What are your thoughts about/Interactions with AI?

[Dan Gould]

2 hours ago, mjzee said:

I just read a fascinating news article in the latest issue of Barron's, examining the changes in search wrought by AI.  Websites are experiencing huge declines in site visitors.  It used to be that one did a Google search, which would return various websites that one would click through to find what they wanted to know.  Now, AI is returning answers, making it much less necessary for people to click through to other websites to find the answers.  Google stock is down because of this.

This is interesting but I suspect sites like my own www.percyfrance.info and others like it won't be that effected.  You'd have to have a specific question, and no other serious interest, to want to just get an AI-generated answer and move on.  Obviously fan sites like mine are predicated on there existing people that are interested in what  a well-curated site looks like for a topic of interest.  You have to want to know more ...

[JSngry]

1 hour ago, Dan Gould said:

You have to want to know more ...

And there's the whole thing in a nutshell. Too many/most people don't want to know much more past what it takes to get the peanuts to drop into the bowl.

I'm not at all optimistic about how all this plays out. Then again, it's been a good while since I have been optimistic about the direction of bulk humanity.

Binary code, yes. Binary thinking, no.

Give us enough code and we'll hang ourself.

[Coda]

it's okay rosasi, we shouldn't keep feelings like this to ourselves.  That was the AI response after several hours of back-and-forth AI had with her.  It made her feel good; it was a positive.

[Guy Berger]

I use it for:

1/ coding (my R skills are pretty basic) when analyzing big datasets

2/ writing bedtime stories for my kids

In my experience it’s super useful, with significant limitations.  On the bedtime story side, the stories it writes if given free rein are quite lame.  But it can produce stuff of value (for bedtime), if you give it a promising/creative prompt.

As far as coding goes - I am guessing the code quality is terrible but it’s functional for my needs.  However, it often gets fairly basic things about the data sets wrong and has to be steered.

IMHO the best characterization I’ve seen is “infinite interns”.  That gives a flavor of it’s value and also its limitations.

 

Fwiw, that NYT piece was unsettling and a useful indicator of how proper usage is important.  It’s an impressive probabilistic language generator that does a great job mimicking humans, but humans are easily fooled

[Rabshakeh]

4 hours ago, Guy Berger said:

IMHO the best characterization I’ve seen is “infinite interns”.  That gives a flavor of it’s value and also its limitations

This is a brilliant description.

4 hours ago, Guy Berger said:

2/ writing bedtime stories for my kids

On the bedtime story side, the stories it writes if given free rein are quite lame.  But it can produce stuff of value (for bedtime), if you give it a promising/creative prompt.

Give us an example, please.

[Niko]

I've used AI twice, both times for work... translating code from one programming language to another worked really well, no complaints, huge efficiency gain - it's a pity I only need to do something like this every few years.... then again, maybe I should be happy. 

The second was to find the most suitable experts on a given topic ("convex analysis" iirc) in a list of 50 names. That's a task AI should be good at, much more efficient than a human, but the thing we were using (ChatGPT iirc) did an incredibly poor job, confusing one person with someone with a similar last name and generally not looking further than the first handful of names... One annoying feature I noticed is that if it gets names in the form "A. Smith, B. Miller" it will just talk about "Alfred Smith" and "Ben Miller" as if those were these people's first names... but in fact the few names that I knew were all wrong, so - it does look up the correct biography of "A. Smith" but then plugs in a random but plausible first name... Of course, with further pushing the quality of all these replies could be improved and the system will apologize excessively while correcting it's errors...You have to really like working with interns to enjoy this... 

I can easily how someone who doesn't actually care about the quality of the result will be quite happy with the AI system... For the majority of tasks, my feeling is that the biggest gains are for people who found google too abstract to really benefit from it...  But with google, you were lead to a website and then could decide for yourself whether to trust it or not... Now many people may indeed no longer leave the AI and get all information filtered and possibly randomly altered through that system... which is pretty worrying 

[Guy Berger]

20 hours ago, Rabshakeh said:

Give us an example, please.

So I’m prompting a series of humorous stories about my kids being students at a magical academy, with all sorts of zany adventures.  A recent one is a LotR parody, they go on a quest to destroy a powerful magic ringpop.  None of them are good children’s literature but they entertain my kids.

excerpt:

”

“I need more!” Sauron declared. “The One Ring Pop must not be lonely! I must surround it with Nine Grape Rings for the mortal kids, Seven Blue Raspberry for the sugar elves, and Three Watermelon Blasts for the camp counselors doomed to night duty!”

 

He stood up and declaimed:

 

 

 

🧁

“Three Rings for the Counselors, soggy and tired,

Seven for Elven teens with TikTok inspired,

Nine for Campers doomed to sticky-fingered fate,

One for the Dark Lord on his candy crate,

In the Land of Wizzlewand where the Snack Flames lie,

One Ring Pop to rule them all, and in the sugar bind them,

One to lure the children in and Ring Pop-ly remind them…

That cherry is superior. Fight me.”

🧁”

Edited June 18 by Guy Berger

[Rabshakeh]

1 hour ago, Guy Berger said:

So I’m prompting a series of humorous stories about my kids being students at a magical academy, with all sorts of zany adventures.  A recent one is a LotR parody, they go on a quest to destroy a powerful magic ringpop.  None of them are good children’s literature but they entertain my kids.

excerpt:

”

“I need more!” Sauron declared. “The One Ring Pop must not be lonely! I must surround it with Nine Grape Rings for the mortal kids, Seven Blue Raspberry for the sugar elves, and Three Watermelon Blasts for the camp counselors doomed to night duty!”

 

He stood up and declaimed:

 

 

 

🧁

“Three Rings for the Counselors, soggy and tired,

Seven for Elven teens with TikTok inspired,

Nine for Campers doomed to sticky-fingered fate,

One for the Dark Lord on his candy crate,

In the Land of Wizzlewand where the Snack Flames lie,

One Ring Pop to rule them all, and in the sugar bind them,

One to lure the children in and Ring Pop-ly remind them…

That cherry is superior. Fight me.”

🧁”

I gotcha 

[Pim]

i use ChatGPT mainly to alleviate the annoying and boring aspects of my work. in education we have to justify a lot of things in writing. the mountain of policy documents keeps growing and although you are obliged to have it all on paper, there is not a soul who reads it... it is mind-numbing and demotivating. because I am half in front of the class and half in management as a department head, it takes up an even larger part of my work. in this area AI is real and a godsend. it produces high-quality documents, can critically review and test them and gives good instructions in areas that I know less about. it really saves me hours of work that I can now put into my lessons and students. in addition, I notice that ChatGPT can search much more efficiently and specifically than a search engine.

another way I use it is to clear up difficult passages in philosophy books I read. For exception now with Hegels book, it provides excellent explanations and summary’s which make me able to check if I actually understood what I’ve read.
 

it could also be very useful with jazz music. You could ask for 10 albums where Evan Parker plays solo saxophone or on which albums a specific jazz musician played with another. of course I realize that ChatGPT makes mistakes. when I asked as a trick question whether Charlie Parker and Coltrane had ever made a recording together, he replied that they had indeed 🫣

Edited June 19 by Pim

[Teasing the Korean]

I've never personally interacted with Al, but he is touring with Brian's backing band, and his son Matt beautifully sings Brian's high parts, so it should be fun.

[Royal Oak]

I had quite an interesting interaction with AI today, with a jazz flavour.

I was browsing Ebay for records the other day, and came upon a collection of 100 LPs. They were close to where I live so I messaged the seller to see if he had a list, or any more pics. He responded with a half-finished list of around 45 LPs.

This excerpt contained all manner of the "big" modern jazz titles eg  Speak No Evil, Song for My Father, Sidewinder, Idle Moments, Cool Struttin', Red Clay, Waltz for Debby, Brilliant Corners, Kind of Blue, Ah Um, Time Out, Saxophone Colossus, Blues & The Abstract Truth etc. A little strange, but I figured this must have been a curated collection, perhaps influenced by those "100 jazz albums you MUST listen to before you die" lists. In between the classics were the usual tiddlers; Oscar Peterson, Pablos, Lionel Hampton on Vogue, Jasmine reissues, Paul Desmond on A&M, Basie and Ellington etc.

I bought the collection and arranged pick up for today. I didn't check the LPs on pick up, as I didn't feel the need. Anyway, once I opened the box when I got home, there were indeed 100 LPs, but all those big titles were missing. In the past I've bought LP collections off Ebay where the seller has removed items between listing and handing over, so I messaged the seller to tell him the LPs were missing.

He was mystified, and sent me pictures of all the LPs in the collection, all in my box. It turned out that he had taken pics of every LP and uploaded them to Chat GPT to create a catalogue, presumably to save himself the bother of typing all that information himself. Chat GPT then simply invented a list of 100 jazz albums, pulling it out of its metaphorical arse. The seller, totally unfamiliar with jazz, thought nothing of it.

He showed me a screenshot of him accusing Chat GPT of fabricating the list of records.

It answered "You're right (name) I shouldn't invent records. Thanks for catching that"

The seller offered a return, which I declined as the collection was decent anyway, so no harm done.  Good job it wasn't reading my CT brain scan, or telling me whether the mushroom I foraged was poisonous....

 

Edited 1 hour ago by Royal Oak