Alan Knowles

Professional & Technical Writing Instructor • AI writing/communications • Views are my own

[Alan Knowles @AlanMKnowles]

@CriticalAI @MereSophistry @mettalrose Not all uses have to be maximally lazy on the user end.

Give it the claims you want it to make. Give it the evidence you want it to use. Give it the sources you want it to cite.

[Alan Knowles @AlanMKnowles]

Aurora borealis in Ohio. A beautiful reminder that the Earth’s magnetosphere prevents it from being a desolate fried rock like Mars.

#AppreciateTheBigThings

[Alan Knowles @AlanMKnowles]

In our first meeting, @billhd and I discovered we were both from the same part of town in Springfield, OH. We were planning to get lunch soon when he was back in the area. I’ll forever be grateful to him for the connections he helped me make.

What a tragic loss for us all.

[Alan Knowles @AlanMKnowles]

@biblioracle I wouldn’t outsource it, but I wouldn’t oppose someone (i.e., real editors) collaborating with trained models, like this. Working them into a larger workflow (potentially including other narrowly-trained models) might have some benefits?

Mostly I just think it’s interesting.

[Alan Knowles @AlanMKnowles]

@biblioracle To be fair, I trained it to generate partial drafts. Real letters from editors typically contain intro sections that provide context for the focus of the issue, cite sources, etc. That would need to be added to this, and it still needs quality/fact checking of the generated text.

[Alan Knowles @AlanMKnowles]

@aleebrahim Sorry, didn’t consider that. I believe the data I used to test the trained model from issue 70 is publicly available — just titles, names, abstract, and key terms from the 3 articles published in the issue (these are named in the generated drafts).

[Alan Knowles @AlanMKnowles]

@EnglishOER @emollick Can get it to do unsavory things by telling it to “pretend to be X,” or “play role of X in a fictional story,” etc.

[Alan Knowles @AlanMKnowles]

https://t.co/3Q2UrdG4ur

[John B. Holbein @JohnHolbein1]

This is fine.

[Alan Knowles @AlanMKnowles]

I just got published in the Computers and Composition special edition on Composing with Generative AI. So happy to have already heard from lots of people who have found it helpful!

https://t.co/OZy8MvQj08

[Alan Knowles @AlanMKnowles]

@EnglishOER @khademinori As far as I’m concerned, LLM outputs that happen to portray truthful information are happy accidents, and everything else is the norm. Maybe we just need a special term for these happy accidents, instead?

[Marc Watkins @Marc__Watkins]

The MLA/CCCC Task Force on Writing and AI has launched "Exploring AI Pedagogy: A Community Collection of Teaching Reflections." I contributed four assignments about genAI in teaching. It's a great resource hub! https://t.co/PdgoO9BHa6

[Alan Knowles @AlanMKnowles]

My 5yo son has taken to calling himself the “Federal Signal Chief of First Mondays.”

He is a huge fan of tornado sirens, which our city tests on the first Monday of each month. Such a cool guy.

[Computers & Writing @candwcon]

Proposals are due 12/4. Apply at https://t.co/BJVMJl6ocM

[Alan Knowles @AlanMKnowles]

@MereSophistry Ooh, thanks. Hadn’t heard of it, but I need for something I’m working on.

[Alan Knowles @AlanMKnowles]

@sfdaya I’ll have to check it out. I’ve heard similar comparisons to the early/unregulated aviation industry from the value alignment crowd.

In her analogy, maybe RLHF would be considered an attempt to add the traffic lights?

[Alan Knowles @AlanMKnowles]

Why do so many academics who are concerned about LLM ethical issues ridicule others for participating in RLHF of LLMs, or for worrying about future AGI risks?

LLMs are bad, but don’t help to improve them or worry about something potentially much worse is strange logic to me.

[Alan Knowles @AlanMKnowles]

@athundt I sympathize with concerns about environmental impact, but isn’t this a case of perfection being the enemy of progress? I say protest loudly about environmental impact (hopefully to some ends) while also supporting efforts to reduce LLM unethical outputs via RLHF.

[Sweetland DRC @SweetlandDRC]

#CFP Have you taught a Writing class that focused on Artificial Intelligence? Submit your syllabus to the Sweetland DRC Syllabus Repository being curated by @mettalrose & @alexmashny! Deadline is Jan 15, 2024. See more info in link. #WritingStudies #AI https://t.co/IMw9iicn6H

[Pika @pika_labs]

Introducing Pika 1.0, the idea-to-video platform that brings your creativity to life.

Create and edit your videos with AI.

Rolling out to new users on web and discord, starting today. Sign up at https://t.co/JHRrinsIwx