Meeting minutes
Collaboration Tools Accessibility User Requirements.
jason: issues agreed last week to add
teatment of generative machine lerning envrionments
integrated envrionments in collaboration tools
janina: mentioned accessibility checkers still an issue?
may have collaborative aspects
some issues interrelated, e.g. ability to get summaries - generative AI
suggests APIs shoudl be developed
raja: on zoom, can get summary and AI optin but have to tell admin
interesting and relevant
janina: yes, have used in some meetings that zoom provides in paid version
could request it
raja: tehre's bias in AI models sometimes
want to add someting that reflects us in meetings
can we choose our own AI?
janina: not up to us to name systm, but need to note how accessibility fits in
jason: potential issues we can raise
can dicuss relaitonship between genrative machine learning systems and prompting/review expected of content with accessibility features
e.g. ATAG expations of reivew and prompting of text alternatives, etc
if auto-generated audio, ggraphical content, subject to same expectations
authors shoudl be promopted
additionally, may be to generatively crate accessibilit
<janina> Recommending a book, "The Alignment Problem," Wikipedia page is here: https://
e.g. captions, alt text, etc
also generative systems ccome in different forms: server cluster, some local, some may combine the two,
so any solution shoudl be capable of implemenaiton regardless of where machine learning is executed
scott: can we separate AI summaries generallyu,
e.g. Microsoft co-pilot, compared to specific collaboration tool?
jason: comment summaires, review summaires as suggested summaries coudl be speciifc CT example so AI processes before human approved
raja: could roy provide uique words for interpreters on AI terms
AI, collaboraiton tools, Microsoft, PowerPoint
jason: terminology in the agenda
interpreter: do get definition list regularly
raja: when Microsoft powerpoint sees slides, automated captions becomes more accurate
assumption is that increasing the predictability
jaspon: in relaiton to collaboration tools, can create long language prompts for machine learning
so background info can be put into prompte that influences what it does in dialogue
i.e. primed with mateirl before generating
secondly, Google and Microsoft developing systems that take into account organisaitons' content
i.e. not just the one your'e working on
janina: we can get a demo of this if group is intersted
janina to follow up on possible demo with Lionel
jason: back to CTAUR< will keep working on both issues
should say something about automated AI generated summaries
other themes?
janina: will spend more itme,
will continue working on it
formal liason to COGA - David Swallow
Media Accessibility User Requirements.
janina: media intreest group rechartered
jason: APA made initial comments on broadcasitn gproposal
going to several standar documents
MAUR: still gathering updates at the moment
Accessibility of CAPTCHA.
jason: some references added
planning to contact W3C security gorups
janina: Lionelmay be able to contribute
goal: up to date when we move to publicaiton wth industry expectaiton
Miscellaneous topics.
jason: announcement from Google - features using automatically created summaires of long web pages with links to sources
so summary poitns back to oriignal URI
raja: does it need to be a different type of usmmary depending on learning and language, e.g. ESL vs cogntiive disability
can we look at that in RQTF?
janina; yes, different summaires are needed and feedback is going into these processes
our liason can help progress this for RQTF
janina: time zone chnages in north america
Scott apology for next week