Rubric Prompt

Prompt Library

Essential Guide to AI for Educators Course

Educating All Learners Alliance

Amanda Bickerstaff

Amanda is the Founder and CEO of AI for Education. A former high school science teacher and EdTech executive with over 20 years of experience in the education sector, she has a deep understanding of the challenges and opportunities that AI can offer. She is a frequent consultant, speaker, and writer on the topic of AI in education, leading workshops and professional learning across both K12 and Higher Ed. Amanda is committed to helping schools and teachers maximize their potential through the ethical and equitable adoption of AI.

Hi everyone, it's Amanda from AI for Education. We have quite a few people joining, so we're gonna have everyone a little bit of time because it's one by one with Zoom, but we're so excited to have you here today.

We just are very excited, but we're going to give everyone just a couple of like a half a minute left before we get started.

But just really appreciate you all being here, whether it's the middle of your day, if it's the evening, if you're taking some time away from the classroom, we just really appreciate you being here.

And we really have our first hello and this is from Budapest. It will be a pretty, a pretty diverse panel.

Like make sure that you're responding to everyone as well. We love to say Clayton, that's my mom, hi mom, always very supportive.

Also New York City Apartments where they buzz every door so hopefully that will not keep happening but very excited to have you here.

Will actually get started. So I want to take it over to Tria, who's going to share like why we're here today.

Hi everyone, thank you so much for joining us. My name is Tria Hutchings. I am the project director for the educating all learners alliance.

We call it ELA for short sometime. So we are a organization of 140 partners. For a part of our alliance each that are committed to supporting students with disabilities and learning differences and we have partnerships from across the United States, across sector.

So we're really excited to be here today and in partnership with AI for Education to bring you this series.

We're hoping that this can answer some questions you might have. And provide a basis for your AI and special education.

Amazing. And we just want to say thank you so much. This was an amazing opportunity for us to really dig into something we care a lot about here at AI for Education, which is a focus on diverse learners and equity.

And I think that this is an amazing opportunity. We know from research that students with IAPs and 5 before plans actually tend to use these technologies more because of the accommodation.

So really excited to have you all here. As I said, I'm Amanda. I'm the CEO and I'm Amanda.

I'm the CEO and co-founder of AI for Education. We're already approaching 300 people watching.

We're already approaching 300 people watching, which is amazing. If you've been with us before or this is your first time, we're already approaching 300 people watching, which is amazing.

If you've been with us before or this is your first time, I want to set this scene. We want this to be interactive.

The chat is your best friend. I want you to get involved. We've got some like missing facts and if you've been with me before you might be familiar with that approach but please use that QA function to gauge with your peers.

This is a community of practice. And we have 4 sessions together that are really amazing for us to be able to do.

So please get involved, say hello. And that also want your problem with us. We'll be using Chat GPT today.

So if you have a Chat GPT account, please feel free. We'll be sharing some prompts and we'll be doing some work together.

And then lastly, if you have a great resource, please share. And I just want to say thank you again to Treah.

I'm gonna have you come off the video. Say, and we'll come however come back on at the end, but we're gonna get started.

We only have an hour together and so just want to say first thank you for everyone joining. We have a super diverse group of participants so love having you here and let's go.

So first I'm gonna ask you to take out your phone. Okay I have actually 2 phones. I have the one that is new and the one that I broke.

And so this one has not been broken yet, so give me time. But what I want to start with is while we start every single introduction genera AI, which is we have experienced with AI every day.

So we might be talking about Chat and other tools, but we have AI in our pocket right now.

And so while it might seem like, oh my gosh, right now. And so while it might seem like, oh my gosh, how did this happen in the last year, how did this happen in the last year? We have had artificial intelligence around us for decades.

And so for me, I have had our official intelligence around us for decades. And so for me, I have a nice handy dandy iPhone.

It recognizes my face. And so for me, I have a nice handy dandy iPhone. It recognizes my face.

And the way, I have a nice handy dandy iPhone. It recognizes my face and the way it does that is it creates a biometric signature of my face and the way it does that is it creates a biometric signature of my face that actually changes every time I look at the technology because maybe, if I'm a guy, I'm doing Movember, I have a suspect mustache or I'm wearing a hat

or slow lighting. And that's a form of artificial intelligence is going to be able to create essentially a signature in my face and then do an action which is this case it opens my phone.

And, so, I, want, to, say, the, AI, is, part, of, our, lives, in, fact, 80 Your fingerprint lock, absolutely. My computer has my fingerprint lock.

So I'll go through a couple of these that are pretty, pretty So this is the opportunity that if you're on a highway and there's a major set of construction, it can reroute you based on real-time data.

It can also tell you, what's the best way to go and or it's quite funny.

It'll say you have a meeting and you know. X amount of time, you need to leave before.

So we have all kinds of artificial intelligence that's part of our maps. We've got things like predictive text and that's why I can never say anything but duck when I'm having a bad day.

00:05:32.000 --> 00:05:43.000

It's going to allow us to complete sentences to complete, you know, do our wonderful. You know, auto correct and help us with spell check.

00:05:43.000 --> 00:06:03.000

We've got things like we talked about in the chat like Snapchat or Instagram or if you're a TikTok or who my TikTokers on the participants, we have all kinds of opportunities where it learns that about you, learns about how you use these tools to be able to provide you recommendations on how to do this.

00:06:03.000 --> 00:06:11.000

And they're also all kinds of predictive and analytical pieces that happen for our students with special with disabilities as well.

00:06:11.000 --> 00:06:18.000

We have lots of different tools that use machine learning. So a lot of tools within your classroom will also have machine learning right now.

00:06:18.000 --> 00:06:27.000

So thank you for playing along. We have a ton of great examples on our phone and we're gonna move forward and we're gonna do a myth and facts and so I know there are almost 400 of us here.

00:06:27.000 --> 00:06:35.000

So it's gonna be fast, we're gonna go fast. But if you believe that this is a myth and when you type an M in the chat and if you believe it's a fact I want you to type an F.

00:06:35.000 --> 00:06:41.000

If you're not sure I can't see you so it's okay if I could see you would ask you to do this.

00:06:41.000 --> 00:06:47.000

But I want you guys to engage as much as possible. So we're gonna go first one. AI can think like humans.

00:06:47.000 --> 00:06:58.000

Is that a myth or a fact? Okay, can already see we have we have mostly missed we definitely have some facts in there so give everyone just another 10 s to answer.

00:06:58.000 --> 00:07:04.000

Okay, so mostly myth and what we have here is it actually is a myth. And this is a first thing.

00:07:04.000 --> 00:07:08.000

If I had you in a room and I want you say it out loud to yourselves. AI is not thinking, it is computing, it is predicting, it is probabilities.

00:07:08.000 --> 00:07:21.000

And so it is designed to make you think it is thinking. But it is not at this stage. We have not created Skynet or data from Star Trek.

00:07:21.000 --> 00:07:32.000

But what we have is these tools that are able to do human like things. And so in this case, because chat QT is a conversational AI, it often feels like it is thinking.

00:07:32.000 --> 00:07:40.000

It has a personality. It says I, but reality is it is not thinking it is computing and we'll talk more about that in a section on the technology.

00:07:40.000 --> 00:07:54.000

But this is a really important idea for us to understand because it really is an underpinning of misconceptions and it's not just for people in our room that we have together, but also students can also think that these tools are thinking not computing.

00:07:54.000 --> 00:08:01.000

But the good news is they are computing so we can learn how to use them to our advantage and also under demystify this moment and time.

00:08:01.000 --> 00:08:06.000

And so next is we're going to go to AI is unbiased. So what do we think? Oh, this is fast.

00:08:06.000 --> 00:08:18.000

So we have mostly Miss. Love this group. I saw some people familiar faces in the in the crowd as well Okay, AI is going to is going to be pretty biased.

00:08:18.000 --> 00:08:24.000

And so I love that you all know this because your working as special educators are working around this world and it's incredibly important.

00:08:24.000 --> 00:08:43.000

AI is trained on the internet. And the internet has never been called an unbiased place. You know, we don't, we don't talk about, you know, how the internet is a really balanced place for both sides of the aisle.

00:08:43.000 --> 00:08:56.000

It is actually a place that really exacerbates and highlights our existing biases. And I think this is really, really important to understand because not just generative AI systems, but all AI systems have shown bias.

00:08:56.000 --> 00:09:12.000

And so whether that is, for example, the work of Ballymani, Joy, who has done work around AI bias since 2,018 on how facial recognition software does a much worse job of non-white non-male faces.

00:09:12.000 --> 00:09:18.000

And that's a form of bias because the training datasets were primarily faces of white males.

00:09:18.000 --> 00:09:24.000

We also have a lot of bias that's implicit. We have all, you know, the majority of the internet is in English.

00:09:24.000 --> 00:09:41.000

It is primarily global north. It is going to have pretty much every ism in represented and so it's really important for us to understand that because of the way these models are trained there is no way to necessarily get completely passed by us at this moment.

00:09:41.000 --> 00:09:50.000

But there are opportunities to build less biased systems by creating better data sets by doing better job of creating systems and training systems.

00:09:50.000 --> 00:10:03.000

So I think those are the while right now, especially if you've ever used a text to image generator, you know, you can see that if you ask for, you know, a CEO, an image of a CEO, it primarily will be a white male.

00:10:03.000 --> 00:10:10.000

And that's, and even if you give it some support, it's going to constantly do that because the data set says that.

00:10:10.000 --> 00:10:18.000

So very important for us to understand is especially when we're thinking about working with these systems. And art says it will also be an older white male.

00:10:18.000 --> 00:10:24.000

Yeah, we definitely will go into a little bit more about bias. So that is an important component to understand.

00:10:24.000 --> 00:10:31.000

The next question is AI detection tools are effective. Do we think this is a myth or a fact?

00:10:31.000 --> 00:10:39.000

Okay, see this one's a little bit splitting the room just a bit. Okay, so, what if I said, do you want them to be effective as educators?

00:10:39.000 --> 00:10:51.000

I think are, we would want, we have a little bit of a difference where I wish I could tell you right now that these AI detectors are going to be reliable, but they aren't reliable at this stage.

00:10:51.000 --> 00:11:12.000

They have both false positives and false negative. So false positives are going to be where it says AI has generated a text but it has not and then it also has false negatives where you have a pretty high proportion up to 20% of AI generated work actually running right through these systems.

00:11:12.000 --> 00:11:21.000

So whether that's turned in, GPT 0, copylex, etc. They both have false positives and false negatives that they are that they're open to sharing.

00:11:21.000 --> 00:11:31.000

And this is something that's really interesting. I have been in many rooms where I get to do this live and I will tell you I've had superintendents, I've had leaders and teachers where their students, their child has been accused of AI cheating.

00:11:31.000 --> 00:11:43.000

And so we think this is a really important distinction to make why AI detection tools may do they do a good job of catching bad AI use and bad AI cheating.

00:11:43.000 --> 00:11:50.000

The same way that you probably are aware of that without using the detector. A student suddenly has a different tone or vocabulary.

00:11:50.000 --> 00:12:03.000

It's a pretty strong signal that that's student did not complete that work. But it is something that there is no magic bullet, there are no x-ray glasses, there is no digital signature that allows for an AI detector to be fully reliable.

00:12:03.000 --> 00:12:16.000

And for example, in July of 2,023, Open AI had an AI detector about chat, BT and what they found is that it couldn't detect the own company could not detect AI writing and they actually closed the tool.

00:12:16.000 --> 00:12:23.000

So this is something where we we say, please don't overly rely on these tools. Because they are not reliable.

00:12:23.000 --> 00:12:31.000

And then finally, we have this idea that they can, there has been research that's shown that these tools can be biased against non-native English speakers.

00:12:31.000 --> 00:12:39.000

So we really want to be clear that this is not a. I. Detection in terms of running it through an A detector is not going to be something that we suggest you doing.

00:12:39.000 --> 00:12:50.000

And as I said, there are tools that are looking at proof of effort or proof of originality. We support those tools because those are looking at how students actually work the process over the product.

00:12:50.000 --> 00:12:55.000

Okay, next one. Everybody ready? AI development has reached its peak. What do we think?

00:12:55.000 --> 00:13:00.000

Fast, we're almost at 400 people. This one primarily myths a couple of facts in there.

00:13:00.000 --> 00:13:11.000

Okay, so love this group. And so the answer is Absolutely not. And if you've heard me speak before, you probably heard me say that like right now, Generative AI is the worse it has ever been.

00:13:11.000 --> 00:13:16.000

It's the worst it ever was yesterday and the worse it will ever be tomorrow. It is so new.

00:13:16.000 --> 00:13:23.000

We are in this world in which we are developing technology that we are, that does amazing things and it is happening at an enormous rate of progress.

00:13:23.000 --> 00:13:42.000

And so we are at a very early stage of this AI development, especially around generative AI. So I don't know if anybody knows what I can guess what the number one use of chat to be T is like what do people use it for the most and please do not say cheating on essays even though I know you're thinking it.

00:13:42.000 --> 00:13:51.000

But what do we think that people are actually using? So for our chat GPT. So as an assistant, emails, planning trips, research.

00:13:51.000 --> 00:13:53.000

Yeah, a lot of these are great and these are absolutely parts that they're using. But it actually is creating more technology.

00:13:53.000 --> 00:14:06.000

It's coding. The number one use case up to 40% of some of these tools are used to actually code more technology.

00:14:06.000 --> 00:14:14.000

So what we're seeing is that not only is generative AI and AI getting better and faster, but it also is something in which we're seeing all technologies going faster, which is really, really interesting.

00:14:14.000 --> 00:14:27.000

And I had this quick, like this wonderful image, this graph from Coacher that looks at how AI development has gone and what we have.

00:14:27.000 --> 00:14:36.000

And so we have Handwriting recognition and speech recognition have been something we've been trying to do since before the dotcom bubble.

00:14:36.000 --> 00:14:43.000

And so this is really interesting and the sense that we have, it took like almost 2 decades. It took a really, really long time for AI to develop in a way.

00:14:43.000 --> 00:14:50.000

We were meeting human parity, which is a sign up here. But we see is that we see a piece around image recognition was much faster and that's because it used humans to help train.

00:14:50.000 --> 00:14:56.000

If you ever did a caption that drove you crazy where you're trying to find all the like, you know, the wheels on a bus or something.

00:14:56.000 --> 00:15:13.000

That's an example of how we actually helped train an AI model to recognize images. But then we see some really interesting stuff happen between 2,016 and now where we see these really rapid development cycles.

00:15:13.000 --> 00:15:20.000

And so we see reading comprehension and like a year, language understanding and all these different pieces like common sense completion.

00:15:20.000 --> 00:15:31.000

And the reason being is that compute power increased significantly. And so we saw that like if you were going to fly from New York City, where I am right now to London, it would take 8 h.

00:15:31.000 --> 00:15:42.000

But if you look at the improvement of processing power between 2018 and 2,020 that eight-hour trip would only take you 19 and 2,020, that eight-hour trip would only take you 19 s.

00:15:42.000 --> 00:15:49.000

And that is before the release of GPT. And so this is something that's really like the engine of why we have this new technology happening so quickly.

00:15:49.000 --> 00:15:56.000

And we're going to continue to see it happen even more quickly. So next one, last one.

00:15:56.000 --> 00:16:03.000

AI will replace many professions like doctors, lawyers, and teachers. What do we think?

00:16:03.000 --> 00:16:11.000

Mostly miss a couple of facts. What if I changed it to AI will train or transform many professions like doctors, lawyers, and teachers?

00:16:11.000 --> 00:16:14.000

There we go. I love it. Yes, we know this is happening very, very quickly. We know that is something that this is coming for, you know, all kinds of jobs.

00:16:14.000 --> 00:16:25.000

We see heavily heavy changes that are happening already and creative pursuits. We know what's happening here.

00:16:25.000 --> 00:16:35.000

We have almost we have over 400 people with us today as a great example of that this is something that is fast changing the way that we work and interact with technology.

00:16:35.000 --> 00:16:50.000

And so we know that there are professions like doctors that already like a radiologist. You know, there are any technology that can not only read MRIs and x-rays at the same like diagnostic level of a doctor, but also do it in a lot faster.

00:16:50.000 --> 00:16:53.000

So we see that there are some, you know, professions that will be heavily augmented or changed.

00:16:53.000 --> 00:17:02.000

We see that lawyers already using this for case brief and also teachers. We know that you're here today learn learn more about this and you're probably using it already in your practice.

00:17:02.000 --> 00:17:09.000

And so in that note, I'm actually gonna ask Amanda on my team, another Amanda, who's amazing who helped build this this wonderful presentation.

00:17:09.000 --> 00:17:18.000

Can you drop in the we have a really simple for question. Like poll, so to speak, to see how, where you are, who you are, etc.

00:17:18.000 --> 00:17:19.000

If you don't mind taking this right now, we'd love to know where you are in your own journey.

00:17:19.000 --> 00:17:26.000

But we are excited to be here today to talk to you about how we can start to use these tools for your practice.

00:17:26.000 --> 00:17:32.000

Whether you're using them all the time or you're just getting started. And so I think it's as it needs permission.

00:17:32.000 --> 00:17:34.000

So we probably need to, yeah, just everyone's get a second. I think I manage to make sure that you can share it live.

00:17:34.000 --> 00:17:50.000

So just give, give us a moment, but thank you for the feedback. And what we're gonna do is while we're waiting on that and hopefully it'll be, fixed in a moment is we're gonna get started with using chat should be tape.

00:17:50.000 --> 00:17:55.000

And so I want everyone I'm gonna go over to my email. I'm gonna go over here.

00:17:55.000 --> 00:18:01.000

This is what it should look like and hopefully it's allowed. Here and we'll definitely.

00:18:01.000 --> 00:18:07.000

Okay. Here we go. Here we go. Let's gonna get started.

00:18:07.000 --> 00:18:13.000

So what I'm gonna do is I'm gonna go to chat GBT right now. Okay, thank you, Caitlin.

00:18:13.000 --> 00:18:17.000

It is working. We got it. Go team. Thank you, Amanda.

00:18:17.000 --> 00:18:18.000

So if you just take a moment and then I'm gonna have you come over here. So this is my Chad Shui T.

00:18:18.000 --> 00:18:26.000

I'm gonna take you on a little bit of a you know a little bit of a tour. I have both versions.

00:18:26.000 --> 00:18:33.000

I have 3.5 which is free and I have 4 which is paid. And what we have here is we call this window.

00:18:33.000 --> 00:18:41.000

Like you see I do a lot of these trainings. We call this a context window. And we'll talk about why we call that a window in a context window in a moment.

00:18:41.000 --> 00:18:43.000

But what I want everyone to do right now is you're going to prompt you to put in this prompt.

00:18:43.000 --> 00:18:53.000

What are you chat GBT? So we're all going to do that together and we're gonna do our first.

00:18:53.000 --> 00:19:00.000

Our first lesson for today. So I've done this and it's a long one and we've got we've got 3 paragraphs.

00:19:00.000 --> 00:19:10.000

It's giving me a lot of information about what GPT is. It's been based on GBT 3, train on vast amounts of data.

00:19:10.000 --> 00:19:20.000

You can interact with a lot of different ways. Does everyone have the same answer? So like, do you have, okay, so Dana, thank you, Denaya, sorry.

00:19:20.000 --> 00:19:25.000

You do not. And if you know why this is, then you're on your way.

00:19:25.000 --> 00:19:27.000

This is because this is something called temperature. So the way that, look, Sarah, yours is a little bit lazier than mine.

00:19:27.000 --> 00:19:35.000

This is the longest I've ever seen it and I've done this about a hundred times.

00:19:35.000 --> 00:19:40.000

And the reason why we have this variance across 400 people is that you have Ch GTT is essentially an application that we interact with, right?

00:19:40.000 --> 00:19:54.000

But it has a model underneath it that's set to certain types of parameters. And one of them is called temperature.

00:19:54.000 --> 00:20:05.000

So the temperature of the model is going to be the level of variation or creativity. So for example, if we all got the same thing, that temperature or variation would be a 0.

00:20:05.000 --> 00:20:30.000

But because this is trash GBT and it wants to be, wants you to feel like it's creative and interesting is it's going to be a temperature of about 0 point 7, which means really highly creative, which is great for us because now we can ask it not just for 10 ideas for how we can support students with dyslexia on a reading assignment, but we can also ask it 50 times. And we can keep asking it the same question and we will get a variation of answers.

00:20:30.000 --> 00:20:42.000

And that makes this in great rainstorming partner. Mix it super creative. It makes it something that is really like such a good opportunity to get in there and to start pushing your own thinking.

00:20:42.000 --> 00:20:52.000

But on the other end it means and if you think that you'll get the same output, let's say you were doing this as a demo in front of a classroom or staff room, it will be different.

00:20:52.000 --> 00:20:59.000

So you cannot rely on this tool having the same output, at least for chat, GBT. So it's just really important to think about this.

00:20:59.000 --> 00:21:08.000

There's a lot of heavy variation between the way you use this tool and the way that you can rely on it to be a reliable tool.

00:21:08.000 --> 00:21:15.000

So that's our first piece and now I want everyone to do I'm going to put this back up on the screen but I'm also gonna drop it into the channel.

00:21:15.000 --> 00:21:23.000

So I want you to drop this in to Chat GBT and so I'm gonna do as this as well and we're gonna try that together.

00:21:23.000 --> 00:21:30.000

I'm gonna start a new context window. I always like to start a new context window because the context window is going to have a certain level of memory.

00:21:30.000 --> 00:21:41.000

Which means and it can remember me this one like GPT 3.5 about 16 pages of content and it can remember our conversation and interact with my conversation within that amount of memory.

00:21:41.000 --> 00:21:45.000

So what I'm gonna do is I'm gonna plop this in and we're gonna talk about it and I hope everyone can see it.

00:21:45.000 --> 00:21:52.000

I'll make it a little bit bigger. So you got create a lesson plan for a mixed stability class on the topic of fractions.

00:21:52.000 --> 00:22:00.000

Include activities for students with very levels of understanding and those who require tactile learning experience. And we're going to do this right now.

00:22:00.000 --> 00:22:04.000

So I'm back that's called a prompt. So if you if you're brand new to this, I love it.

00:22:04.000 --> 00:22:05.000

We're so glad to have you here. But this is called a prompt. When I asked Chat GT a question, I'm prompting it.

00:22:05.000 --> 00:22:14.000

Okay, so sorry Chris, I won't let you copy from the chat. Maybe then follow along or you can, it's not too long.

00:22:14.000 --> 00:22:22.000

So, and or feel free to experiment if you want to do a slightly different one. But what we've done is we've actually now with Chat GT, we've got it's identified where there would be a fraction level.

00:22:22.000 --> 00:22:33.000

You know, looking at the context, it looks at fourth to sixth grade. So I don't even have to give it to that.

00:22:33.000 --> 00:22:46.000

Give that context. It's got some adjectives, some materials including manipulatives, a duration, and so we've got lesson plan, we've got the introduction, hands on exploration.

00:22:46.000 --> 00:22:52.000

Fractions on a number line guiding practice, closing and assessment. And then there's some extensions at that mixed ability.

00:22:52.000 --> 00:22:58.000

So it's identified advanced students. It's got his identified additional hands-on support for students who need special support.

00:22:58.000 --> 00:23:06.000

And this is where I think it's really, really interesting. So now I can say now, now add an extension.

00:23:06.000 --> 00:23:09.000

So for students.

00:23:09.000 --> 00:23:17.000

That are. Dislike. We're going with that.

00:23:17.000 --> 00:23:25.000

And for students that. Get discouraged.

00:23:25.000 --> 00:23:31.000

Trying something new. So I'm giving a couple of different things, so mixed ability.

00:23:31.000 --> 00:23:33.000

So not just in terms of students with the disability, but also just like some of us get discouraged.

00:23:33.000 --> 00:23:46.000

And so here we go. So it's gonna go now and it's gonna include around anxiety and so it's going to have the same type of lesson but now what we have is like some clear visuals of minimal text.

00:23:46.000 --> 00:24:06.000

You've got a couple of different places in which it's been added. It talks about actually, you know, having smart fonts and formats, specific support and also decided with students like actually maybe take a little bit of a break for mindfulness and this is just an example we're going to do a lot more in the rest of our time together to kind of dig into this.

00:24:06.000 --> 00:24:12.000

But what you'll notice is it's a pretty good start. I know that those of you in this room are experts.

00:24:12.000 --> 00:24:23.000

Are you in in whatever you're an expert in. And if you are an expert special education teacher, or you're an expert in all kinds of different ways, I think that this is going to be really, really interesting.

00:24:23.000 --> 00:24:33.000

In the sense of we have this opportunity, to, start to play and use our expertise, but also use Chashvity and other tools to extend and to do more with what we have.

00:24:33.000 --> 00:24:45.000

And so I think this is something that is really interesting for us to start with. So. While this is exist like we're starting with this very basic prompt, we're going to come back and do some more prompt engineering and show what's possible.

00:24:45.000 --> 00:24:53.000

And I love that Sarah, I love that you're getting excited about it and that Christian, you know, this is great.

00:24:53.000 --> 00:24:54.000

It can translate into languages, but it definitely is an English language. It's been trained on the English language.

00:24:54.000 --> 00:25:04.000

You are, it depends on how you use it. You can use it. I would say that if you use it you can use it.

00:25:04.000 --> 00:25:08.000

I would say that if you can interact with it with English and then have it, I would say that if you can interact with it with English and then have it translate, that if you can interact with it with English and then have it translate, that tends to work better.

00:25:08.000 --> 00:25:22.000

But it can recognize over a hundred languages. And so for example, I did a session with a tribal college in this mark, North Dakota, and we were able to interact with Chastity T in Lakota, which was amazing and something so cool.

00:25:22.000 --> 00:25:27.000

It wasn't amazing at it in terms of it got, you know, some the grammar wrong, but it was really cool to be able to access it in other languages.

00:25:27.000 --> 00:25:43.000

But it works best in English at this stage. So for An, you can't change the temperature of chat to VT, but you can for other models if you're building your own bot with something like Poe or others you can change the temperature.

00:25:43.000 --> 00:25:49.000

And then, and I'll leave that one. We have some bad at math. We'll talk about that in a moment, but we're gonna go over to like a tech talk.

00:25:49.000 --> 00:25:56.000

And so we're gonna spend a ton of time on this. We're talk a little bit about the technology in the sense that, you know, why, why does it work the way we are?

00:25:56.000 --> 00:26:13.000

And I had a really great question already come in from Melinda about the idea of bias is that these systems are not necessarily biased because they have their thinking and they are biased, you know, engines, but they are biased because of the way that we train them and we'll talk about why that happens right now.

00:26:13.000 --> 00:26:22.000

So thank you for the point of clarification. But our official intelligence has been around. We've essentially Socrates talked about thinking machines a long time ago.

00:26:22.000 --> 00:26:34.000

Like 2,000 years ago. And so we have this idea of, artificial intelligence as a theory about creating machines that can think and act like humans.

00:26:34.000 --> 00:26:42.000

So you can think about robotics, you can think about this idea of actually creating, you know, chat, GPT, this conversational AI.

00:26:42.000 --> 00:26:52.000

And then we have machine learning. And so that's the next subset down. And that's where we have something like predictive text where it takes a whole bunch of data and then it's going to give us an output.

00:26:52.000 --> 00:26:59.000

Based on how we how we use it. So for example, if I ask Siri what the temperature, it'll be able to tell me the temperature outside.

00:26:59.000 --> 00:27:07.000

So it does one thing really, really well. What that was though is that it actually didn't, wasn't that helpful in terms of it's very, very expensive.

00:27:07.000 --> 00:27:23.000

It takes a lot of time and effort and data to build machine learning models. And so for example, a chat bot, a sorry chat bot, but a system like Quill that looks at Felix at writing, they had 200 machine learning models to help with writing.

00:27:23.000 --> 00:27:30.000

So a model for checking punctuation, for spell check, for, you know, things like active and passive voice.

00:27:30.000 --> 00:27:42.000

And so these are all something that's going to be something that's important for us to understand is that that was pretty expensive and when we think about technology we're trying to create better technology, faster technology.

00:27:42.000 --> 00:27:51.000

And this is something in which we went from deep learning because what do we want to do if we want to create, you know, technology that things like humans is that we look at the brain.

00:27:51.000 --> 00:27:57.000

So we looked at the brain and then looked at that and created something called a neural network. And that's led to what we have today.

00:27:57.000 --> 00:28:03.000

So we when we talk about today, we're talking about a form of artificial intelligence called generative AI.

00:28:03.000 --> 00:28:09.000

And what we just saw was a lesson plan that was created based on all of its training data and how I asked it a question.

00:28:09.000 --> 00:28:20.000

And so General, I can create new text, audio, images, video, or code. And a lot of other things that are happening for based on what it's been trained on and the way that we use it.

00:28:20.000 --> 00:28:28.000

So I'm gonna ask everybody if I ask you to complete the sentence the sky is what is the most common next word of the sky is.

00:28:28.000 --> 00:28:36.000

Blue. Okay, but what if I changed it now or in the book? Chicken little. What would this guy be doing in chicken little?

00:28:36.000 --> 00:28:49.000

Falling, absolutely. So that's your, that's another piece. We talked about temperature, which I'll talk about in a moment again, but also now we're talking about the way that we work and so in the case of Chat TVT, it actually predicts the next word.

00:28:49.000 --> 00:28:54.000

So if it's in this case the sky is the most common next word is blue, but it can also be clear or usually and that's based on the training data set which we'll look at momentarily.

00:28:54.000 --> 00:29:07.000

So if we change to a book that's specific is that we're looking at something that is going to be specific to like the training data set.

00:29:07.000 --> 00:29:13.000

And so that's how this works. And so if you're creating an image, instead of it being the next word, it predicts it's actually the next pixel.

00:29:13.000 --> 00:29:21.000

So it'll build it by pixels. And so in this case. We have now an example and this is part of our free student curriculum.

00:29:21.000 --> 00:29:25.000

So if you love this and you're a big tech nerd, this is for you. We won't spend more than a couple minutes here, but this is how these work.

00:29:25.000 --> 00:29:34.000

So GPT is called TEAS for a transformer. And what we have here is you have a bunch of texts from the internet.

00:29:34.000 --> 00:29:42.000

So in the case of 3.5, which you're going to use this text, it's text only for the more expensive version is going to be text and images.

00:29:42.000 --> 00:29:50.000

What happens is that comes into the transformer and it turns it into probabilities or relationships between words like the sky is blue.

00:29:50.000 --> 00:29:57.000

That creates a model that then checked by humans and then can say, because remember these tools aren't thinking, so it has to be told what's appropriate or inappropriate.

00:29:57.000 --> 00:30:04.000

And what happens is, and that's also where temperature is set. So thank you so much to Dana for being my helper here.

00:30:04.000 --> 00:30:11.000

I love you Dana. Amazing to have you involved, but that is the idea of like the level of variation that's set at the model.

00:30:11.000 --> 00:30:13.000

So how much variable and they play around with that to be able to get to the right level of good interaction.

00:30:13.000 --> 00:30:20.000

And then it's fed back into the model and then it is released and that's when we get to use it.

00:30:20.000 --> 00:30:26.000

One thing we have to understand though is that when a model is released, It means that the training data set stops.

00:30:26.000 --> 00:30:33.000

So for example, if I go to chat to be T right now, and you might already know this, if I ask it.

00:30:33.000 --> 00:30:41.000

What? How up to date? Is your training data. What I'm gonna see.

00:30:41.000 --> 00:30:47.000

Is it September 2021? It's actually I think September 2022 now and if I go to GPT 4 it's different.

00:30:47.000 --> 00:30:55.000

So I will say, are you sure? Pretty sure it's 2022. But what you can see is that if something has changed, no, it's a bad.

00:30:55.000 --> 00:31:01.000

What we have now is that if I ask it questions like Do you know if Travis Kelsey?

00:31:01.000 --> 00:31:05.000

And Taylor, well, like I say, who is? Do you know who?

00:31:05.000 --> 00:31:07.000

I think, yeah, it should be January, 2022. So this actually a great example of a hallucination.

00:31:07.000 --> 00:31:19.000

It's hallucinating or making up stuff. About itself, which is crazy. So it's actually wrong, but it's very confident and it's wrong answer.

00:31:19.000 --> 00:31:27.000

But if I ask it a question like who's dating Travis Kelsey, it would not be able to do that because the training data has stopped.

00:31:27.000 --> 00:31:34.000

So that's something it's really, really important. If you're in Ontario, I know we have some Ontario people here, you have new standards that were came out after the training data set.

00:31:34.000 --> 00:31:40.000

So if I asked for a lesson planned, a lesson plan.

00:31:40.000 --> 00:31:47.000

For. . The class on Fractions.

00:31:47.000 --> 00:31:53.000

With Ontario. Standards. What's going to do is it's going to give me Ontario.

00:31:53.000 --> 00:32:00.000

I know I spoke Kelsey wrong, Alexa. I, I know. I, well, it's like it'll get it like I'll show you as well.

00:32:00.000 --> 00:32:04.000

It's gonna get it wrong again. But what we can see though is it has standards that are represented.

00:32:04.000 --> 00:32:16.000

Let's see, actually it didn't increase like. It includes Ontario. Alright, and standards.

00:32:16.000 --> 00:32:20.000

And there we go. These are n, these are going to be the old standards, not the new ones.

00:32:20.000 --> 00:32:26.000

And I'm going to ask it for Chelby Kelsey. I'll spell it correctly now.

00:32:26.000 --> 00:32:29.000

Who is Travis Kelsey dating?

00:32:29.000 --> 00:32:38.000

It will also not know. So, but this is an example of why it's so important to understand how this technology works or doesn't work because it does have knowledge cut-offs.

00:32:38.000 --> 00:32:43.000

And for grace, if you're using chat ct for it's knowledge cut off is April, 2023.

00:32:43.000 --> 00:32:53.000

So but this is all what we're thinking about in terms of you have these models that work as specific way and it's important for us to understand the capabilities and eliminate limitations.

00:32:53.000 --> 00:33:06.000

And this is something that's really, really important. These tools, like we said, this is the worst they've ever been and will be and so they do have things where they have bias in the system, they also can have hallucinations or inaccuracies.

00:33:06.000 --> 00:33:12.000

So the way that we're gonna do this, the best way we think to use generative AI is to think of it as a brainstorming partner.

00:33:12.000 --> 00:33:18.000

So if you are a special educator, You are so busy like I think the special educators have the hardest job in education and I've seen that very strongly.

00:33:18.000 --> 00:33:40.000

I sought special education when I taught. I was not a special education teacher, but I was in a high complexity school and I cannot, I will never forget how difficult it was to try to meet the needs of my students and what's happened since I left the classroom is we have really ramped up the level of work we have to do.

00:33:40.000 --> 00:33:53.000

That is, and love you, I love that if you're in the audience, I just want to say thank you for being here and all the work that you do and taking the time and the fact that Ela is supporting this work I think is just so amazing and important for us to be able to understand.

00:33:53.000 --> 00:33:58.000

And so just thank you for being here. You are superheroes. You're definitely definitely a superhero for me.

00:33:58.000 --> 00:34:09.000

But what it is is this can be a great way for you to be a, have a brainstorming partner, to be able to get that level of like, okay, I need a whole bunch of great ideas.

00:34:09.000 --> 00:34:22.000

I need a whole bunch of activities to get my students involved. I want to be able to show them how to to really get deep into some things that are previously very difficult, like going back to the idea of dyslexic students.

00:34:22.000 --> 00:34:23.000

You no longer have to type, you know, you don't have to have it perfectly spelled.

00:34:23.000 --> 00:34:38.000

It will recognize the way that I type and I can be a little bit more confident that I won't look like I don't know or I'm dumb because I'm unable to do some like I'm unable to.

00:34:38.000 --> 00:34:47.000

You know, do things that we expect like our non dyslexic folks will expect us to be able to do and there will be no judgment there and I think this is so important.

00:34:47.000 --> 00:34:54.000

So this is a great way to have a pro brainstorming partner and it's also something in which it really can save you time.

00:34:54.000 --> 00:35:03.000

And with a good prompting, your creativity as educators and some resilience, meaning like you keep going, you can create some amazing things.

00:35:03.000 --> 00:35:17.000

And we're gonna do in this series. So this is the first of a series. So what we're going to do is the next one we're actually going to look much more deeply into planning before we move into IP goals and then we move into the IP process and then finally we're going to focus on working with students with disabilities.

00:35:17.000 --> 00:35:26.000

So this is a really great way to think about this. And so like Jody is like, this is a great way to build some differentiation and will be focused on that as well.

00:35:26.000 --> 00:35:37.000

And so we're going to go into, so Katie and others, this will be recorded so you can share with your colleagues as well or watch at a later date.

00:35:37.000 --> 00:35:46.000

So what we have is some ways in which generative AI will work really well right now. So we've got written communication, lesson planning, differentiating content, which is just asked, brainstorming and also translations like we said before like Estonian.

00:35:46.000 --> 00:36:04.000

It's not going to be perfect but it does a pretty good job or if you're an expert in that language or you have someone that you can use to take a check over, even maybe your student, you can do this and you can actually have, some really amazing differentiating content.

00:36:04.000 --> 00:36:12.000

And so the last piece we're going to talk about that's really important as we talked about it before, it is a silly name, y'all.

00:36:12.000 --> 00:36:21.000

Hallucinations is not a name I would have suggested. This is what technologists call, Lin, a model messes things up and is no longer accurate.

00:36:21.000 --> 00:36:31.000

So this is something that happens a lot. It can happen up to 20% of time and someone actually asked us about why it's so bad at math is because it's not actually doing math and traditional sense.

00:36:31.000 --> 00:36:40.000

If I ask for, you know, a word count or to do basic calculations, if a tool is not using a calculator, it's actually not going to be able to do that.

00:36:40.000 --> 00:36:41.000

But it will say that it can and so it can make up things like fake leaks to books or articles.

00:36:41.000 --> 00:37:02.000

YouTube videos. It can give you a date that's wrong that we saw September, 2021 is not the right training data set for chat GPT and so we have to be really really careful that we're always double checking our work that the human and loop that the expert that you are is going to be something in which it's really important.

00:37:02.000 --> 00:37:09.000

So that the hallucinations, we're looking at this and right at this stage there are no models that have been able to fully eliminate hallucination.

00:37:09.000 --> 00:37:13.000

Because it's kind of how it works, it's always making stuff up because it's not thinking.

00:37:13.000 --> 00:37:17.000

And so in the case of this, it is always thinking, I'm sorry, it's always making stuff up and then sometimes it's wrong and sometimes it's not.

00:37:17.000 --> 00:37:25.000

Okay, so we're gonna go back to we're gonna go back to, what we need to do.

00:37:25.000 --> 00:37:40.000

We're gonna we're gonna talk about privacy before we leave, but I'm gonna go back here and we're actually going to look at our prompt library and we're gonna do a lot of this together, but I wanna make sure that we're having an opportunity to,

00:37:40.000 --> 00:37:45.000

I spelled that wrong, but here's our prompt library. So Anne, it's a great question.

00:37:45.000 --> 00:37:51.000

Hallucinations are because that's it's always making stuff off based on probabilities. And so based on the training data set, it's going to have different probabilities.

00:37:51.000 --> 00:37:59.000

But Realistically, because not thinking, it doesn't actually know if something is wrong or right.

00:37:59.000 --> 00:38:08.000

And so it often will and it's also how it's been told to act. So for example, Chix VT has a system prompt that says, don't say I don't know.

00:38:08.000 --> 00:38:16.000

So it will continue to answer even if it is it's outside of its realm of ability. So I know that we're going to have some time for Q&A.

00:38:16.000 --> 00:38:22.000

So if you want to put some Q&A into the actual Q&A. So if you want to put some Q&A into the actual Q&A piece, yes, I will put that into the chat right now, Sarah.

00:38:22.000 --> 00:38:30.000

Thank you for identifying that for me. This is our prompt library, which is one of our favorite. It's actually what we launched the AI for Education with was our prompt library.

00:38:30.000 --> 00:38:33.000

And if you've used it, I know I see some familiar faces here. If you used it before, we loved having you.

00:38:33.000 --> 00:38:41.000

But this is an opportunity to start to baby do some preparation work for our next next time together next week.

00:38:41.000 --> 00:38:52.000

Our share, but we have all kinds of different types of prompts. And remember, a prompt is where we actually ask Trash GBD how to do this and it makes us a difference in how we approach it.

00:38:52.000 --> 00:38:59.000

So like the way we ask a question because it's a technology actually matters. So we have lesson planning, we have professional development.

00:38:59.000 --> 00:39:10.000

I've used it to help build some professional development that we do with you all, social emotional learning, which I know is so uniquely important to student to special educators and that work with students with disabilities.

00:39:10.000 --> 00:39:21.000

And we also have some special needs pieces. So we'll do a lot of this next week. And so you can see that we'll look at things like accommodations, we'll look at some planning pieces, and we'll do that together.

00:39:21.000 --> 00:39:29.000

But I always like to do in this session. I always like to do my favorite prompt in our first ever prompt, which was a rubric.

00:39:29.000 --> 00:39:34.000

And so if you know me, you know, that I really dislike, creating rubrics.

00:39:34.000 --> 00:39:38.000

And so this is the first prompt I ever used on Chat GBT ever. And so, and thank you everyone for being so great and supporting each other in the chat.

00:39:38.000 --> 00:39:51.000

It's so good to see you guys helping each other out. And so we have here, Oh, that's not right.

00:39:51.000 --> 00:40:02.000

That's just the URL. Okay, so we have here is we have the way that our prompt library set up is we have a prompt that you can change based on what you need.

00:40:02.000 --> 00:40:08.000

You have a prompt that you can just cut and paste. And then we have a couple of ways in which you can, this one has 2 versions and then some ways in which you can keep working with it.

00:40:08.000 --> 00:40:15.000

And so for this case, what I'm gonna do is I'm actually gonna change it up for our for us today though.

00:40:15.000 --> 00:40:26.000

So we have we're setting the scene we'll talk about that more next week but you want to give it a you know tell it what it is you know it's the internet is huge and that's a training data set.

00:40:26.000 --> 00:40:34.000

We always like to give it a specific role to take and then very specific directions. And so our directions really, the more specific we can be the better.

00:40:34.000 --> 00:40:49.000

So what I have here is a fifth grade class studying engineering. And through the egg drop challenge and we want to create that and I want to we're gonna create a just a standard rubric now and so if you've never seen this before wait for your socks to be officially blown off.

00:40:49.000 --> 00:40:59.000

It might take you, you know, 45 min to write a rubric. It's not taking you 10 s and maybe it's not perfect and we'll show you how to work with that more where I can start to say.

00:40:59.000 --> 00:41:07.000

Eck protection, presentation, collaboration, collaboration, teamwork. So what I can do now is say change to a 4 point rubric.

00:41:07.000 --> 00:41:18.000

Remove the teamwork aspect. Because it's individual. And create. Yeah, we'll start there.

00:41:18.000 --> 00:41:24.000

So we're going to remove that. So we're going to redo this. And here we go.

00:41:24.000 --> 00:41:29.000

So it's changed it to a 4 point rubric. And it has removed. That's it.

00:41:29.000 --> 00:41:37.000

It has changed their movie. Just remove the teamwork component, which is super cool, right? So kept the presentation and communication, but this is where it gets really great for differentiation.

00:41:37.000 --> 00:41:43.000

Now what I can do is create A checklist version.

00:41:43.000 --> 00:41:49.000

Oh, this rubric. So for my students that don't do well with charts like this.

00:41:49.000 --> 00:41:55.000

I can do that. So now I can create a differentiated version. The formatting will be a little bit wonky, so like.

00:41:55.000 --> 00:42:01.000

Take it with a grain of salt. It's not going to be perfect yet, but I can also say now and create.

00:42:01.000 --> 00:42:13.000

3 guys to completing. This activity. One for students. That, need visual support.

00:42:13.000 --> 00:42:20.000

One for students that need. Ajestional scaffolds. Or.

00:42:20.000 --> 00:42:27.000

Decision making and one for.

00:42:27.000 --> 00:42:41.000

Our students that are. Side impaired Okay, so this is where I think this is where it gets really interesting is that now I can actually create 3 different guides.

00:42:41.000 --> 00:42:46.000

So I have a guy for, and, and, and, and visuals like a bit have visual impairment.

00:42:46.000 --> 00:42:52.000

If that. Ones around. The guide for students seeing additional support.

00:42:52.000 --> 00:43:02.000

But this is where it gets really, really fascinating for you all is that differentiating for your students, especially if you are a special educator.

00:43:02.000 --> 00:43:12.000

It sometimes feels like a mountain, a mountain and you're on Everest and you're trying so hard to support your students and to meet their different goals.

00:43:12.000 --> 00:43:22.000

What happens is it just you just don't have enough time or you just need that extra bit of support or you could maybe you're just a little bit tired right maybe it's just hard to get your head in that creative space that often has to happen.

00:43:22.000 --> 00:43:29.000

And so this is an example of how we can do this. And the more we'll talk about this much more in the later series where we get deeper and deeper into how to use these tools.

00:43:29.000 --> 00:43:32.000

And how to think about them in targeted ways. To support your practice. But if you haven't been using these tools, this is a great opportunity.

00:43:32.000 --> 00:43:45.000

My challenge to you all between this week and next week or if you're watching this on recording is to try it out use our prompt library or just try it out.

00:43:45.000 --> 00:43:54.000

Think of a piece that you really want to do more. Or are you, tend to have to like put aside because you're just so busy, and be able to do this.

00:43:54.000 --> 00:43:59.000

And I love that we have so many great examples of what you can think about whether using UDL, working backwards.

00:43:59.000 --> 00:44:04.000

There's a lot of great opportunities here. So I'm just gonna, we're gonna have some time for questions.

00:44:04.000 --> 00:44:09.000

So I have a couple of questions. If you have them. Please like, put this into the QA because it's hard for me to be able to follow the chat.

00:44:09.000 --> 00:44:16.000

So if you have specific questions that you'd like me to answer. Please put them in the Q&A section and then we'll do some answering.

00:44:16.000 --> 00:44:34.000

And we'll be able to talk about that together. But I have a couple of things I want to do before we I let you guys, I let you guys go, which is the first thing is we're going to look at why it's so important for us to be thoughtful about how we use these technologies.

00:44:34.000 --> 00:44:40.000

And one of them is because you are working with students that have protected bases like they're protected. We have student privacy data and personal information.

00:44:40.000 --> 00:44:53.000

We have a lot of information that should not be shared. And so I think this is just really, really important for us to understand that it's always important for to be very, very cautious about what you share.

00:44:53.000 --> 00:45:01.000

And we'll be reiterating this in every single session because while you might want to be able to build specific, you know.

00:45:01.000 --> 00:45:09.000

IP goals or plans or accommodations for a specific student. These, these technologies are leaky.

00:45:09.000 --> 00:45:14.000

They can actually, you know, if you're using Chatch VT or other technologies, that data can be used to train this system.

00:45:14.000 --> 00:45:18.000

It can be it could be sent to multiple places through APIs. You just have to be very, very careful and please never include any personally identifiable information.

00:45:18.000 --> 00:45:41.000

About a student, a parent, or even a colleague or yourself. With an AI tool. So we will show you some examples of how to do this, of how to kind of get like redact and or to remove, but just very, very important that you're thinking about these because your students are protected.

00:45:41.000 --> 00:45:49.000

Their data is protected, their privacy is protected, even though this technology could be really, really cool. So what we're going to do is we're actually going to get a question and the answer is we'll spend the last 5 min.

00:45:49.000 --> 00:45:58.000

To be able to answer a couple, to tell you about next steps. And we're going to go to the questions.

00:45:58.000 --> 00:46:04.000

Please put some questions. Please put some questions. Please put some questions and we're gonna go to the questions. Please put some questions in. I know I have we have almost 400 people here.

00:46:04.000 --> 00:46:07.000

I know I have, we have almost 400 people here. I'm sure we have a whole bunch of questions.

00:46:07.000 --> 00:46:09.000

And so, we have almost 400 people here. I'm sure we have a whole bunch of questions.

00:46:09.000 --> 00:46:18.000

And so, Melinda, I think we talked about this a bit about this idea of bias. So data, like we can build better technology that is less biased and actually right now there are a ton of organizations that are working directly against that.

00:46:18.000 --> 00:46:32.000

So if for example, if you use something like Bard, or other tools, they actually put in their image generators be diverse and they're trying to work against that bias but we need better trading data sets. We need better systems.

00:46:32.000 --> 00:46:41.000

So it's just something to be aware of now that these systems will replicate bias and we just need to be aware of that being the case.

00:46:41.000 --> 00:46:52.000

Okay, thank you for that question. Melinda, we've got from Carol. This idea for lesson plans.

00:46:52.000 --> 00:47:18.000

Okay, so we're gonna talk a much more about that. The more specific can be by even including standards or in this case school policies cutting and paste them in or in the case of using something like a clawed or paid version of chat twot you could actually upload the actual policy and or the standards and interact with it that way so you get really targeted lesson plans or accommodations.

00:47:18.000 --> 00:47:24.000

The more open we are to asking a question, the more open, like less, less information we give, the more generic the answer will be.

00:47:24.000 --> 00:47:33.000

Because remember it's creating probabilities and the more like it's going to give you a generic answer unless you're very specific on what you ask.

00:47:33.000 --> 00:47:40.000

So thank you for that. That question, Carol. Next question. Oh man.

00:47:40.000 --> 00:47:55.000

So Naz has some some deep hitting questions about disinformation. It is really interesting. Everybody disinformation is something that and misinformation and inaccurate data, something that's happening a lot.

00:47:55.000 --> 00:48:12.000

Then a couple of different ways. One is we have lawyers and doctors who have been fined or disbarred for for essentially making up like using General AI like Bard or Chat to make up they didn't they didn't mean to but they asked it about building some case brief and it was wrong.

00:48:12.000 --> 00:48:22.000

It was made up. Same thing with some academics. We've seen misinformation. There was a 1,200% increase of internet scams after the release of chat GBT.

00:48:22.000 --> 00:48:29.000

We're seeing Joe Biden robo-calling people and it was not really Joe Biden. It was using an AI voice generator.

00:48:29.000 --> 00:48:39.000

So there's a huge focus and actually like building a couple things. One is we need better systems that are built by the technologists and better safeguards and regulations.

00:48:39.000 --> 00:48:48.000

And so if you're if you're in this room and you're a technologist or you have a site line too, someone in that can be some regulatory body, love the work you're doing.

00:48:48.000 --> 00:48:55.000

Hopefully this is something you're working on. But also we need to build critical literacy around how to like how to be aware of how to not trust and be critical users of these.

00:48:55.000 --> 00:48:58.000

Because they can create things that are inaccurate. In fact, they do it all the time.

00:48:58.000 --> 00:49:07.000

And also people are using it in ways that can be pretty damaging in ways that can be pretty damaging.

00:49:07.000 --> 00:49:11.000

So thank you, Nas. Always happy to talk about ethics and we'll be doing that together as well.

00:49:11.000 --> 00:49:20.000

So, from anonymous attendee, love you anonymous. We have other tools like Chats, VT over bar and other platforms.

00:49:20.000 --> 00:49:29.000

We don't have a ton of time today, but I think what we'll do is we'll pull together a little bit of a guide that we can share with you all potentially in either the second or third session about how to compare these.

00:49:29.000 --> 00:49:41.000

Chats to BT, especially for is considered to be the best model on the market. It does cost anywhere between, you know, 16 if you're buying it for the year to 25 if you're using it as a team.

00:49:41.000 --> 00:49:42.000

A month so it is not inexpensive and I know that it's very hard to ask people to add an expense to your life.

00:49:42.000 --> 00:49:56.000

So, but it is known to be the best tool. You've got things like Claude, which are can take almost like Harry Potter and answer questions about you can upload a document.

00:49:56.000 --> 00:50:02.000

I don't suggest uploading Harry Potter because it's copyright protected, but if you had your policies, it's copyright protected. But if you had your policies, you could do that.

00:50:02.000 --> 00:50:08.000

It's also a great writer. You don't have to remember all of this. We's also a great writer. You don't have to remember all this.

00:50:08.000 --> 00:50:15.000

We'll make sure that this is something that we bring up next time. Bard, which is now going to be called Gemini, is it's actually connected to the internet and will be able to search, which is really great.

00:50:15.000 --> 00:50:20.000

So it actually can give you a little more comfort that it's that it's going to, be accurate.

00:50:20.000 --> 00:50:30.000

And then the final thing is something called perplexity, which is a great, it's a generative search engine where you can, it's kind of like Google plus chat TBT and it's really cool together.

00:50:30.000 --> 00:50:40.000

And I think if, Amanda or someone on my team, if you don' Claude, that would be great.

00:50:40.000 --> 00:50:49.000

And so those would be awesome to do. And also being as what we say here, being also is up to date as well as connected to the internet.

00:50:49.000 --> 00:50:58.000

And then also, yeah, so great. We have a lot of sharing this chat too around image generators like Dolly 3, the journey, and Microsoft image generator.

00:50:58.000 --> 00:51:05.000

So we're gonna run out of time. I think I have time for a couple more. So I'm gonna take a look and see.

00:51:05.000 --> 00:51:12.000

Dana about using the tool for students have difficulty with right running expression. I have troubles generating.

00:51:12.000 --> 00:51:14.000

You can create, I mean, send the starters and a great way to build like worksheets, essentially differentiated worksheets that you can do.

00:51:14.000 --> 00:51:24.000

This is something that it's just really like What do you want it to do and try it out?

00:51:24.000 --> 00:51:39.000

My, Jane, I want you to go try and see if you can actually create a couple of differentiated pieces where maybe you create a couple of like sit on starters and then give some space for the student actually is scaffolded to creating their own.

00:51:39.000 --> 00:51:46.000

For, Lisa, great question about oral, data. So something called automatic speech recognition.

00:51:46.000 --> 00:51:53.000

I can tell if I have Chat GPT on my phone, I can talk to it and it can talk back to me, which is super cool.

00:51:53.000 --> 00:52:04.000

So we talk about accommodations. A lot of the different tools actually now have ASR. I know a lot of people that use this that are uncomfortable with text are, low vision.

00:52:04.000 --> 00:52:09.000

So that's a great way to do that. The iPhone app is amazing for that.

00:52:09.000 --> 00:52:20.000

So will you see more and more of multimodality being able to be part of these systems. So I'm gonna go.

00:52:20.000 --> 00:52:29.000

We'll definitely be sending a recording. So Julie will do that tomorrow. let's look at see.

00:52:29.000 --> 00:52:37.000

I think I have time for 2 more questions.

00:52:37.000 --> 00:52:46.000

Okay, for Amy, good question. Amy asked a question about personally identifiable information and what policies should we be doing.

00:52:46.000 --> 00:53:07.000

And so this is something that's really important. I think that we want to think of this as an extension of existing laws and pieces and that what happens though is it feels so natural to put in like Johnny has X right like and maybe we give it some identifying information because it feels quite different the way we interact and we want to get more information than we would normally.

00:53:07.000 --> 00:53:20.000

So what we suggest is you really think of this as the extension of any technology. And the same way that you would not, you know, open up your your grade book or your your notes, anecdotal notes to any other technology, would not want to do that with us.

00:53:20.000 --> 00:53:33.000

So we suggest really extending your existing AUP or academic integrity. Sorry, your acceptable use policies to include this so that people are just aware that this is that are form technology even though you interact with it differently.

00:53:33.000 --> 00:53:38.000

So I think that will, we will definitely tackle more privacy. We'll, do that again next week.

00:53:38.000 --> 00:53:55.000

For those who are wondering, privacy, safety, data is something that Ela and I share. We share a lot of, focus on and the fact she had actually gonna ask TREA to come back on and we could talk about next steps but this is something that's really great.

00:53:55.000 --> 00:54:17.000

There's so many great questions. I just want to say you know for us to be able to have such a rich conversation at this early stage our entire goal was we will we are building this series which is great as a partnership so we're going to be able to do this work together to actually like identify some maybe some key areas that we can actually work on and incorporate.

00:54:17.000 --> 00:54:24.000

And so this is something where, you know, if you have good ideas, we have a feedback form that we're gonna put in the channel right now, Amanda, if you don't mind putting that channel.

00:54:24.000 --> 00:54:31.000

If you want cool things, if you want to know more, please. Complete that form. And I just want to talk about like how you can continue learning with us.

00:54:31.000 --> 00:54:37.000

So really quickly we have our prompt library which we shared. We've got a free 2 h course and we also have our newsletter.

00:54:37.000 --> 00:54:44.000

You get a hangout with us, which is super cool. And then I'm gonna hand over Treya about how you can keep in touch with Ela who are why we are here today.

00:54:44.000 --> 00:54:45.000

So just everybody got Jackie Aurora, Aaron, everyone that has helped on the ELA team.

00:54:45.000 --> 00:54:52.000

Just thank you so much and this is how can we keep in ball with you all.

00:54:52.000 --> 00:54:59.000

Thanks, Amanda, and thanks so much for such a great presentation. That was really fun to kick off this series and stay tuned with us.

00:54:59.000 --> 00:55:07.000

But if you want to learn more about Ela or the educating all learners alliance, we've got all kinds of great information on the screen.

00:55:07.000 --> 00:55:21.000

You can visit our website at educating on learners. Org. You can send us an email if you've got a great resource we've got a fantastic resource library that includes links out to AI for education but we actually go beyond just AI and focus more specifically on supporting.

00:55:21.000 --> 00:55:26.000

Students with disabilities and inside the classroom. So lots of more generalized resources as well. And you can send us emails.

00:55:26.000 --> 00:55:37.000

They visit our website. We're excited to welcome you into the Alliance.

00:55:37.000 --> 00:55:52.000

That's so great. And thank you so much. I mean, we are so lucky to be able to do this work with everyone and like so what's coming up okay I know I will tell you all in my it is not as easy to run a webinar and answer 400 people's questions. So I know I missed some of your questions.

00:55:52.000 --> 00:56:03.000

So I apologize that we could get to everything. I know there's some still these really big questions that you have all and we will make sure that we are getting you that that time and if you don't mind answering that feedback form.

00:56:03.000 --> 00:56:09.000

Partner free get you to do it after you leave here today, but I see it 317 people off, so please give us that feedback.

00:56:09.000 --> 00:56:15.000

We will come back. We're going to focus the next session much more on planning. So we did a couple of prompts today, but more as an a boost.

00:56:15.000 --> 00:56:26.000

You're getting a little bit of a taste. And what we're going to be doing is we're going going into the planning process specifically for working with students with disabilities as special educators.

00:56:26.000 --> 00:56:41.000

And so while this was a more general topic because this is our introduction. Today we are going to be going much more deep into how to support you as special education, technologists, teachers, leaders, all the things I know you are and the piece.

00:56:41.000 --> 00:56:45.000

So let's do like make sure that if you want to share this. We'll do that.

00:56:45.000 --> 00:56:52.000

We'll then move on to the IP process and then I'm really excited the last session will be focused on working with students with disabilities.

00:56:52.000 --> 00:56:57.000

We'll be using some real examples of how to use things like Canva for students that are non-verbal.

00:56:57.000 --> 00:57:03.000

And so we'll be doing a lot of work on that, which is really exciting. We'll also be highlighting some great special educators as well.

00:57:03.000 --> 00:57:09.000

So we'll be introducing that. But I want to say thank you to everyone that supported us today.

00:57:09.000 --> 00:57:16.000

My team, Amanda, has been done an amazing job of getting these resources ready for you all. Dan, who's so helpful and all these everything.

00:57:16.000 --> 00:57:20.000

And I just want to say, wherever you are, I know some of you, it's late.

00:57:20.000 --> 00:57:23.000

Go to bed. If it's the morning, good, you know, we're excited to have you here.

00:57:23.000 --> 00:57:28.000

If you're back going back to teach, you have our support, we're just so glad that you took a little bit of time with us today and we look forward to having you join the next session.

00:57:28.000 --> 00:57:36.000

Thanks everybody.