Thanks Alan and hello good afternoon and welcome back from lunch. I have a very short 25 minutes to talk about a lot of things.

So it's going to be a very fast rant. I hope we have some time for questions but if not we've got the rest of the day.

The title is slightly misleading and I had a slightly different approach to this when I proposed the title but things have developed recently and so I'm squeezing in a few different topics.

So what I'm going to do is, well first of all I'll introduce myself so these are my roles at the moment. Most people know me as the founder of Moodle but I'm involved in a lot of other things.

I'm going to tell you right up front what the takeaways of this session are. So first of all here's the chain of logic that I want to get across to you.

Number one, AI will be the primary creator of educational content in the future.

Number two, there is a crisis that knowledge quality is degrading. Number three, I have a proposal for a solution that we can all be involved in.

And number four, that can impact the whole world so it's fairly important.

I'm going to justify all this, don't worry. The second thing I want to get across to you is a bit of a metaphor shift.

We often think about the internet and OER in general as a library. We're going to produce publications and put them out there into the library of the internet.

I do not believe that's the best metaphor for the future. We need to think about the internet as a student that we are teaching.

And we'll get into that too. So what I'm up to.

My role now is head of research at Moodle. What that means is that I'm out of the bureaucracy I created and I have a budget to do whatever the hell I want to do.

Which is a lot of learning. This is an AI generated image obviously. All the AI generated images have this symbol on them just so you know.

And that's the practice I encourage you to do with your own AI content.

Now this is not a great image but I put this up here because I put the same kind of prompt into the new bit Apple iOS system.

So this is generated on board an iPhone as a completely standard core feature of every iPhone in the future.

And while it might not be as clear because it has access to all my private photographs and things it's actually a pretty good likeness of me.

And that is a general trend you're going to see is this increased personalization of AI and it's building into the very tools that you are using all the time.

So the research team there's only a couple of people that's not as impressive as this.

But you're welcome to join. We have an area on Moodle.org where we have quite a very healthy community of people discussing particularly AI and online learning together.

We have a meeting every two weeks actually and we usually get between 20 and 40 people discussing these topics.

So let's talk about trends.

So think about the future of education and technology. We need to look at the whole world because it's bound up with everything that's happening.

What is it to be human? What is it to be part of the universe? You know existence. It's all connected.

So these are the kinds I made the time to write down a list of all the things that I'm studying.

What I've been looking at for the past well always really but particularly in the last year.

And these are those things that I'm researching. So I have feeds and things that I've selected and I'm spending a lot of time studying all of these things.

Particularly how they all converge and they're all connected. So I won't go through them all but you can see some of the things that are there.

Down the bottom here though I'm spending a lot more time being involved with my local community than I ever had before.

And that gives me so much satisfaction. And if anybody here involved in a local community thing of some kind,

I've been seeking them out, finding them, joining them and it's just really satisfying.

And as somebody who's made a whole last 20 odd years of online learning, it's really good to find.

So really I'm a lifelong learner. Anyone here see who would call themselves a lifelong learner? Yes, that's what you're here, right?

That's what it's all about. Now you probably want to know what's happening with Moodle.

Some of you want to know some have got a couple of slides because I can just get that out of the way.

I've conducted research over the past year including three surveys of the community, a lot of engagement, these fortnightly discussions I'm talking about, a lot of analysis, etc. etc.

So this isn't just pulled out of the air. But the basic direction of where Moodle is going to be heading with AI is with educators to save admin.

Nobody likes admin and admin can be simply converting a document into a form to put it online.

It's a lot of clicks, it's time consuming, it's hard sometimes.

So some of the projects getting on the roadmap are generating resources and activities from uploaded files.

Not just asking in AI but I have a bunch of stuff that I've already collected and used and I want to turn that into an online course.

Helping you with configuration, reviewing your learning design with suggestions.

There's one down here, pre-grading assignments.

Grading assignments is one of the most tiring things that an educator faces and if you had help with that, you're still making the final decision.

But support with that admin task is an important one and there's other things you can see there.

Now, so as I want to say though, the goal of all of these things is to produce really high quality courses.

So that you have a baseline truth of that subject. Whatever it is, but you've got a quality course and that's really important.

It's all about improving quality. Most online courses don't get to a good level of quality because of all the hard work and admin involved in making one.

With students, the overall thrust is now that you have a high quality course about personalising, customising and fitting that course into your life as a student.

So you've got a base truth of a great course with a lot of interaction and all the best pedagogical practice and now the AI is helping the admin of the student in engaging with that course.

And that means things like generating personalised content. It's an English course but I only speak Spanish.

Okay, well, it translated for me. Or it's a university level course and I'm in high school but I'm interested in that subject.

So adjust the reading level. I used to be in construction and now I'm learning physics.

Convert the examples into examples that use my previous knowledge and experience. So make it more relevant to me.

Those kinds of things. And so on. So I won't go on about that. I could go on about that for a whole hour but I won't.

I love this one here particularly simulations of realistic scenarios.

So you're learning how to be in a particular situation with people and you can practise that situation over and over again with AIs that are role-playing and you can practise in a safe environment.

So these are all ideas that came out of the research as the most priority ideas from the Moogle community. We can go outside the Moogle community by the way much.

Now let me talk about the automated world that is coming. You are here. The AI you're seeing now is the worst AI you will ever see again.

It is improving exponentially. There is a little bit trangier in the last few weeks, month maybe to talk about.

So I think that it's not going to scale any more than it is and so on. That is not true. There is so much work and billions of dollars and thousands of companies and millions of people working on solving these problems.

And I'm reading the papers. There are solutions for most of the problems. Most of them.

On the Gartenhub cycle I think we're about here. There is a bit of going down a bit, going downwards thinking about it. I'm not sure if this applies fully but definitely there is a lot to come.

What large language models do essentially is summarise everything they've seen. That's essentially what they're doing.

So they look at the internet and they summarise the internet and turn it into a product and say what do you think about this?

That's kind of what we do. We're all in education. Why are we in education? Because we want to get people to see things and think about things so that we can influence their future behaviour.

And that's how we work as well. It's not surprising because language models are based on neural networks much like our brain.

And you've no doubt seen, I don't know, I'm clicking this a little bit here, you see what things can happen. There's video generation.

There's just a brand video, some AI generated videos. They can generate them because it's seen videos about us and videos.

It's not plagiarism really at all. It's generation just like we generate ideas too.

I'm on a timeline here. Text and search. Have you noticed that anybody here has used perplexity?

So they were the first to really go harder to search but now you notice all the big platforms are getting into search and more than getting into it, they're making it by default.

Perplexity is quite good in that it will go off and search the internet for data. It doesn't get it from its own brain. It's searching things, finding relevant results and it summarises those and it actually is doing a search for you.

So it's a researcher.

And here, I did it last night I think, I just said give me a summary of your review of open education resources in the last decade and then there's a button at the top where you can turn that result into a webpage and it becomes another piece of content on the internet.

Music generation. I made this little song yesterday.

In a quiet classroom as the doors swing wide, the world comes in.

Not a real person. It's a completely generated public sector.

Knowledge share freely and every end reach, burning together.

Now, world generation. There are whole worlds, people that generate, even if they play Minecraft, somebody put a thousand AIs into Minecraft, they're all playing Minecraft as players.

They started forming little democracies. They have all these beautiful stories. It's like a whole little community building up as they all learn to live with each other and they start helping each other and so on.

On this, another game was released recently where the whole game is generated in real time, like what you see was a Minecraft type game again, in real time.

And you could say turn left and you would see what's over on the left and it's all being generated on the fly. There is no model, there is no programming, there's nothing behind it.

It's all just video generated on the fly.

Has anybody chatted with your voice to chat GPT, used the multi-modal theme? Right, I won't show it now, it's quite impressive.

But talking to machines is going to be everywhere all the time, part of our future.

This is a podcast. Generated using notebook LL.

So that's turning, actually generating some text using complexity and then turning it into a podcast discussion.

It's Moodle. Moodle that opens the world's platform. We hear so much about it, it's used by ancient leaders all over the world.

And Mexico, oh they're not just using it, they're basically running stories. It's very possible.

Now all the myths and reality, I'm a big fan of the Vision Pro, I've been using VR since 1996.

Finally this Vision Pro, anyone try the Vision Pro? Apple Vision Pro? Yeah, one of you.

Finally it's a proof of concept that you can have something on your face that produces digital things in your environment that are beautifully represented and connected with the real world.

So you get a window, I can get a window, I can put it up there on the wall, it could be my computer screen or some app or something on the wall.

I can go back home to Perth, I could come back here this week, it will still be on this wall here in Brisbane.

Like that's an amazing amount of integration with the real world and the digital and the real world being combined.

Now that sort of experience is being miniaturized, right now it's a bit clunky, it's $6,000 straight, $3,000 US.

But it's kind of a peak into the future, everything is being miniaturized. These are real glasses, these wave by technology.

And you have screens happening on something as thin as these things that, a lot of you wear glasses, first thing you do in the morning you grab it off the thing, you put it on your face.

If that's connected to your phone, for a lot of the processing, it can be very lightweight and it means that the internet is with you all the time.

Robots are an extension of AI into the real world. We have an intelligence that is now embodied in the world.

It can walk around, see things, interact, react to things as we do and so it can learn and gather data from the real world as we do.

And that's a very different prospect from an LLM sitting inside a data center somewhere on the internet.

This is now, it's here with us in the world.

There are dozens of companies building these. This one is a one from China that is $16,000 US only.

It folds up like this, it weighs 35 kilograms. It has full use of fingers and motion.

It's open source, you can actually program it yourself.

That technology for the movement of the robot is getting cheap and small and light.

This movie we've seen at Bicentennial Man is around the corner.

This kind of robot that you have at home that you say, "Hey Bob, look, finish the laundry, I'm going out.

When I get home, have dinner ready for me in an hour, that will be possible in probably five years and you'll probably have to buy one."

And if I'm wrong, it's about five years, it'll be seven or something because of regulatory things and all sorts of other issues, but technically possible.

And a conference we go to in five years may have instead of staff serving us food and drinks out here robots doing that.

Because as soon as any organization sees that an automated solution is better or cheaper or faster or safer, they're going to do that.

Instead of paying a person to do some mundane job, if you can pay $16,000 and have your own robot that does that job 24 hours a day, even in the dark,

what choice do you think that organization is going to make?

And so there'll be a series of micro-decisions happening all around the place and you're just going to start seeing these things everywhere in the street.

If you haven't seen the movie I, Robot, go watch it again.

I was very dismissive of it when it came out and it has Will Smith, so you have to get over that.

But if you watch it again, it's actually surprisingly on target, so it's a good one to watch.

So what is the best case scenario that we have in this world?

We have a possibility of post-scarcity.

Right now a lot of our society is based on scarcity. There's a limited amount of stuff.

It's possible we could have a utopia.

We might solve the problem of universal basic income, so it's okay if there's 70 or 80% unemployment in the decades to come,

because robots are taking care of it all and they're taking care of us as well.

They're doing, you know, we're all okay, we're supported.

This is assuming a lot from our governments, right?

But unfortunately no one is talking about that in their election campaigns.

You might say, "What would we do?" Well, this is Maslow's hierarchy of needs.

These things will be largely automated.

Physiological needs, food, war, rest and so on, security, safety.

So we're going to focus on that.

There's plenty to do in there because a lot of us don't get a lot of time for that because we're spending eight hours a day working.

So earn money to pay for food and survive.

We're all working for machines, right?

A university is a giant machine.

Every organisation is a giant machine that we've constructed.

We'll have more and more free time to learn, pursue your hobbies, support your projects, engage with the local community,

save the earth and care for each other.

And less and less grind working for organisations just to earn a living.

We know that will work out because this is what rich people do.

As soon as you don't have to work for a living, just look at those people.

What do they do?

Well, they're supporting charities and they're hanging out with their friends and they're looking after their family

and they're going on holidays and they're having a wonderful time.

Wouldn't it be great if we all had that life?

It's not impossible to imagine because some of us already have that.

The worst case, well, Cyberpunk 2077, corporations are renting everything to us and we're under their control.

We've got the worst case scenario maybe of the Trump Musk revolution.

I don't know, who knows?

But it could be that.

These AIs are also the most powerful mass surveillance devices ever invented.

So it could go very, very bad.

So very quickly, just to address the title, what could education feel like in 2034?

Well, there are two main types of learning that I like to think about.

You've got your big chunks and your little chunks.

The little chunks is your informal learning.

So you're curating your own good trustworthy feeds, your gurus.

You probably do this now, right? You follow people you like.

You get feeds of information. They're blogs, they're Twitter's, they're whatever, patreons, right?

You search out these things and that's your ongoing learning.

You'll also have AI in the closest device, it might be on your face,

that is constantly helping you learn about what you're looking at.

I was in Costa Rica lately.

I've seen a lot of plants and animals I've never seen before.

I'm all whipping out my phone all the time. I'm just identifying this, identifying that.

I'm just learning what I'm looking at.

And if that was in my glasses, it would just be pre-labelled for me.

I'm just looking at the world and I'm going to say, "I'm in learning mode right now.

"Just show me what I'm looking at."

I could be looking at the audience and it could have all your names floating above your heads.

Or if I'm looking at a jungle, so this is the name of that plant. It's the name of that animal.

And then you've got your formal learning and a lot of you work in these institutions now.

I believe there is a massive future for these organisations, but they do need to reform.

So you've got your high quality courses.

So when you do want to learn something, you want to go to your local university

and hang out with people who also are into that thing and learn it in the best, most efficient way you can.

AI will be helping fit that into your life.

You can trust what's going on there because it's all transparent and it's lifelong.

I mean, why leave university ever?

University is just part of your life. You've got time for it now.

Alright, there is a crisis, and I'm really running out of time.

There is a crisis happening though. It's not all rosy.

Now there are many crises.

I've come over to the guy who talks about the multi-crisis situation, but

you know, let's just call it a cluster.

I'm going to focus on one of these things because the disinformation and misinformation.

Now you know this slide and you also probably know that a lot of these sustainable development goals,

which we are not on track on meeting by the way, we're all doing a very bad job.

But we, because you're here, I know you believe education is the root problem.

The root thing we need to solve because it helps solve all the other things, right?

Quality education though requires quality information.

And so does AI because it's summarizing everything you've seen, right?

Oh, you can take a photo if you want to be quick.

So these slides will be on the link at the end of your seat.

So this is me having a chat with Karma Migo, and I convinced it that it's subtracting 2 from 15 gives you 12.

Like, it's very easy to sway current chatbots in the wrong direction.

This paper shows that a shocking amount of the web is already machine translated.

So it's already been converted not always perfectly by current tools.

You may have seen this paper that actually got into, actually got published.

I can't remember if this one was in nature or not, but they had some AI generated images that were so clearly wrong

that they passed review and got published.

Here's another one. There are thousands of examples.

Look, I'm very sorry, but I don't have access to real-time information or paper.

Someone's used some sort of chatbots to generate their paper and didn't check it very well.

And neither did the reviewer of Radiology Case Reports.

Here's another one.

Certainly, here is a possible introduction for your topic.

This is a Elsevier. What a lovely publisher.

This one's in nature, right?

Now, this is interesting because AI models collapse when you recursively train them on AI generated data.

And that's a very interesting point we need to keep in mind.

They had some pictures showing when they trained an image generator on images that had generated

that they start off being able to do lots of dogs and they start becoming more samy

and then they end up with this mishmash and they turn their dog into some sort of a chocolate chip cookie.

Okay? Five minutes.

So, this paper talks about it's increasingly difficult to train newer versions of LLMs without access to data

crawled from before the internet started to be filled up with AI content.

And some estimates are that there's a 60% of content out there already is AI generated

because some people are just like slamming the internet with all kinds of junk.

There is also a massive cultural bias and I wish I had more time to talk about this

but even as an Australian I feel the US preponderance of LLMs.

And imagine if I was in Nigeria. You should imagine that.

Because there is a huge bias. Most of it is scraping things off the internet

so only people who like to put a lot of stuff on the internet get represented.

The biggest educational data set that we have out there currently is fine web edu.

It sounds fine doesn't it? But it isn't because what they've done is take a lot of web scraped data,

point it in AI added it and said does this look educational?

And it goes yes. Okay, so they will include that in the subset.

And this is what passes through educational resources in most AI.

This is a data set that AI, a lot of AI tools, a lot of language models use as training.

This is what it looks like.

I was a bit involved with the open source organization in defining the new open source

AI definition and the main thing that's been added over open source software

is that you need to be able to show where the data came from and what quality it was

and give someone else the chance to rebuild that system using the same data.

Here's the solution.

This is a proposal and I'm bringing this to you now because I want us to discuss this.

I believe Open Education Global is ideally located to be the organizing entity

with all these members to think about OER,

curate, collect, review and build an OER collection

that we can format into a data set.

Let's call it the OER data set, why not?

And advocate that all the AIs use that data set.

This is human checks, reviewed data, stuff that you're all working on and advocating for.

Bring it together.

OER Global is kind of in the name, right?

How are we not already doing this?

Okay, I appreciate the extra minute there.

We're going to get into this on the forums,

but some things we need to discuss, are there already other aligned projects

we can support or join that are doing thinking similarly?

Is there strategic partnerships?

Is there distinct funding that we have in OER Global enough?

This is a big technical thing, it's mostly just a people contributing thing.

Or do we need to find new funding?

Could we negotiate licensing for commercials?

Maybe we say, hey, open AI, you should use this and they'll go, yes please,

because we don't have enough data, we want to use it.

Pay us a million dollars a year and that will help fund us to make it bigger and better.

What tool should we use?

I do make one, MoodleNet, which is cease being a research project by way

as proper funding in MoodleNet, so it's kind of a bit grow.

But there's Calibri, there's Merlot, there's many other platforms.

Let's discuss, right? I don't really care.

I care that it happens, I don't care what tool gets used.

How do we empower everyone to contribute?

And how is there more ways to increase trust in this whole thing?

So that's some starter topics, I think we should discuss it.

This is the QR code you need to grab.

It takes you to the OE Global Connect discussion for this session.

And very shortly, I'm going to post a link to a proper discussion about the project.

And I'm going to include, I'm going to record a video of this and put it up there as well for anyone who wasn't here.

But that's where you should go.

So back to the takeaways.

That's what I'm talking about. This is a logical flow.

And to me it seems inevitable.

I feel like it's really important that someone does this and I feel like we're in a good position.

If OE Global can't do it, someone else should do it, but we need to do it as educators.

And I want to leave you with that thought.

The internet is no longer, we shouldn't think about it as a big library.

Just put things out and hope someone wants them or use them.

It comes along to the library. A lot of things never get read. They never get used.

We have to think about the internet as something that we're teaching.

We can't complain about AI without teaching it.

We've got to teach it to be better.

Just like you wouldn't complain about a bad student, right?

You'd go, "Okay, let me educate you. Let's educate the internet."

So thank you very much.

(Applause)