วีดีโอ

Excellent Discussion, Thanks to the organizers and experts

Your disclaimer touched on my expectation that there will soon be AI-managed education to reduce the cost and make on-demand education accessible. Coursework will have a much broader knowledge base (not limited to instructors' resources) with case studies researched and drawn on current news and activities. Loved the idea of snippets of useful virtual content mixed in with traditional content delivery -- helps with retention.

Thanks for a very informative and thought provoking lecture.

Fascinating panel discussion - - very thought-provoking dialogue among the 4 experts. Listening as a high school teacher, it was interesting to learn of a trust issue among some teachers in NC to use ITS, and that Chat GPT has broken down some of those barriers. The topic of standardization at the start of the discussion was particularly intriguing, because standardization could minimize some of the costs, and therefore, accessibility for lower-income districts could be increased. Thank you for covering so much territory in only one hour!

Thank you for sharing this talk! Really inspiring me to think about how to do my research in the similar field ~

Bender's critiques on LLMs certainly add spice to the academic stew, focusing heavily on labor and environmental woes. But is she too caught up in the gloom, missing the tech evolution train? It's like she's zeroing in on the shadows in a masterpiece, ignoring the bigger art movement. Sure, her points have weight, but are we risking an oversight of the transformative tech potential here? Seems like a sharp mind potentially boxed in by pessimism.

Using AI to have a socratic dialogue, as a tutor opens up new pedagogical insights; because in the age of AI banning it altogether does not work or help rather coming up with ways that nont only enhances learning but also engages the learner actively. Thank you for the very informative session

Sound?

Interesting talk. Thank you for sharing. For the question "why have we not seen these broad scale systems change" at 52:00 I think it is because it is one of the "most important" mechanism for addressing the future, the risk appetite is very low to make mistakes. And the system has become too big to change and universities and technical institutes need to start perhaps using AI to code different experiences for entry, which leaves schools to customize education for the communities they serve.

Promo-SM 👉

Thank you for an interesting session. Could you please provide the link to the Google document you shared?

❤❤❤

The sound seems to disappear half way through this very interesting panel?

Helpful presentation. Thanks!

this shit wrinkled my brain

Very interesting!!

39:27 Fascinating! In that scenario, it might be better to avoid watching news channels like Fox News when children are present, don't you think?

This was really fun to watch and educational! Thanks for sharing, I will check what Khan academy offers for first graders..

Dear Mike, thank you for your seasoned, insightful , and ever-learner-centered perspective in our hottest topic in edtech since the WWW!! Warm regards, Roy

excellent presentation, tank you

Thanks for sharing this publicly! Listening to Emily Bender is such a refreshing experience during this (hopefully temporary) moment of collective irrational AI exuberance. The comment during Q&A about universities apparently considering using LLMs in academic advising would make for a good (or awful?) example of the problem of accountability when institutions choose to hand communication duties to generative AI. I work at a university; academic advising is part of my job. One major part of advising is helping students put course schedules together, which seems simple but requires the sort of planning that we know LLMs can't do (e.g. respecting prerequisite structures when planning sequences of courses to take over the next two years). Are universities prepared to clean up the messes that LLMs will inevitably create if tasked with this? Registration errors are probably the most straightforward type of damage control universities would have to do when LLMs give bad advice. There are worse. Students also look to their advisors when dealing with personal crises, underconfidence, conflicts with instructors, the tough decision of whether to drop a class or change their major or leave college altogether, those kinds of things. Say an advisor insults or undermines or misleads a student There will be (or ought to be!) recourse and accountability. But if an LLMs does it... what then? Even if a noble/foolish administrator promised to "take responsibility" for the bad actions of an LLM, this is a contractual kind of responsibility, like a co-signer on a loan being responsible to the bank when the other person stops paying. While it's sill important for someone to take responsibility on the institution's behalf, I'm guessing an administrator's heartfelt apology for the behavior of a chatbot would ring hollow. There's an inherent moral responsibility underlying human communication that just ain't there when we delegate it to a human-sounding machine. Even in an extreme case where the LLM says something gratuitously offensive and someone from the institution is fired as a result... that kind of "accountability" is qualitatively different from the kind that we'd put on a person if they said what the machine said.

Thank you Prof. Bender ❤

has emily ever really used a LLM? i have the strong impression that she never has spent much time trying to use it as a "partner" for delving into complex subject fields. this has nothing to do with the way children learn language or understanding defined as being grounded in the physical world, but this is not a problem at all for many use cases. in fact, real language in use is for the most part not (directly) grounded. most human language is relating to other language most of the time, and this is what LLMs are doing too. they are far from perfect, but they are stunningly good in this. the parrot-metaphor is totally misleading because the "prediction", that is the forming of statements, is obviously based on a complex model of semantics that the "learning machine" has built in numerous complex processes of training and fine-tuning. of course, this is nothing like a human "mind" (and talking about AGI is bullshit), but it is a very capable agent or medium of "discoursive intelligence", represented by the enormous amount of language data it has been trained with.

So good! Thank you!

One comment on the "Make LLMs 'safe'": If we were able build a Device that is able to fact-check and filter biased contents from ChatGPT's output, we wouldn't need ChatGPT any more; the Device itself would be the tool were are looking for in ChatGPT.

Wonderful talk

Really helpful Mike, calming explained backed by real examples

Thank you.

Thanks Mike, that was so well articulated. Brillant slides.

Is the PowerPoint presentation available?

I just listen one of those professor's offline lecture, and find Graile's TH-cam account. Nice job.

Note that when ChatGPT generates a sample course syllabus at 11:05, the course textbook it suggests ("Teaching with Technology: A Practical Guide" by K.J. Willis and M.D. Johnson) does not actually exist; it is a typical ChatGPT "hallucination."

Thank you

Will you help me to grasp some construct