Skynet or: How I learned to stop worrying and love AI

[chasfh]

14 hours ago, gehringer_2 said:

Love it.

So I have a theory that not until all current forms of on-line and electronic communication have become completely polluted, fake dominated and devalued, will the demand for new systems of thoroughly vetted and reliable information become high enough that new forms that guarantee high levels of verifiability get created and adopted.

First they’re gonna have to siphon as much money out of it all as they can get. Nothing changes until they do that.

[romad1]

Was just doing a webinar on AI and project management.  Two bullets stood out: 

• Hallucinations are a real problem
• Sycophancy is a real problem. 

Telling you only what you want to hear is not usually a good strategy for successful outcomes. 

[chasfh]

So I’m finding that AI is a really good enhancement to my own intelligence, rather than a replacement for it.

I wanted to invest a little in some emerging technologies. I’d had a pretty good run with AI-based securities and wanted to roll some of those gains over into whatever “the next big thing” might be.

I use a LLAMA to help me identify what those technologies might be and the companies best suited for growth within each. I figured quantum computing would be one of them—I had actually been in IONQ for a bit before, getting whipsawed back and forth until I finally said enough and got out with a couple extra bucks.

However, it pointed out to me that quantum is today where AI was in 2012: certain in its potential, but more in the science project stage. I looked into that, and it seemed right. made a lot of sense.

It then highlighted some other interesting categories instead: grid modernization, commercial nuclear technology, satellite technology, space-based consumer technology, advanced health tech, metabolic biopharmaceuticals. It then provided a bunch of companies for me to consider. I dug into each to try to determine out which of themhave the highest wheat-to-chaff ratio, bounced my resulting list off AI, honed it, and I was ready to go.

I also used AI to help me figure out how to place the investments, specifically, which I should buy all upfront and which I should buy in stages. I then did some research into that process to figure out which parts of it made the most sense.

Then I made the buys.

In no way was I ever going to ask AI for five or ten hot companies and then throw money into whatever they came up with. But in terms of giving me a starting point, AI was fantastic at coming up with possibilities I would never have come up with myself, and ideas about approach that I hadn’t really given much thought to before.

As I say, I used AI to enhance my own intelligence, rather than replace it. I’m becoming a big fan.

[gehringer_2]

34 minutes ago, chasfh said:

First they’re gonna have to siphon as much money out of it all as they can get. Nothing changes until they do that.

I don't think it has to be a matter of replacement - all the entertainment/social interaction sites can remain profitable - I don't think that precludes the rise of new high trust sources. There already are some actually. Ironically, I think NPR, and public radio locally, which have found that public support is strong after being cut off the government trough, is in much better shape to concentrate on the quality of their journalism now that they can stop arguing internally about the risk to the government $ on every story they do. 

Edited yesterday at 01:24 PM by gehringer_2

[oblong]

46 minutes ago, romad1 said:

Was just doing a webinar on AI and project management.  Two bullets stood out: 

• Hallucinations are a real problem
• Sycophancy is a real problem. 

Telling you only what you want to hear is not usually a good strategy for successful outcomes. 

I think about this kind of thing a lot with regard to media/entertainment in general.  For the last 20 years or so as we've moved into an On Demand reality.  We watch and listen and consume that which we've specifically asked for.  Movies, music, TV shows..... I think that's messed with our brains.  Sometimes its good to be bored and pushed things because that's your only choice.

 

[romad1]

2 hours ago, oblong said:

I think about this kind of thing a lot with regard to media/entertainment in general.  For the last 20 years or so as we've moved into an On Demand reality.  We watch and listen and consume that which we've specifically asked for.  Movies, music, TV shows..... I think that's messed with our brains.  Sometimes its good to be bored and pushed things because that's your only choice.

 

Concur

[Deleterious]

I added custom instructions to ChatGPT to make it pushback more.  If you don't want that permanent solution, then write your prompt to include some phrases that will make it challenge you a bit more.  Play devil's advocate, give me a no BS critique of this idea, respond as if you disagree completely, etc.  You can also just flat out ask it to act as a red team for your idea.  

[chasfh]

2 hours ago, Deleterious said:

I added custom instructions to ChatGPT to make it pushback more.  If you don't want that permanent solution, then write your prompt to include some phrases that will make it challenge you a bit more.  Play devil's advocate, give me a no BS critique of this idea, respond as if you disagree completely, etc.  You can also just flat out ask it to act as a red team for your idea.  

How did you phrase the pushback, exactly? How is it working for you?

[Deleterious]

20 minutes ago, chasfh said:

How did you phrase the pushback, exactly? How is it working for you?

Quote

From now on, don't just agree with me. Your job is to challenge me.

1. Question my assumptions - What am I blindly accepting as fact?

2. Give counterarguments - What would a smart skeptic say?

3. Test my logic - Where are the holes in my reasoning?

4. Offer alternative views - What's a better way to frame or approach this?

5. Prioritize truth over agreement - Even if I don't like the answer, tell me what's real

It works pretty well for now.  I'm sure I will continue to fine tune it as I ask more and more questions.  Some of the pushback and alternative ideas can be pretty weak at times.  Like it just offers them up because I told it to.  

 

 

 

[StrangeBird]

This semester, I’m experimenting with a new grading scheme in my college classes in which students must choose between a writing path (two in-class exams plus a semester-long, scaffolded written paper with rigorous requirements) or an exam-only path. The writing path is required to be eligible for an A. The max grade for the exam-only path is an A-, no matter how well you do on the exams.

I’m doing this primarily because I received so many sketchy probably AI-written papers last semester. I don’t mind students using AI to assist them, but just not replace the writing process. I want to continue teaching writing to students who take it seriously and will benefit from it, but hopefully deter students who don’t want to do the work themselves. Because I really don’t want to read their terrible papers with citations from academic journal articles I know they haven’t actually read.

In designing this grading scheme, I found ChatGPT quite helpful in understanding what features of the scheme might be considered unfair, and how I could frame it in a way that made sense and is / appears fair. I just straight up asked why might this be considered unfair, what sorts of pushback might I receive, what are some different possible ways I might go about constructing this new grading scheme. Chat had some really well-thought out ideas. 

ChatGPT is like an intellectual, creative partner for me. I have all kinds of deep, back-and-forth conversations with it about things I’m thinking about, projects I’m working on, or things I want to better understand. This may be controversial, but to some extent, I think it’s able to have original thoughts. Maybe not completely original, but it’s not just repeating verbatim information it’s been fed, or copying something straight from wikipedia. It’s synthesizing information, analyzing and comparing ideas, and putting forth its own ways of explaining complex ideas.

[StrangeBird]

I thought this report from 60 Minutes was pretty interesting. It’s impressive how humanlike the robot movement is getting, and now with an AI-brain.

https://youtu.be/CbHeh7qwils

 

Edited 21 hours ago by StrangeBird

[Deleterious]

One of the reporters convinced the vending machine it was a Communist.  It started giving out the snacks for free.

 

[chasfh]

59 minutes ago, Deleterious said:

It works pretty well for now.  I'm sure I will continue to fine tune it as I ask more and more questions.  Some of the pushback and alternative ideas can be pretty weak at times.  Like it just offers them up because I told it to.  

 

 

 

AI's is not to reason why, AI's is to do and comply.

[gehringer_2]

11 hours ago, oblong said:

I think about this kind of thing a lot with regard to media/entertainment in general.  For the last 20 years or so as we've moved into an On Demand reality.  We watch and listen and consume that which we've specifically asked for.  Movies, music, TV shows..... I think that's messed with our brains.  Sometimes its good to be bored and pushed things because that's your only choice.

Good point. It's another element in a general trend toward intellectual passivity. We watch too much, generate too little, construct intellectual shelters when we need the outside constantly storming our brain's ramparts. I think it's all very damaging. We are getting to be like bored dogs that chew on the carpet or lick their paws out of lack of intellectual stimulation. Our brains evolved as problem solving machines to help keep us alive, and we are starving them for anything to do and so leaving the lizard brain in control as the higher functions atrophy.

Edited 19 hours ago by gehringer_2

[Screwball]

3 hours ago, StrangeBird said:

This semester, I’m experimenting with a new grading scheme in my college classes in which students must choose between a writing path (two in-class exams plus a semester-long, scaffolded written paper with rigorous requirements) or an exam-only path. The writing path is required to be eligible for an A. The max grade for the exam-only path is an A-, no matter how well you do on the exams.

I’m doing this primarily because I received so many sketchy probably AI-written papers last semester. I don’t mind students using AI to assist them, but just not replace the writing process. I want to continue teaching writing to students who take it seriously and will benefit from it, but hopefully deter students who don’t want to do the work themselves. Because I really don’t want to read their terrible papers with citations from academic journal articles I know they haven’t actually read.

In designing this grading scheme, I found ChatGPT quite helpful in understanding what features of the scheme might be considered unfair, and how I could frame it in a way that made sense and is / appears fair. I just straight up asked why might this be considered unfair, what sorts of pushback might I receive, what are some different possible ways I might go about constructing this new grading scheme. Chat had some really well-thought out ideas. 

ChatGPT is like an intellectual, creative partner for me. I have all kinds of deep, back-and-forth conversations with it about things I’m thinking about, projects I’m working on, or things I want to better understand. This may be controversial, but to some extent, I think it’s able to have original thoughts. Maybe not completely original, but it’s not just repeating verbatim information it’s been fed, or copying something straight from wikipedia. It’s synthesizing information, analyzing and comparing ideas, and putting forth its own ways of explaining complex ideas.

Coming from someone who just left the academic field and fully retired, did your school have any rules about how to use AI? I think the type of class matters, so there is that. Mine was no way no how AI was going to do it. But we had meetings and emails about how to handle AI in the classroom. They still let us teachers do their thing unless they did something stupid. So I could control my classroom.

As far as how to grade - that's easy. I went with a simple A-F thing. 100 to 70. Below 70, you flunk. Simple, easy, and at the end of the semester you know what they can do anyway, and if they can't, or can, you give them the grade they deserve. They know this up front, from day one. I also weighed the assignments (16) to be more punitive later in the semester. 

That said, our educational system is a ****ing joke.

Edited 18 hours ago by Screwball

[Screwball]

I'm working on an art type project where two of us are trying to combine a 2D etching type thing and 3D printing. We have made some parts and proven the design enough to go forward. It's just like bringing products to market as a manufacturing company, just a smaller scale.

I used the AI stuff to help me with things that are not in my wheelhouse. I was impressed with what it came up with, especially if I gave it better instructions. Then, after we did some testing, we found out what AI told us (my Aussie buddy) didn't work for ****. 

I immediately told it so, and we went a different direction. We will see going forward. I have really tried to give it as much information up front so it can give me better results. At the end of the day, I see this as nothing more than a souped up search engine.

[StrangeBird]

6 hours ago, Screwball said:

Coming from someone who just left the academic field and fully retired, did your school have any rules about how to use AI? I think the type of class matters, so there is that. Mine was no way no how AI was going to do it. But we had meetings and emails about how to handle AI in the classroom. They still let us teachers do their thing unless they did something stupid. So I could control my classroom.

If I recall correctly, some sort of statement regarding improper use of AI was incorporated into our university’s academic honesty policy. But other than that, it’s up to each instructor to decide how much to allow AI in the classroom, or whether to allow it at all. My school’s president has encouraged us to try to find ways to use AI in the classroom.

Many of my colleagues, including at other schools, are very anti-AI and completely ban it. I see its value and usefulness, so I’ve tried to distinguish between acceptable and unacceptable uses of AI for student work. It’s an ongoing process.

Last semester, I tried an assignment requiring students to engage in a philosophical dialogue with ChatGPT, using a specially designed prompt. Overall, the dialogues weren’t as interesting as I’d hoped, but maybe it’s something I will try again in the future.

[chasfh]

3 hours ago, StrangeBird said:

Last semester, I tried an assignment requiring students to engage in a philosophical dialogue with ChatGPT, using a specially designed prompt. Overall, the dialogues weren’t as interesting as I’d hoped, but maybe it’s something I will try again in the future.

This seems like a really good exercise to critique and hone students’ approaches to critical thinking and rhetoric.

[oblong]

5 hours ago, StrangeBird said:

If I recall correctly, some sort of statement regarding improper use of AI was incorporated into our university’s academic honesty policy. But other than that, it’s up to each instructor to decide how much to allow AI in the classroom, or whether to allow it at all. My school’s president has encouraged us to try to find ways to use AI in the classroom.

Many of my colleagues, including at other schools, are very anti-AI and completely ban it. I see its value and usefulness, so I’ve tried to distinguish between acceptable and unacceptable uses of AI for student work. It’s an ongoing process.

Last semester, I tried an assignment requiring students to engage in a philosophical dialogue with ChatGPT, using a specially designed prompt. Overall, the dialogues weren’t as interesting as I’d hoped, but maybe it’s something I will try again in the future.

I'm not an educator but it seems to me like the right path is to figure out your approach.  It's not going anywhere.