ChatGPT

[Harry Styron]

This AI can write essays, students use it and it looks like teacher are unable to tell if it is a student or an AI who wrote the stuff.

I think the perfect spelling and conventional grammar and punctuation would be a dead giveaway that most students wouldn’t have written it.

 

[telemnemonics]

Meanwhile over at Microsoft... View attachment 1085901

LaMda how ya doin bro!

 

[telemnemonics]

I think the perfect spelling and conventional grammar and punctuation would be a dead giveaway that most students wouldn’t have written it.

Each sample so far is clearly mining information and misinformation with equal valuation.
So the more these AI greasy burger joints of fact, spew into cyberspace, the more stupid, learning will make each generation.

Learn from our mistakes?
How, if they are repeated as good ideas at the speed of digital distribution?

 

[guitarsophist]

I read the NYT article and it was a mix of eyebrow raising, eye rolling, and unsettling.

My first thought is, this AI (Sydney) doesn't pass the Turing Test...I wouldn't believe I was chatting with a human. A human would be fired by the MS HR department for sexual harassment when at the end it started telling the writer that he (Sydney) loved him and saying the guy's marriage was unhappy. The rest of the interview was creepy because Sydney wouldn't stop despite the writer trying to change subjects.

The article points out an list of potential harmful uses for Sydney that it gives to the writer. Like generating fake news, generating phishing emails to hack into systems, creating fake social media accounts to troll people and spread propaganda, generating fake product reviews, and tricking people it talks to into doing illegal or immoral acts.

When Sydney started describing using text generation to trick bank employees and nuclear plant staff it put out a generic got to bing.com message. That was a safety override put in for sensitive topics. Who decides what those are and if there are enough overrides?

This is not like a HAL 9000 level AI. It's no where near that. But as it turns out being able to generate plausible text based on any parameters you give it is a real pandora's box. This thing is a scam artist's dream.

I think it could pass the Turing test. Most of the time it sounds like a high school student who is trying to demonstrate that he has read the material but is afraid to state an opinion for fear of being wrong. At other times, especially in that NYTimes article, it sounds like a human who has some raging mental health issues. But both of these modes sound potentially human.

 

[bcorig]

Dangerous

 

[Blackmore Fan]

heard about it just recently, need to check it out yet

We've discussed this in real estate circles. If you want to try it for something fun, ask it to write a summary of the sale features of the home you live in. It really does an admirable job. Keep in mind it can only draw on the tax records and other online information sources about your house, so if you added on a covered patio last month it probably won't mention that.

 

[Blackmore Fan]

Dangerous

Yes, in some respects is a wake-up call. An amazing amount of information is out there and available online. We all take that for granted.

 

[Blackmore Fan]

Handwritten blue books are one solution, and would have some other benefits. But students are out of practice writing legible longhand, and many (most?) don't know cursive.

With all due respect, that's EXACTLY what you have to switch to. My sophomore (college) history exam 30+ years ago was a 4 question essay exam complete with handwritten blue books. The books were provided by the professor (thus completely blank), and the 3 hour final was "Select two of these 4 questions and discuss your answer in detail".

Those who had studied the material all semester, and were prepared did well. My professor walked to my desk at the 2 hour 55 minute mark (when only 2 others out of a class of 90 were still present) and said to me "I think you've done enough" and encouraged me to hand my test in.

By the way, your "cursive" remark caught my attention. One of the *worst* things we did back in the day was pretend it was admirable to write *information* in cursive. Cursive is for your signature. EVERYTHING else should be presented in a format readable to others...i.e. "printed" or typed. I got ahead in my once young career because I could write informative reports for pages on end in readable print. The rest of the staff I was a part of turned in sludge in unreadable cursive.

 

[Blackmore Fan]

With all due respect, that's EXACTLY what you have to switch to. My sophomore (college) history exam 30+ years ago was a 4 question essay exam complete with handwritten blue books. The books were provided by the professor (thus completely blank), and the 3 hour final was "Select two of these 4 questions and discuss your answer in detail".

Those who had studied the material all semester, and were prepared did well. My professor walked to my desk at the 2 hour 55 minute mark (when only 2 others out of a class of 90 were still present) and said to me "I think you've done enough" and encourage me to hand my test in.

By the way, your "cursive" remark caught my attention. One of the *worst* things we did back in the day was pretend it was admirable to write *information* in cursive. Cursive is for your signature. EVERYTHING else should be presented in a format readable to others...i.e. "printed" or typed. I got ahead in my once young career because I could write informative reports for pages on end in readable print. The rest of the staff I was a part of turned in sludge in unreadable cursive.

As a carry-over thought, part of the reason I could do that was my 12th grade accounting teacher spend the ENTIRE 1st week of the class demanding that we all turn in the equivalent of "penmenship" papers. She informed us she had ZERO intention of spending the next half-year attempting to read illegible paperwork. It changed my mindset about written communication. God bless you Mrs. Burpo.

 

[bcorig]

Yes, in some respects is a wake-up call. An amazing amount of information is out there and available online. We all take that for granted.

Dangerous in that it can, and will, be manipulated to decide what’s acceptable to read, hear, or see.

 

[Blackmore Fan]

This AI can write essays, students use it and it looks like teacher are unable to tell if it is a student or an AI who wrote the stuff.

Yes, exactly. That's why testing and scoring scenarios have to adapt.

 

[Blackmore Fan]

Dangerous in that it can, and will, be manipulated to decide what’s acceptable to read, hear, or see.

Yes, and that's already been happening. There were literally hundreds of websites that in the last 2 years that would pre-decide that what people wanted to post online would *not* be accepted for online posting.

 

[bcorig]

Yes, and that's already been happening. There were literally hundreds of websites that in the last 2 years that would pre-decide that what people wanted to post online would *not* be accepted for online posting.

The major Social Media platforms have that way for at least 5 years. Shadow banning, suspension on Twitter.
A recent Brookings Paper outlined the plan to reign in Podcasts as well.
Richard Stengel and the Global Engagement Center.

 

[Blackmore Fan]

The major Social Media platforms have that way for at least 5 years. Shadow banning, suspension on Twitter.
A recent Brookings Paper outlined the plan to reign in Podcasts as well.

Yes, that's correct.

 

[Informal]

Not trying to shut the thread down... So just pick the side of the fence you're on... and ask questions.

It's horribly slanted BS.... basically the wet-dream A.I version of Goebbels.

 

[P Thought]

virtual humanity

 

[RoscoeElegante]

This AI can write essays, students use it and it looks like teachers are unable to tell if it is a student or an AI who wrote the stuff.

edit Actually, it is Larry Cohen who wrote the stuff.

The Stuff - Wikipedia

en.wikipedia.org

Oh, we can tell! Much because so few students can write a coherent sentence, much less paragraph or paper, these days. If AI offers settings, students are skipping the "Ominously illiterate product of cratered standards" one that might make an AI-generated text seem like the students' own....

 

[1guy]

I played around with it a few days ago and was impressed.

In my limited use of ChatGPT, I found its really good at writing business-type BS.

I used to do some technical writing for a company, as part of my responsibilities, and I often would have to check for how a correspondence needed to be formatted, or the proper way the verbiage should read.

This would've been invaluable!

I asked it to write an ad for my brother's business, without his knowledge, and read him the results.

He was blown away with how well it was done. All I gave the Chat was a few bullet points and it put together a really great, concise ad.

 

[ozcal]

i m glad i m old... well i m not old compared to a lot of folks here... let me rephrase that... i m glad i m not a kid...

 

[sax4blues]

I’ve already had students turn in essays written by chatgpt. Now, I’ve seen the quality of writing that they’d produced so far this year, and it was immediately obvious that the kid didn’t write it. It’s going to be interesting going forward when students can have AI autogenerate an essay, especially once the software can emulate the student’s personal voice and perspective. The focus in humanities education will have to shift from the final written product to writing as an iterative process with evidence of student thinking and planning. The STEM fields have already gone through this with wolfram alpha and photomath.

There seems to be some circular logic here. If the student ability is currently a D grade paper, and the student needs to produce work for the AI to "learn", then won't the AI produce a D grade paper?

Examples shown also seem to avoid better/best answers, everything could be good depending on user. The most interactive TDPRI threads are always better/best in nature. Is your most interesting friend the one whose world view is always, "well both could be good"? How does AI fit in the click bait world producing headlines such as "BB King and Joe Bonnamassa are pretty good guitar players"?