly good at this

    A ChatGPT conversation. The prompt is “Write code to add two numbers in python”. The answer explains the + operator and gives a short Python program with an add_numbers function, followed by an explanation of the code.

    Prompt: Write code to add two numbers in python

    In Python, you can add two numbers by simply using the + operator. Here's a simple function to add two numbers:

    def add_numbers(a, b):

        return a + b

    number1 = float(input("Enter the first number: "))

    number2 = float(input("Enter the second number: "))

    result = add_numbers(number1, number2)

    print("The sum of the two numbers is:", result)

    In this code, the add_numbers function takes two arguments, a and b, and returns their sum. The user is prompted to enter two numbers, which are then passed to the function to perform the addition. The result is then printed.

    View original slide 15

    How do we know what modern AI will be good or bad at?

    A puzzle-piece icon beside the question.

    View original slide 16

    Component sheet

    Centre signal: style

    Navy carries structure, pink marks the active state and the one accent per area. Barlow Semi Condensed for headings and labels, Barlow for reading text.

    Type scale

    Body text is 18 px with 1.55 line height. Nothing is set below 14 px; 14 px is for tags and captions only.

    Display · 44/48 · Semi Condensed 700Find a session
    Page title · 40/46 · 700MondAI Roundup, October 2026
    Section · 30/34 · 700Recently added
    Item title · 24/28 · 700Assessment failure as a driver of change
    Label · 18 · Semi Condensed 600Presentation slides
    Body · 18/28 · Barlow 400The October MondAI roundup is now online, with written chapters expanding on yesterday’s session and links to the original sources.
    Meta · 15–16 · Semi Condensed 600, muted27 October 2026 · Talks & showcases
    Tag · 14 · Semi Condensed 600Recording

    Colour tokens

    Ratios are computed in the page from the hex values. Pink is never used for body text on the tinted fills.

    PairSampleRatioUse

    Buttons

    One primary per screen. Secondary for the same level of choice elsewhere. Quiet for utilities such as copy link and feedback.

    Focus: 3 px pink ring with 2 px offset on light surfaces, white ring on navy. Minimum height 44 px.

    Material-kind tags

    Icon plus word, one calm fill. Kind never depends on colour.

      Search dropdown, with live results

      Type to filter sessions and courses. Arrow keys move, Enter opens, Escape closes. The count is announced to screen readers.

      Course card and booking button

      On a course page the booking button is the primary action. In a list it drops to secondary so a screen never carries three primaries.

      Feedback panel

      Entry point, open state and sent state are on the session page. Sent state shown here.

      Thank you, your feedback is sent

      Anonymous unless you added your name. Goes to the AI Competency Centre team.

      Error and unavailable message

      Says what happened, then offers a text route, another route and a contact.

      The recording is not available yet

      It may not have been published, or the player may be blocked where you opened this page.

      V2 · Centre signal

      Design notes

      The idea: keep the Centre’s existing navy and pink identity and its solid bars, but make navy the structure and pink only the active state and one small accent per area (a short tab on section rules, the current item in a list, the current slide). Content sits on white, with grey stages behind lists and slides, so contrast steps do the grouping.

      The slide reader is a stage: each slide is a white 16:9 panel on a light grey ground with a number band, a title, content and image arranged like a slide, and it holds up at phone width where the 16:9 ratio gives way to height that fits the content.

      How each journey is served

      J1 Revisit a session
      Session page keeps all materials in one list beside the selected material; the reader links back to “Recording and notes”.
      J2 Read slides as a reader
      Numbered slide panels, sized images, contents list, previous/next, quiet “View original slide N” link. Works with CSS alone.
      J3 Choose and book
      Course cards show level, length and format as labelled values; the booking button is labelled as the University booking page opening in a new tab.
      J4 Find by a remembered word
      Search dropdown with title, type and date, with matches marked. Restoring the query on Back is a build matter and is not mocked.
      J5 Ask for help
      A navy consulting strip on Home ends in one action; the strip is deliberately slim because this track fixes the layout.
      J6 Share a session
      “Copy session link” is a quiet button at the head of the session page.
      J7 Give feedback
      Quiet entry point on the session and reader pages; a three-field form (rating, free text, optional name) with the privacy line, then a sent state.
      J8 Maintain
      Not a visual question. The style uses one component per record type: event row, material tag, course card.
      J9 Something fails
      One amber notice pattern that names what happened and gives a text route, another route and an email. No-results search uses the same wording style. The reader’s scripts-blocked state is shown at #reader-noscript.

      Why a form, not an assistant

      A short form fits the “under a minute” test, is easy to point at an Oxford Microsoft Form, and needs no data-handling approval beyond what a form already has. A conversational assistant would need a hosted service approved for Oxford data.

      Left out on purpose

      • Real navigation, filters, calendar, trainer pages and the About-this-site surface: this track fixes the layout and only varies style.
      • Presenter names: the content pack holds none, so the slot says so.
      • Real logos, a real video player and real booking dates.
      • Dark mode, motion and illustration.

      Mock-up shortcuts

      • The reader shows the BETT 2023 sample deck inside the October MondAI session’s “Presentation” entry. This is illustrative; the sample belongs to a different talk.
      • The slide total (16) is set once on the reader wrapper as a CSS counter; the generator that wraps real readers would write each deck’s own total.
      • The contents list, current-slide highlight and previous/next bar are the only things that need scripts. Without scripts, the contents list and scrolling still work.

      Round 1 changes

      Acted on

      • Standalone reader.html with all 16 slides as static markup, CSS counters for “Slide N of 16”, per-slide previous/next anchors, a contents list (a native details on phones) and a marked, liftable reader style block. The session page’s “Open the slides” now points to it.
      • With scripts off, index.html shows the home page only (not every surface stacked); the #reader, #reader-noscript and #section-N hashes still open the styled deck from CSS alone.
      • Navy limited to the header band, primary buttons and the pale slide icons’ backing circle. The consulting strip, reader bar, materials heading, player slot and title slide are now white or pale grey. Pink is a short rule, active mark or bullet.
      • 390 px header is one row: logo pair and a Menu button holding the links and search. Logo placeholders drop their small caption on phones.
      • Slides: capped at 760 px wide, image height capped at 230 px, icon images at 96 px, title slide no longer forced to 16:9. More white space between sections.
      • Page bottom padding reserves room for the review button; every link, button and input has a 3 px pink or white focus ring.

      Declined

      • A sticky previous/next bar in reader.html: it cannot work without scripts, so each slide carries its own links instead. The sticky bar stays in index.html as the scripted enhancement.

      Escalated

      • Dominik, Ella: with pink reduced to an accent, is the Centre identity still recognisable enough, or should the header band carry a thin pink rule again?
      • Dominik, Ella: the reader.html contents list is written twice (side panel and phone drop-down) so it works without scripts; confirm that doubling is acceptable for the generator.

      Open questions

      • Dominik, Ella: is pink as accent only too quiet compared with the existing library, which fills whole header bars in pink? Round 1 made it quieter still.
      • Dominik, Ella: can Barlow Semi Condensed be self-hosted as one font file for the Canvas entry page, or should headings fall back to the system sans?
      • Dominik, Ella: should the slide stage be white-on-grey (as here) or should the title slide of each deck carry the navy block shown on slide 1?
      • Ella: should the session page list presenters from the event record? The pack does not include them.

      Surfaces

      #home#session#reader#style#notes

      States

      #search (live results)#search-empty#booking#consulting#unavailable#feedback (open)#feedback-done#reader-noscript