0:00 What's the difference between generative AI and agentic AI?
0:05 Well, they're two distinct approaches to artificial intelligence.
0:08 And I think we're all familiar with generative AI, things like chat bots and image generators and the like.
0:15 And they are really fundamentally reactive systems.
0:22 They wait for you to do something, specifically they wait for you to prompt them
0:29 and once you prompt them, their job is to generate some kind of content based upon what you provided in the prompt.
0:40 And they're using patterns they learned during training
0:44 The things that it can generate, well that might be some
0:48 text or it might be an image or it may be a piece of code or it maybe some audio.
1:00 These are all sorts of things that we can generate with generative AI and
1:04 they're essentially sophisticated pattern matching machines.
1:07 They've learnt the statistical relationships between words and between pixels and between waves.
1:14 And they've learned that from massive data sets.
1:16 So when you provide a prompt, GenAI predicts what should come next based on its training,
1:22 but it's work does end at generation.
1:25 It doesn't take further steps without your input.
1:29 Now, agentic AI systems, by contrast, those are not reactive.
1:36 They are proactive systems.
1:40 Now, like generative AI, they often start with a user prompt,
1:45 but that prompt is then used to pursue goals through a series of actions
1:54 and an agentic system basically goes through a bit of a life cycle.
1:58 So the way this works is it kind of first of all perceives it
2:03 perceives its environment if you like and then once it's done that it can decide an action to take.
2:13 Once it's decided that action, it can then execute that action,
2:18 and then once that action has been executed, it can kind of learn from the output and then round and round we go,
2:27 all with minimal human intervention.
2:31 Now, both of these AI approaches often share a common foundation.
2:38 And that common foundation is large language models or LLMs.
2:45 LLMs serve as the backbone for chatbots and yeah
2:50 there's actually other tools that are used for some of these other generative things,
2:54 diffusion models typically for images and audio,
2:56 but for chat bots we use LLMS and LLMs also provide the reasoning engine that powers agentic systems,
3:05 but before we go any deeper into that let's talk about some real world applications and use cases.
3:12 Now, maybe this doesn't put me in the best of lights,
3:15 but I don't think I'm the only one using generative AI
3:19 to help with the task of content creation and especially creative content creation.
3:28 Now, before work this morning, and this is completely true, I used the chatbot to help write the next chapter of
3:37 my Nelson Demille fan fiction novel and right now
3:40 you're probably thinking how profoundly cool and absolutely non-nerdy this guy is,
3:45 but for many of us gen ai does help with daily tasks.
3:50 Like let's consider how a Youtuber
3:53 might use a generative AI system to review scripts and suggest
3:57 thumbnail concepts and maybe even generate background music,
4:01 but at each step, there is a human.
4:05 There is a human creator and that human creator is looking at this generated content
4:10 and they are reviewing it, check it's what they want,
4:15 probably isn't, so then they are refining it as well and they are really going through and directing this whole process.
4:25 The AI generates possibilities but the human curates them.
4:29 Now, agentic AI that kind of thrives in scenarios
4:33 that require ongoing management and consist of multi-step processes, so not just one thing at a time.
4:43 So consider a personal shopping agent.
4:47 Given a product to purchase as input, it actively hunts for availability across platforms, it might monitor price fluctuations,
4:55 it might handle checkout processes, and it might even coordinate delivery.
4:59 Largely by itself, seeking input only from you, only when it's needed.
5:05 But how does it do that.
5:07 Well, it turns out that the LLMs that are behind much of generative AI
5:11 can also be used to provide reasoning capabilities to AI agents.
5:16 So this essentially here, we're using gen ai's ability to kind of
5:20 think in inverted commas there, and it's thinking through problems,
5:25 and this has a name.
5:27 It's called chain of thought reasoning.
5:34 and this is what LLMs are so good at.
5:38 It's a process where the agent basically breaks down a complex task into smaller
5:42 logical steps, kind of like how humans tackle difficult problems as well.
5:47 So let's imagine one.
5:49 Let's imagine that we want to have an agent that is planning a complex task like organizing a conference.
5:56 So what it's going to do is it's going to use gen ai to generate an internal dialog.
6:01 And that dialog might go something like this.
6:03 It might say.
6:04 First I need to understand the conference requirements of the size, the duration, the budget, that sort of thing.
6:09 Then I should research available venues matching those parameters.
6:14 Then it might think well for those venues that meet those requirements I now need to check availability and so on.
6:20 It's effectively the agent really kind of talking to itself to explore the problem space before taking action.
6:28 Gen AI is basically the cognitive engine driving an agent's decision-making.
6:33 Now looking ahead, the most powerful AI systems probably won't be purely generative or purely agentic.
6:41 They're going to be intelligent collaborators,
6:44 that will understand when to explore options through
6:46 generation and when to commit to courses of action through agentic action.
6:53 Like an agent that would know when to generate the next chapter
6:58 of fan fiction so it's ready after, I don't know, a video shoot. Maybe, uh...
7:04 Maybe it's ready right now.
