Episode 233: understanding AI - part 2 - Aligning AI and Principles of use
Episode 233: understanding AI - part 2 - Aligning AI and Principles of use
Embedded Files
Understanding AI - Part 2 - Aligning AI and Principles of use - episode 233
Understanding AI - Part 2 - Aligning AI and Principles of use - episode 233
“Focus on the practical use of AI” I continue the series to Understand AI. This week I share the topic of Aligning AI and mention the 4 principles suggested by Ethan Mollick in the book, Co-Intelligence - Living and working with AI. I begin by sharing the idea of the paper clip theory suggested by a philosopher. The worry that AI could either create the apocalypse or our salvation. Then I go through each of the principles Mollick suggests as a way to guide us in our use of AI.
Show Notes: Hi Friends! I hope you enjoyed listening to this episode. Below are all the references.
Quote From the Episode
Quote From the Episode
“It’s easier to think of AI as a person because it behaves more like a person than a computer software that we're used to… but also still remember that it’s not an actual person.”
Key Topics
Key Topics
How large language models (LLMs) work
Why people fear AI (apocalypse vs. salvation thinking)
The “paperclip experiment” thought experiment
The importance of alignment in AI
How everyday users contribute to AI training
The need for boundaries when using ChatGPT
Four principles for using AI responsibly
Examples of prompting and personas
Planning real-life tasks with AI assistance (like landscaping)
Four Principles of Using AI (as explained in the episode)
Four Principles of Using AI (as explained in the episode)
Always invite AI to the table – try it, experiment with it, and learn how it can help with tasks.
Be the human in the loop – remember AI predicts, guesses, and can hallucinate; your judgment matters.
Treat AI like a person, but describe who it is – give it a persona with context, constraints, and expectations.
Assume it’s the worst version of AI you’ll ever use – expect rapid change, stay flexible, and keep learning.
Sources
Sources
Sources & Mentions
Sources & Mentions
Co-Intelligence: Living and Working with AI — Ethan Mollick
General Conference, The Church of Jesus Christ of Latter-day Saints
Camille’s mention of upcoming landscaping project + using AI to help design it
Transcript
Transcript
I am Camille Johnson, and this is Finding the Floor.
Stories and reflections of midlife motherhood, family, and finding meaning in it all.
Join me as I share a little piece of my life and figure out what I want to be when I grow up.
Hey friends, welcome to Finding the Floor.
And today I'm going to continue the series about learning about AI.
And today we're going to talk a little bit about aligning AI and principles of use.
Can you believe we are, we're in October.
It is finally feeling like fall here.
And it was our general conference.
And so you'd have breaks in between.
And it was funny because my daughter was wearing a sweater and it was like in the 80s.
And I was like, aren't you going to just
And she's like, I am protesting.
It should be cooler fall weather right now.
But then this week, it's cooled down.
So fall in Cleveland is the best.
So it is now when the leaves are changing and there's so many fun fall things to pick apples.
And I just feel like there's a lot in this area.
And has anybody been thinking about Halloween costumes?
I've loved doing family costumes over the years.
They haven't always like worked out as some of my kids wouldn't want to go along, which was fine.
almost all of us one way and then someone else who just had to be this particular thing.
So we don't know, we're kind of struggling.
It's so funny now that we're back down to just one kid at home.
The teenager image, it's just so funny.
So we'll figure it out, but it's just a fun little dilemma.
We just, we have fun dressing up and our ward has like a trick or treat.
We just have to hope and pray that maybe it's still nice.
temperature drops and it's snowing or freezing rain.
does and then how maybe I can use it.
I've started using AI to help me with my summaries at my website.
I figured out how to make a transcript, which was really cool on a Microsoft Word.
There's this way that you can load an audio file and it'll make the transcript.
So I'm slowly trying to add transcripts to my
So that's been kind of nice so that I don't feel overwhelmed on creating that.
We've already kind of made this format for each webpage.
So that's something that I've been doing and I have another thing that I'm going to try next week.
I'm going to see if I can use ChatGPT as I try to re-figure out my
front landscape because it's always getting eaten by the deer.
And by mid-summer it just is eaten and then just weeds grow.
So I've been doing some research on deer resistant plants and all that.
Like I'm going to see if it can plug in a picture with measurements.
I just feel like, why don't I use it in a way that really helps me make my life a little bit easier?
So to begin, last time we left off with an understanding of what large language models are, or LLMs.
And that is mostly what ChatGPT uses.
When you search in Google, you're typing lots of words, and that's what their AI uses.
And then AI has developed and been trained to then create something because of those weights.
and it's kind of mathematical when it goes back to itself and then spits back out.
And so there is this belief that AI can either
be apocalyptic and destroy all humans if it gets to a certain point.
And Ethan Molock kind of talked about, you have on one hand the apocalypse and on another salvation.
And he gives this analogy of the paperclip experiment.
So this philosopher came up with this idea of the paperclip.
So there's an AI that's all its job is to make paperclips.
But then this AI becomes a super AI, meaning
Then it also kills off all the humans because they might prevent it from making more paper clips.
a super intelligence which is like smarter than a human.
As of now, those two things have not happened.
So what Ethan Moloch is saying is that we're going to not worry about those ideas of the apocalypse.
And how many of us have watched movies where the AI takes over?
I think of like the Avengers movie where they make
But what Ethan Mollick says is that what AI needs is it needs to be aligned in the right way.
Because it's way expensive for companies to try to experiment with all these types of ideas with AI.
But if we choose to learn to use it, then we are also adding that input to the training of AI.
less bias and represent the whole world, if that makes sense.
Who knows how it will be helpful, but we have to try to use it.
And the hardest thing about that is that there's no
Like we were working on my ideas for my website and how to format it.
And then it was just adding more and more ideas.
Well, would you want to do this?
And I was like, oh, can you show me maybe how this would look?
So it would do that and then keep on asking me questions.
And I was just like, whoa, I've gone way further than I thought I was going to.
So I think all of that is helpful to like
All right, so first principle is always invite AI to the table.
He's saying, figure out AI and see how you like it as you are doing some of your tasks.
Try to find ways where you can.
use, an AI assistant or ways that you can begin to use it that will help you.
which I still don't understand how it learns.
That's a podcast for a whole nother day and maybe philosophy.
But so principle one, always invite AI to the table.
And then principle #2 is be the human in the loop.
How AI is trained on using AI judgment is they don't actually know something.
They are predicting something.
And the one thing to know about AI is that they want to please you and make you happy.
And it always says this at the bottom of ChatGPT, like can give you incorrect answers.
They'll just come up with things that aren't even real.
because they're trying to please.
So the second principle is be the human in the loop.
Third principle is treat the AI like a person, but describe who it is.
You can't quite understand it, but it can understand you.
And so in this way, it's saying treat it like a person, but describe who it is.
You want to tell how you want it to think and from what perspective.
It's unaware of the real world, but can learn from your feedback.
It's suggestible and gullible.
So you need to give clear and specific AI persona
in your questions to help you with getting the right answer.
And it needs context and restraints.
And then you kind of refine and redirect if you don't like what the answer it is giving you.
And so it was really helpful that there were ways that
I want to make the most of like my marketing strategy or whatever.
So my guess is it probably would be good to have AI
be the persona of an expert landscaper with knowledge about deer resistant plants.
It is important to say that you
Again, think of AI as a person.
And that's one thing that you have to remember that AI is also not a person.
And he talks about that, but it's easier to think of AI as a person
because it behaves more like a person than a computer software that we're used to.
But then also still remember that it's not an actual person.
Page updated
Google Sites
Report abuse