Nobody cares about "AI" (Chatbot: I disagree.)

[Khamon]

Good writing assignments are scaffolded so that the students have to turn in a bibliography and thesis statement, outlines, and/or other types of pre-writing. That will tell me a lot about what the student knows. However I will not hold an interview with every student on every assignment (4 major assignments x 72 students = 288 meetings per semester (and my department has a lighter teaching load than many others: an adjunct in English may have 120 students per semester)). There is no time, and in the workplace no one asks you why your work is crap.

You would, though, take the time to interview a student concerning a paper that you thought might not be their work. Perhaps I'm misinterpreting "using generative AI tools on a history project." We do have to believe that the adults handled the situation properly.

My English 101 instructor did take the time to interview me after giving my first paper a zero because she believed that it was from a mill. I answered all of her questions and offered to take her to the books I had referenced in the Library. She would have none of it; the paper was copied and received a grade of zero. My final grade was a solid A but she never revoked that zero.

 

[Veritable Quandry]

Yes, I would allow them an opportunity to defend their work. I also need more than a gut feeling that it's "too good" before I go that far. Usually you can find key phrases in a Google search.

 

[Khamon]

I would have expected the student to get a 0 for that paper.

I would not. Half, seriously half, of my students cannot read for content, outline, or take notes. This is true every semester. Some barely read at all. Yet they have high school diplomas and are admitted to college. They come from all over the country to this military post-secondary school, that the armed forces academies refer them to, primarily because they cannot read. Everybody passed the binary/hexadecimal conversion test last week. But they can't read or write worth a poot.

 

[Argent Stonecutter]

They don't get a 0 for cheating?

 

[Khamon]

They don't get zeros for anything including cheating or not submitting the work at all. They drift through and are passed through. Being military, and residential, we deal with some trauma when they realize that the adults in their life let them slide by without being prepared to succeed. I've witnessed students calling parents, teachers, counselors, coaches, and screaming, crying, or both while they tell them that they can't pass their classes because they were allowed to slip through high school with no requirement to learn.

It is unfathomable that a high school graduate is stuck in developmental English, because they read at a fourth grade level, and developmental Math because they can't add three four-digit numbers correctly (I kid you not). But we see it every semester. Multiple sections.

 

[Khamon]

In retrospect to the AI discussion, we actually don't know how extensively the student used the AI, or how diligently the instructor analyzed the student's knowledge and methodology. We're subjected to lazy, or intentional, reporting. The reporter may not have asked ample questions to give us enough facts to know the truth or the editors may have purposely omitted that information. Sloppy and misleading reporting is normalized to the extent that it gets published, and repeated, and discussed.

 

[Free]

Sloppy and misleading reporting is normalized to the extent that it gets published, and repeated, and discussed.

Kind of how AI succeeds as a business model.

 

[Veritable Quandry]

Some Apple researchers are dumping on the actual intelligence part of AI.

Apple study exposes deep cracks in LLMs’ “reasoning” capabilities

Irrelevant red herrings lead to "catastrophic" failure of logical inference.

arstechnica.com

So while I was showing my students how to use AI for topic development, I gave them a big warning about checking facts. Then I asked CoPilot how many "r"s are in "strawberry." It replied 2. I prompted it to spell "strawberry" and asked again. Answer was still 2 even though it clearly used 3. So I asked it why there are two "r"s in strawberry and it gave an interesting history of the word, where the two roots had a single "r" in each. But it still could not acknowledge that "strawberry" has 3 "r"s in it, and the reason is not entemology but the way LLMs turn words into tokens that they can process.

 

[Free]

So while I was showing my students how to use AI for topic development, I gave them a big warning about checking facts. Then I asked CoPilot how many "r"s are in "strawberry." It replied 2. I prompted it to spell "strawberry" and asked again. Answer was still 2 even though it clearly used 3. So I asked it why there are two "r"s in strawberry and it gave an interesting history of the word, where the two roots had a single "r" in each. But it still could not acknowledge that "strawberry" has 3 "r"s in it, and the reason is not entemology but the way LLMs turn words into tokens that they can process.

That's kind of like when you ask pirates how many 'r's are in the word ar, they'll say 3.

 

[Veritable Quandry]

 

[Free]

Oh great, an invisible phishing solution.

Invisible text that AI chatbots understand and humans can’t? Yep, it’s a thing.

A quirk in the Unicode standard harbors an ideal steganographic code channel.

arstechnica.com

What if there was a way to sneak malicious instructions into Claude, Copilot, or other top-name AI chatbots and get confidential data out of them by using characters large language models can recognize and their human users can’t? As it turns out, there was—and in some cases still is.

The invisible characters, the result of a quirk in the Unicode text encoding standard, create an ideal covert channel that can make it easier for attackers to conceal malicious payloads fed into an LLM. The hidden text can similarly obfuscate the exfiltration of passwords, financial information, or other secrets out of the same AI-powered bots. Because the hidden text can be combined with normal text, users can unwittingly paste it into prompts. The secret content can also be appended to visible text in chatbot output.

The result is a steganographic framework built into the most widely used text encoding channel.

 

[Argent Stonecutter]

The whole pirate R thing is because they were often from a part of England that still used the rhotic R that died out in the old country and became the American accent. For North Americans, every day is "talk like a pirate day".

 

[Dakota Tebaldi]

The whole pirate R thing is because they were often from a part of England that still used the rhotic R that died out in the old country and became the American accent. For North Americans, every day is "talk like a pirate day".

Hmmm....maybe yeah, but I'm not so sure. Like, aside from the "arrr" thing in particular, whenever pirates talk it sounds very obviously Irish. To me, at least.

But, we're also talking about a thing that could very easily have been completely made up for movies, like maybe the very first pirate character who had spoken lines in a movie just made up this accent and that was officially the pirate accent forever after.

 

[Erich Templar]

Hmmm....maybe yeah, but I'm not so sure. Like, aside from the "arrr" thing in particular, whenever pirates talk it sounds very obviously Irish. To me, at least.

But, we're also talking about a thing that could very easily have been completely made up for movies, like maybe the very first pirate character who had spoken lines in a movie just made up this accent and that was officially the pirate accent forever after.

Close. The late great Robert Newton basically created the pirate 'accent', and every movie pirate since has been impersonating him.

 

[CronoCloud Creeggan]

Like, aside from the "arrr" thing in particular, whenever pirates talk it sounds very obviously Irish. To me, at least.

Irish and Scottish dialects are also Rhotic, as are certain West Country accents still...which is where Robert Newton is from. (and was traditionally pirate land as well). You can find examples of the Somerset accent online. As I sometimes say, Americans speak proper Queen's English....Queen Elizabeth the First! (her accent would have been Rhotic so she would have sounded more like an old-timey American)

 

[Khamon]

The defendants' motion to dismiss the complaint, filed last week, said RNH admitted "that he used an AI tool to generate ideas and shared that he also created portions of his notes and scripts using the AI tool, and described the specific prompt that he put into the chatbot. RNH unequivocally used another author's language and thoughts, be it a digital and artificial author, without express permission to do so. Furthermore, he did not cite to his use of AI in his notes, scripts or in the project he submitted."

Student was punished for using AI—then his parents sued teacher and administrators - Ars Technica

With this information, yes, he was wrong to not cite the use of the AI tool. The grading is a bit harsh for the infraction but that's an instructor's prerogative. It is good to have some factual information.

 

[Free]

That should fix things.

Penguin Adds a Do-Not-Scrape-for-AI Page to Its Books

The world's largest publishing house doesn't want its copyrighted work used to train AI models.

gizmodo.com

 

[Free]

New rule: Do not work as a model for any company involved with AI.

Human Models Horrified to Discover Their Faces Are Being Used for AI Propaganda

Human models were paid for their likenesses to train AI. Now, to those humans' horror, their faces are being used to peddle propaganda.

futurism.com

 

[Dakota Tebaldi]

Tired of feeling obligated to occasionally call your elderly parents so they think you still care about them? You can pay this AI to call and talk to them instead, and it will send you a notification if it detects something wrong with them, so you can ignore them with confidence!

 

[Free]

Claude AI Gets Bored During Coding Demonstration, Starts Perusing Photos of National Parks Instead

Anthropic's Claude AI is purportedly capable of using computers like a human. And apparently, that includes procrastinating.

futurism.com

While its developers were trying to record a coding demonstration, the latest version of Claude 3.5 Sonnet — Anthropic's current flagship AI — got off track and produced some "amusing" moments, the company said in an announcement.

It's perilous to anthropomorphize machine learning models, but if this were a human employee, we'd diagnose them with a terminal case of being bored on the job. As seen in a video, Claude decides to blow off writing code, opens Google, and inexplicably browses through beautiful photos of Yellowstone National Park.

In another demo attempt, Claude accidentally stopped a lengthy screen-recording in progress, Anthropic said, causing all the footage to be lost. We're sure that wasn't intentional on the AI's part.