Why parents need to engage with AI now

Rob: Welcome to the National Virtual Expo. Rob Malicki on Gadigal land in Sydney, and welcome, parents. This is one of the only sessions I think I can confidently say, when it's labelled "parents," we probably have no kids showing up. So thanks for joining me at the National Virtual Expo. I'd like to introduce you to Aaron Driver from the University of New England. Aaron, welcome back to the National Virtual Expo.

Guest: Thanks, Rob. It's great to be back with talk number two.

Rob: We've already done a session on the first day of this expo about studying with AI and how to use that -- if you're a student-parent as well -- to actually improve your studying habits. So parents, if you didn't see that session, please go back and watch it. You'll find it in the session catalogue below. But today is about fundamentals. Maybe let's get straight into it and talk about the three things parents need to know now about AI, and then at the end we can get into three actions parents can take literally right now, today, to start feeling comfortable about this topic.

Guest: Great. Three key things. First, absolute number one: this is just not optional for your child or children. The children who thrive and do really well in the workforce in the next decade will be the ones whose parents took it seriously right now. You're probably here because your kid's future is on the line and you love them and you want to make sure they're well prepared. So that's number one -- it's not optional.

Guest: Number two: you don't need to understand AI intimately. You don't need to be an expert to guide them. A lot of parents we find, at university and in the schools work we do, disengage because they feel a bit stupid, like their kids know more about it than they do. But you don't need to know how a car engine works to teach a kid road safety. You just need to know the rules, the risks, and when to sit in the passenger seat with them. Stay engaged.

Guest: And three: the danger. We're going to have an AI safety chat today, and AI safety is so important. We have a mandatory AI safety course for all of our students at UNE -- we're the first university in the country to provide a full AI platform for all 22,000 of our students, which we started late last year. Students don't get access to that platform until they complete the safety course. But here's the thing -- the danger isn't the technology. The danger is the surrender. The real risk is a child who stops thinking because the machine thinks for them. And that's what should keep parents up at night. The good news is it's entirely preventable with the right habits. We call it cognitive offloading, or cognitive surrender. If your child is learning that the AI does the thinking for them, they're in trouble. But we'll talk about that today and ways to work against it.

Rob: I'm really excited about this point in history. My kids are sitting on the cusp of this too. So parents watching along, Aaron and I are literally in the same position as you right now. Our kids are coming up to that point in their schooling and we have to be having these conversations -- we practise what we preach here. But to me, the opportunity for people right now is one of the most exciting in the history of humanity. That might sound dramatic, but imagine buying Bitcoin in 2011, or being the first into a gold rush. This technology is already pervasive, and it's going to infuse every aspect of the workforce and many aspects of our lives. So you can do one of two things: you can scale back and resist, or you can see the amazing opportunity and reach out and take hold of it. And it doesn't have to be a big thing to begin with.

Guest: Run toward the giants. That's your only option now. Change is becoming a constant and the pace of change is accelerating, so you've got no choice but to lean in. I often use the parallel of fire as a technology. When fire came along, it allowed us to cook our food, which meant we could consume more calories. It allowed us to keep predators away at night, which meant we could get more REM sleep. Consuming more calories and getting more sleep led to a dramatic increase in the size of our brains, and a lot of changes down the track. You wouldn't have said, "I'm going to stay away from that technology." Of course, fire can burn you and cause a bushfire, so you've got to be careful. But you're not going to run away from it. And I think we don't know what the final long-term outcomes of AI will be, but the only way to have a role in that is to engage with it.

Rob: You can't run away from it anyway, can you? Technology doesn't care how you feel. It's here now. You can accept that, or it's going to be there anyway. The tech always wins in the end.

How large language models actually work

Rob: Let's start to break this down for parents. We've called this a fundamentals session, so where would you like to start?

Guest: Let's talk about what we mean when we use the term "model." A large language model is a new type of AI invented in a famous 2017 paper by researchers at Google DeepMind. The paper was called "Attention Is All You Need," because it's about the attention mechanism in a transformer stack -- all very technical-sounding jargon. But this technology was only invented a few years ago and it's already swept over the world. Let me share my screen and show you how it works. I'll use our AI platform at UNE, which we've named MAGIC, after our founder Sir Robert McKnight -- UNE was founded to serve people in regional communities, we were the first university to do distance education, then the first to do purely online degrees, and now the first university to really embrace AI for all of our students.

Rob: And I was really grateful when you accepted the invitation to come and share this with us, because universities are dipping their toes into this at varying speeds, and UNE is really at the cutting edge.

Guest: It's a pleasure, Rob. So here's a useful way to think about AI: imagine your new digital intern. It's brilliant, it's read almost the entire internet, it's incredibly fast -- but it has zero life experience, no common sense, and it's on its first day on the job. Let me show you what that means and show you how the AI thinks, because the AI isn't actually thinking. The task of the AI is to predict the most likely next word. Under the hood it's predicting tokens, which are mathematical representations of words, but we'll just say words.

Guest: So here I've given it this instruction: "Complete a sentence with only the single most statistically likely next word. Here is the sentence: The first person to walk on the moon was Neil..." And it returns "Armstrong." The reason it returned that is because it has seen that sequence of words millions of times. It correctly predicts the next word. This is the power of AI -- it recognises patterns in vast amounts of text, images, sounds, and videos. The AI didn't think. It just detected that pattern in its training data.

Guest: Now let's watch what happens when I give it a question it can't possibly know the answer to. Same instruction: "Complete this sentence: This morning for breakfast, I ate..." It's come back with "a bowl of cereal." So the AI has said I ate a bowl of cereal for breakfast, even though I haven't. What's the AI doing here? It's not lying. It's just doing what it's designed to do: complete the most likely next word in the sentence. Every conversation you have with the AI is just predicting the next word, just doing so incredibly fast. It feels like you're talking to a person, but it's not thinking, it's not reasoning, it doesn't have any actual embodied understanding of the world.

Rob: That's a really good example of how the AI thinks.

Guest: And what that leads to is what's called hallucination -- where the AI apparently makes things up, like telling me I had a bowl of cereal. Hallucination is when an AI confidently makes things up because, to the AI, it's just the most likely next word. Its intention is good, but it often comes up with made-up information. So the important thing for parents to understand, and to explain to your kids, is that the AI is a prediction machine, not a thinking machine. It is always just making a highly educated guess about which word comes next, word by word.

Rob: So when we talk about "models," we're essentially talking about different ways of doing that prediction. Each one is provided by a different company. They're always working on their model -- their system for predicting what the next word is. Is that right?

Guest: Exactly. You get small models that are fast and cheap to run, but they're not as capable because they don't have as many patterns in their training data. And then you get larger models which are more capable. The main three model companies in the world are Anthropic, which makes Claude; OpenAI, which makes ChatGPT; and Google, which makes Gemini. They each have their own model families, and they're each good and less good at different things. As you use different models you get a feel for it.

Assistants, agents, and tools: the AI ecosystem explained

Guest: The important thing to understand is this: you have an underlying large language model, or LLM, that powers the conversation. The layer on top of that is your assistant or your agent. In this case I've just created a chat assistant, but you can choose other assistants like academic design or academic research. Those assistants are like personalities -- each has a system prompt that tells it what its task is. Your job is to run research for me. Your job is to teach me things. Our students have an assistant called Study Mate, which is designed to help students study. It can use any model as its brain, but the assistant itself is built with its system prompt and instructions.

Guest: It's like a digital person: you've got the personality, but you've got a brain that powers it. And the last piece is the tools the AI can call -- a search tool, a database, the weather, the time, and all sorts of other tools. Those three things -- the assistant or agent, the model, and the tools -- make up the ecosystem of what your children will be using when they work with AI.

Rob: I love that. When I first started hearing about AI agents, my mind went straight back to the 1990s Terminator movies. But when you niche down into it, it's actually just a specific configuration designed to do one thing really well. It's a bit like being in a big company -- you might have the office manager handling scheduling, the account manager interacting with clients, and a secretary dealing with emails. An AI agent is a similar concept: a specialist doing one thing.

Guest: That's a great way of explaining it. And it's worth noting the difference between an assistant and an agent. An assistant is something static, like a chat bot. Whereas an agent has agency -- it can carry out actions, report back, talk to other agents, and keep going. We've seen these AI personalities evolve from assistant to agent in the last six months. So if you hear your kids talking about assistants and agents, now you know the difference.

Rob: And that evolution in just the last three or four months has been extraordinary. The ability to create your own agents has been there for a while, but to do it at scale -- asking one agent to create multiple agents to do multiple things -- that has taken an incredible leap forward just recently. And that's the advantage of embracing it rather than approaching it with fear.

The power of AI as a research tool -- and the risk of cognitive surrender

Guest: I want to show parents how powerful this is, because when parents understand the power, they can understand the benefits it can provide to their children. Here I've asked a researcher assistant to do a mini literature review on supply and demand for a high school economics student. It ran several parallel searches across the tool to cover the key dimensions, got good results, and then of its own accord decided to run more targeted searches to fill out the picture. Then we get the mini literature review with an executive summary, the top five papers each with a one-sentence insight, gaps and limitations in the literature, and an annotated bibliography. That's a few days' work done in a minute and 37 seconds.

Rob: I remember when I was at university -- 25 years ago -- if I wanted to research something, I'd go to the library, look up a list of topics on a computer, get the title of an academic paper, walk up to the top floor to find the shelf reference, pull out the journal, flip through to find the article, and then read the whole thing to work out whether it was even relevant to my assignment. That would literally take hours.

Guest: And this is where the cognitive surrender point comes back in. It is a student's job to take that paper, jump into a new tab, open Google Scholar, find that paper, download it with their library account, and read it. We're not saying the AI does everything for them. We don't want your children cognitively surrendering. But as a research tool, it's incredible.

Privacy, free tools, and why paid subscriptions matter

Rob: You've got me thinking about investing in the tool, if you can. That's one of our three actions we want to get to. It's honestly a fantastic investment -- we're talking around $30 to $40 a month. We'll talk more about the three actions at the end of the session.

Guest: Yes, and I'm glad you mentioned it because I want to talk about privacy and the advantages of a paid tool. When you have a free tool, your data is being captured by the AI company and they are training their future models on your interactions. So the AI is capturing how you -- or your children -- think, study, learn, and approach the world. In all of these tools, whether it's Claude, ChatGPT, or Gemini, you can go in and turn off that training in the settings -- just Google how to do it. Sometimes on a free account you don't have the option to turn it off, but on most accounts you can. Students at UNE are never trained on their interactions with the AI. Their interactions are completely private and stay in the country. Not being trained on is a really important thing for you to keep an eye on as a parent.

Rob: This is one of those key things that's really good to understand. Every time you see your kid using something for free -- downloading an app, using something on the internet -- there's a reason why it's free. If you're not paying for the product, you are the product. They're monetising you and monetising your attention. So definitely keep that in mind. Or send your kid to UNE and they'll get the AI for free.

Three actions parents can take this week

Rob: Let's talk about our three actions. The first, which we've just mentioned, is invest if you can. This is honestly a fantastic investment, and you can use it yourself. If you're watching this session and still feeling overwhelmed, invest in a $30 or $40 subscription for one month, sit down after the kids have gone to bed with a glass of wine, and ask the tools to explain themselves to you. Ask them to train you on how to use them, because that's the power of them -- you can literally ask them to do anything, and they're very good teachers.

Guest: Point number one: get them into an AI service this week, if they don't already have one. Claude, Gemini, ChatGPT -- my personal favourite is Claude, but pick one. Set it up with them. Sit down with them while they use it for the first time. You wouldn't hand your kids a car and walk away, so don't hand them this either.

Guest: The second thing I'd suggest is have the jobs conversation with them, but don't catastrophise. You don't need all the answers. You just need to open the door and say, "For the job you'd like to do -- what do you think AI means for that?" And listen more than you talk. The goal isn't to solve the AI jobs question. It's just to signal that in your family, that's the sort of stuff you talk about and that you're in it together. Even just asking some questions and listening is hugely beneficial and way better than doing nothing.

Rob: And on that -- we do have a session at this expo on AI and the future of work. If you want to have a bit of an advantage in that conversation with your kids, please come to that session. We're going to talk about what this technology means for the workforce, for degrees, and for particular careers. It'll give you some of the base knowledge you need to have that conversation.

Guest: I can't wait for that chat. So, one: get them into AI services this week. Two: have the jobs conversation with them this week. And the third one is watch for cognitive surrender. You will see it. Homework done in 30 seconds with no thinking. When your child accepts an AI output without question, that's the moment for you to lean in -- not with punishment, but with curiosity. "What do you actually think about this? What could you have added to that assignment? Does it bother you that it doesn't sound like you?" Just ask a single question. "That's what the AI thinks -- but what do you think?" Sit down with them. Watch for the surrender and ask a question. It's just a great habit to get into.

Rob: What popped into my mind -- and it might be a bit awkward depending on your relationship with your kids -- is ask them to teach you how to use it. If you're not feeling confident with this technology, go and say, "I want to understand this better -- can you show me how to use it?" Once you put them in the position of being the expert, it'll be entertaining for you, and you'll get to see how they think and interact with the tool in a non-threatening way. I think that's an easy way to open the door for those conversations.

Guest: That's such a great idea. Any teacher will tell you -- you learn so much by teaching a subject. So not only will you learn, but your child will learn as well.

Closing thoughts and resources for parents

Rob: This has been a fantastic conversation. Something we've created for parents to help get you started is a worksheet you can download from the session description below. Everything we've talked about throughout this session is in there to help guide you through the early stages. I personally believe the opportunity is there, and don't wait -- if you walk away with anything from this expo, this is one of those things you really want to get into straight away.

Guest: Full credit to you, Rob, for the worksheet idea. And the upcoming sessions on AI safety and AI and jobs really complement this chat. I just encourage parents: run towards the challenge. Sit with your kids, lean in, and use the worksheet. Really sit down, ask some questions, have those conversations, and engage with them. There literally is no more important technology you could engage with your kids on than this right now.

Rob: Aaron Driver, Chief AI Officer for the University of New England -- thank you so much for joining me at the National Virtual Expo. We'll catch you again a little bit later in the week.

Guest: Absolute pleasure, Rob. Thanks for having me.