Boris Steipe @boris_steipe@qoto.org
- Sentient Syllabus Project
- http://sentientsyllabus.org
- Sentient Syllabus Analyses
- https://sentientsyllabus.substack.com
Joined Nov 2022
Is @scholar.social somehow blocked from #qoto?
I was surprised to find that I can't tag colleagues like Sarah Eaton (@drsaraheaton@scholar.social), and none of her tweets show up when I look for hashtags she recently used, like "AcademicIntegrity (today)" or "PaperMills (Feb. 15)".
Did I misconfigure something?
Some enlightenment would be much appreciated. Thanks!
"Postplagiarism" – are we there yet?
Sarah Eaton ((@drsaraheaton@scholar.social)), who has written on ethical issues in academia for more than a decade, thinks so. In her newest blog post ...
... she nails six tenets to the cathedral door. I'm not going to summarize, it will take you longer to read my summary than her original thoughts. Short, precise, and insightful.
I take a slightly different track in the Sentient Syllabus analysis on academic misconduct (https://sentientsyllabus.substack.com/p/generated-misconduct
), but I agree with all her conclusions.
#SentientSyllabus #ChatGPT #HigherEd #Education #AcademicIntegrity #Academia #PublicationEthics #ResearchIntegrity
Boris Steipe
boosted
This is interesting stuff, you can pass fairly simple story conflicts in #chatgpt and get interesting answers :
give me 5 reasons a princess in a fantasy story might decide to stay with the villain instead of going with the dashing hero--
Stockholm Syndrome: The princess may have developed a psychological bond with the villain due to prolonged captivity or trauma. She may feel a sense of loyalty or even love towards her captor, despite his evil deeds.
Shared Goals: The princess may share similar goals or aspirations with the villain, which she feels the hero does not understand or support. For example, she may want to bring about change in the kingdom that the villain is also fighting for, but the hero may not share this vision.
Fear of the Hero: The princess may have been raised to fear or mistrust the hero, perhaps due to propaganda or false information spread by the villain. She may believe that the hero is a threat to her and her kingdom, and therefore chooses to stay with the villain.
Desire for Power: The princess may be power-hungry and see the villain as someone who can offer her more control and influence than the hero can. She may choose to ally with the villain in order to gain more power or leverage over her own kingdom.
Personal Reasons: The princess may have personal reasons for staying with the villain, such as a past connection or unresolved emotional issues. She may also be motivated by a desire for revenge against someone she perceives as having wronged her or her family, and believes that the villain can help her achieve it.
I just came across an excellent take on job perspectives in the era of #GenerativeAI by Adrien Book (@thepourquoipas; thanks to Andrew Woolridge @triptych for posting).
https://www.wearedevelopers.com/magazine/how-to-protect-your-career-from-chatgpt
Adrien knows what he is talking about, he looks at specific careers and shows specific ways in which they may be disrupted - but most importantly, he develops a brief, concrete perspective for each, on how to respond. My redux of his advice: (a) emphasize the authentic human experience, and (b) make the AI work with you, not for you. This resonates completely with what we have been writing in the Sentient Syllabus Project.
https://sentientsyllabus.substack.com
http://sentientsyllabus.org
Good article, one of the best I have seen on the topic (and I have seen a lot 
.)
Although - Adrien! There's no free lunch yet for grading. "Bastille day" is indeed the anniversary of the storming of the Bastille, and the French national holiday. This would have lead to a rather embarrassing (and successful) remarking request. How to use generative AI in grading / marking is an unsolved question. We only know that we must use it, because the detailed assessment of generated essays does not scale.
#SentientSyllabus #ChatGPT #HigherEd #AI #Education #University #Academia #Writing #careers
Surprisingly, it reads and produces some SMILE strings – although not the chiral version (there it gets confused, eg it might produce a string with 4 O and insist there are six). But it might be fun to explore where the boundaries are, and if a long, expressive prompt could shift them.
🙂
So much has been happening in #AI and #HigherEducation but I finally finished the next update on the #SentientSyllabus resources:
https://sentientsyllabus.substack.com/p/resource-updates-2023-02-24
Some highlights: the _Syllabus_Resources_ got a makeover to align it with previous writing on Academic Integrity. The _Course_Activities_ page includes new observations on teaching programming, and creating graphics; as well as a new section of activity design to prevent generative AI use. The _Understanding_AI_Issues_ page includes new thoughts on alignment, and the potential of personalized AI - which is no longer a distant reality but available right now.
The Sentient Syllabus Project is a public good collaborative for practical and actionable resources for university educators. Resources include sample text for syllabi, thoughts on learning objectives and classroom activities, and some background material. I also publish newsletters and analysis on Substack.
Drop in, have a look around, and share with others.
:-)
#ChatGPT #Bing #HigherEd #AI #Education #University #Academia
Did you ever compare CoPilot with ChatGPT?
Boris Steipe
boosted
I think I don’t want to ever go back to a world without #ChatGPT: yesterday I spent ~6 hours building a feature from scratch, a tableview with search on a remote database with a segue to a PDF loading from a FTP server. I did that easily, pretty standard stuff.
But before closing off, I thought of adding a color highlight to the substring matching the search query.
I was pretty tired, but AI gave me a working method in seconds.
Would have it been hard to do it myself? No
But I wouldn’t have.
Boris Steipe
boosted
Students are #writing my usual first paper assignment: just 2 paragraphs.
As usual, they say it's WAY harder than their usual paper assignments.
Today I asked, "How is the usual paper easier to write? Isn't it longer?"
“Yeah", one replied, "but writing 10-15 pages is easy. It can be mostly fluff. Even a #chatbot can easily do it. Proofreading is the hardest part.”
Why are my 2 paragraph essays so tough? And why can't chatBots do it?
(Toot 1 of 2)
Boris Steipe
boosted
I've been playing with #ChatGPT to see how niche it's knowledge can go.
So far, it has given me suggestions for troubleshooting my RV furnace which keeps overheating, and helped me plan out the next phase of my off-grid solar installation by helping me calculate voltages, amperages, and figuring out what wire gauge to use.
This is nuts.
We only see the hype, not the numbers. I take the current slowdown re. #Bard to mean that Google is not seeing a meaningful drop in search volume. There doesn't seem to be the bleeding sense of urgency that was tangible in Google's Paris event two weeks ago. And I'm not surprised: the newly constrained #Bing does not appear to offer much more than, well, Bing.
#ChatGPT on your desktop?
FlexGen paper by Ying Sheng et al. shows ways to bring hardware requirements of generative AI down to the scale of a commodity GPU.
https://github.com/FMInference/FlexGen/blob/main/docs/paper.pdf
Paper on GitHub - authors at Stanford / Berkeley / ETH / Yandex / HSE / Meta / CMU
They run OPT-175B (a GPT-3 equivalent trained by Meta) on a single Nvidia T4 GPU (~ $ 2,300) and achieve 1Token/s throughput (that's approximately 45 words per minute). Not cheap, but on the order of a high-end gaming rig.
Implications of personalized LLMs are - amazing.
This is cute - but it's a frequent error that you see under high load. Try to copy your prompt and reload the page. It's typical answer would be to thank yopu for pointing out its error. It's actually fun to find it can learn _within_ a conversation.
🙂
Boris Steipe
boosted
Discovered a new use for #chatgpt: Practice tests. Tell it what you're studying and what kind of questions you want. Fine tune the difficulty if you need to. The questions and answers are sometimes wrong, but if you use it for reviewing stuff you've already learned, it's usually fairly obvious and gives you a starting point.
😞
You are conflating knowledge about experience with the actual experience. That's not the same. ChatGPT gave you a perfect answer in both cases.
I think you are asking the right question, linking UBI to AI concerns. After all, when we outsourced physical labour to the steam engine, we were still employable as knowledge workers. But what is left once we outsource that too?
Although, when I read your article, I noticed you saw three strategies for the future: quantity (content mills), community (subscribers), and manipulation (outrage-baiting). I am curious that you did not mention "quality" as another option.
I understand the dystopian perspective that quality only matters if it can be heard, but for academia (the topic that I am engaged in, with the #SentientSyllabus project) quality is really the only option: to understand how the human perspective can go beyond the AI, _and_ for society to understand how that creates value.
A win-win proposition requires both: creating quality on the creatives' side, and encouraging culture on the recipients' side, a culture through which individuals recognize and values what is being produced as an enrichment of their own lives. This is nothing other than the ideal of education.
When I imagine how this could translate into perspectives for the content-writing economy, I realize that we share many concerns. And conversely, the community-building angle you mentioned is something we may be paying too little attention to (It is implicit in the "shared commitments" I describe in my analysis of academic misconduct - but making such communities explicit may be productive. https://sentientsyllabus.substack.com/p/generated-misconduct )
More to think about. Thank you.
Good prompt. :-)
Since you are not asking this question to ChatGPT itself, I presume you are asking how to get beyond the obvious. Perhaps start with reading between the lines of what's going on between Google and Microsoft.
https://sentientsyllabus.substack.com/p/reading-between-the-lines
There's more analysis on my Substack, and if you need even more, feel free to get in touch.
Have fun!
- Sentient Syllabus Project
- http://sentientsyllabus.org
- Sentient Syllabus Analyses
- https://sentientsyllabus.substack.com
Joined Nov 2022
