presentation/thames-valley-ai.html at master · digiguru/presentation

809 lines (695 loc) · 35.7 KB
<!doctype html>
<html lang="en">
		<meta charset="utf-8">
		<meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0, user-scalable=no">
		<title>Being the Human in the Loop: Thriving in the Age of AI-Driven Burnout - (Thames Valley AI - v4.1  - 16/09/2025)</title>
		<link rel="stylesheet" href="dist/reset.css">
		<link rel="stylesheet" href="dist/reveal.css">
		<link rel="stylesheet" href="dist/theme/black.css">
		<link rel="stylesheet" href="dist/theme/AND.css">
		<link rel="stylesheet" href="custom.css">
		<!-- Theme used for syntax highlighted code -->
		<link rel="stylesheet" href="plugin/highlight/monokai.css">
		<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.5.2/css/all.min.css">
		<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-RZP8CG4KLJ"></script>
  window.dataLayer = window.dataLayer || [];
  function gtag(){dataLayer.push(arguments);}
  gtag('js', new Date());
  gtag('config', 'G-RZP8CG4KLJ');
		<div class="reveal background">
			<div class="slides">
				<section data-background-image="assets/robot-future2.png">
					<section data-auto-animate data-state="focus-bg">
						<h1 class="standout">Being the Human in the Loop</h1>
					</section>
					<section data-auto-animate>
						<h1 class="standout">Being the Human in the Loop</h1>
						<p>Thriving in the Age of AI-Driven Burnout</p>						
					</section>
					<section data-auto-animate>
						<h1 class="standout">The 20th version of this talk</h1>
						<p class="fragment fade-up">Since I first ran this talk</p>
						<p class="fragment fade-up">over 64,000 projects have been made</p>
						<aside class="notes">
						This is the 20th version of this talk and it’s going to be the most honest and open iteration I’ve ever done because I’m going to be sharing a story about Burnout, and how I survived it.
						Since I first ran this talk over 64,481 projects have been tagged with ai on github, and many many more AI announcements. 
						</aside>
					</section>
					<section data-auto-animate>
						<p>over 64,000 projects have been made</p>
						<img src="assets/tools.jpg">
					</section>
					<section data-auto-animate>
						<p>Example...</p>
						<p class="fragment fade-up"><img width="50%" height="50%" src="https://www.ee.ucla.edu/wp-content/uploads/ee/ozcan-Optical-Generative-Models-scaled.jpg" width="80%" height="80%" /></p>
						<aside class="notes">A team of researchers at UCLA led by Aydogan Ozcan.
							have developed a new type of generative AI that uses laser beams and optical components instead of traditional GPUs.
							This "light-based" system generates images with millions of times less energy and almost instantaneously, 
							It also introduces a new method for ensuring privacy, as images are encoded in a unique optical pattern that only the correct decoder can reconstruct.</aside>
					</section>
					<section data-background="black">
						Things are moving fast
					</section>
				</section>
				<section data-state="intro" data-background-image="assets/adamhall.jpg">
					<section data-auto-animate data-state="focus-bg">
						<h1 class="standout">Adam Hall</h1>
						<aside class="notes">
						Hi - BTW I’m Adam Hall. I know this because that chap looks like me.
						</aside>
					</section>	
					<section data-auto-animate>
						<h1 class="standout">Adam Hall</h1>
						<p><img src="assets/clients.png" width="40%" /></p>
						<aside>
							I’ve worked with scores of brands, many of which you will recognise.
						</aside>
					</section>	
					<section data-auto-animate>
						<style>
							.logo-text-line {
    display: flex;         /* Turns the <p> into a flex container */
    align-items: center;   /* Vertically aligns the text and logo */
    gap: 0.5em;            /* Adds a small space between the text and logo */
    justify-content: center; /* Optional: Centers the whole line on the slide */
.logo-text-line img {
    /* Set the logo height relative to the text size */
    height: 1em; 
    width: auto; /* Maintain aspect ratio */
						</style>
						<h1 class="standout">Adam Hall</h1>
						<p class="fragment fade-in-then-semi-out" data-fragment-index="1">Former Agile Coach</p>
						<p class="logo-text-line fragment fade-in-then-semi-out" data-fragment-index="2">
							<span>AI Principal at</span>
							<img src="assets/AND_logo.png" alt="AND Digital Logo">
						<p class="fragment fade-in-then-semi-out" data-fragment-index="3">Amateur AI Artist</p>
						<p class="fragment fade-in-then-semi-out" data-fragment-index="4">Burnout Survivor</p>
						<aside class="notes">
							I used to be an Agile Coach, where I was obsessed with helping teams break down work into smaller iterations and get faster feedback loops.
							I’m currently an AI Principal, where I obsess with how to help teams embrace AI to make the work they do more agile, with faster feedback loops and more fun.
							I’m also a reputable AI Artist who is on a number of Creative Partnerships with big Image Generation AI services.
							I’m also a burnout survivor and I’m here to share with you all how to protect yourself from burnout and give you a framework to succeed in AI Innovation
						</aside>
					</section>	
					<section data-background="black">
						<p>“Speed up the feedback loop”</p>
						<aside class="notes">
							My motto is “Speed up the feedback loop” - and I’m going to share how to do this carefully.
						</aside>
					</section>
				</section>
				<section data-background-image="assets/robot-art2.webp">
					<section data-state="focus-bg">
						<h2 class="standout">
							Being an AI Artist
						</h2>
						<aside class="notes">
							I joined the AI Art community on 12th July 2022 when midjourney dropped it’s first model. Every since then I’ve been on a journey to make AI art and share it with people.
						</aside>
					</section>
					<section data-auto-animate>
						<p>I want to share beautiful art with the world</p>
					</section>
					<section data-auto-animate>
        /* The main container that holds all the grid items */
        .grid-container {
            display: grid;
            /* This is the key property: it creates 5 columns of equal fractional width (1fr) */
            grid-template-columns: repeat(5, 1fr);
            gap: 1.5rem; /* This creates space between the grid items */
            padding: 1.5rem;
            font-family: sans-serif;
        /* Styling for each individual cell in the grid */
        .grid-item {
            border: 1px solid #ddd;
            border-radius: 8px;
            padding: 1rem;
            text-align: center;
            background-color: #f9f9f9;
            box-shadow: 0 4px 6px rgba(0,0,0,0.1);
            transition: transform 0.2s ease-in-out;
        .grid-item:hover {
            transform: translateY(-5px); /* Add a slight lift effect on hover */
        /* Styling for the paragraph text */
        .grid-item p {
            margin: 0 0 10px 0; /* Remove default margins */
            font-weight: bold;
			font-size: small;
            color: #333;
        /* Styling for the image within each grid cell */
        .grid-item img {
            max-width: 100%; /* Ensures the image is never wider than its container */
            height: auto;    /* Maintains the image's aspect ratio */
            border-radius: 4px;
        /* --- Responsive Design --- */
        /* For tablets and smaller devices */
        @media (max-width: 992px) {
            .grid-container {
                grid-template-columns: repeat(3, 1fr); /* Change to 3 columns */
        /* For mobile phones */
        @media (max-width: 600px) {
            .grid-container {
                grid-template-columns: repeat(2, 1fr); /* Change to 2 columns */
    </style>
						<div class="grid-container">
                    <div class="grid-item">
                        <p>V1 - Feb 2022</p>
                        <img src="https://cdn.midjourney.com/13e7a134-63cb-4c9a-8eec-12eb6706a39e/0_2.png" alt="Midjourney V1">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V2 - Apr 2022</p>
                        <img src="https://cdn.midjourney.com/09bdc755-cfd2-4e1a-9d64-f29053b407cb/0_0.png" alt="Midjourney V2">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V3 - July 2022</p>
                        <img src="https://cdn.midjourney.com/6a28a794-5703-4634-afd9-b82fcdac048f/0_3.png" alt="Midjourney V3">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V4 - Nov 2022</p>
                        <img src="https://cdn.midjourney.com/acae9b21-6922-4caa-a2a5-f4ac53eda796/0_3.png" alt="Midjourney V4">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V5 - Mar 2023</p>
                        <img src="https://cdn.midjourney.com/158883d9-3922-4f31-9d45-99a8e308767f/0_2.png" alt="Midjourney V5">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V5.1 - May 2023</p>
                        <img src="https://cdn.midjourney.com/0e8522d4-74fa-4530-81a4-0719b9319034/0_3.png" alt="Midjourney V5.1">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V5.2 - June 2023</p>
                        <img src="https://cdn.midjourney.com/79d29616-0eac-4b53-9090-791996269aa1/0_1.png" alt="Midjourney V5.2">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V6 - Dec 2023</p>
                        <img src="https://cdn.midjourney.com/11d066a8-0b2e-4752-8b5d-50faeb95376e/0_1.png" alt="Midjourney V6">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V6.1 - July 2024</p>
                        <img src="https://cdn.midjourney.com/e8e205ec-b4ee-4f34-a84e-fe4a19ee7088/0_1.png" alt="Midjourney V6.1">
                    </div>
                    <div class="grid-item fragment fade-up">
                        <p>V7 - Apr 2025</p>
                        <img src="https://cdn.midjourney.com/2cfe2f1f-a779-4446-b023-82efafc9a698/0_1.png" alt="Midjourney V7">
                    </div>
                </div>
						<aside class="notes">
							Midjourney has revved it’s model approximately once every 4 months, slowly getting it better and better over time. For example here’s an output from it’s very first version.
							ANd now I’ll upgrade the version until we have today’s model. Please shout out if you can guess the prompt at any time 
							Watch as we skip forward about half a year with each version until we arrive at todays.
							Darth Vader as the girl with a pearl earring
							And this is why Disney is suing Midjourney, right?
						</aside>
					</section>
					<section data-background="black">
						Always be trying out the latest model
					</section>
				</section>
				<section data-background-image="assets/robot-happy.png">
					<section data-state="focus-bg">
						<h1 class="standout">AI Art for Business</h1>
						<aside class="notes">
						Exploration of the latent space in AI art is fun, but I'm here to share real world ways of using AI art in your business
						</aside>
					</section>
					<section data-auto-animate>
						<h2>LLM Image Classifiers</h2>
						<p class="fragment fade-up">Explore with <a href="https://gemini.google.com/" target="_blank">Gemini</a></p>
						<aside class="notes">
							One of the things I see people do as they start to play with LLMs is they start piping images into an LLM to get it to do classification. This is a smart idea, you can use a mutli-model LLM like Gemini or ChatGPT. You just upload an image and you can ask questions about the object. This is a technique we tried recently on a client where we had to discover the type of article uploaded by users.
							Note: this is fairly expensive. Typically you are using a sledgehammer to crack a nut. You’ll succeed but you’ll have spent all your budget on a sledgehammer. I only advise it when you are playing with PoCs. 
						</aside>
					</section>
					<section data-auto-animate>
						<h2>Training your own model</h2>
						<p class="fragment fade-up">Explore with <a href="https://teachablemachine.withgoogle.com/train" target="_blank">Teachable Machine</a></p>
						<aside class="notes">
							Another great tool you can use is “Teachable Machine”
							This allows you to train an AI model with bespoke images.
							Rather than try to explain it I thought I’d do a live demo.
							Here’s teachable machine. We’ve used it to do strange experiments like checking pipes for defects, by taking photos of defective pipes and then using this to spot moments where pipework could be defective from a video of a camera being pulled through it.
							So the advantages of this is that you have more control. It’s also a gateway drug to more extreme versions of Machine Learning. The really cool thing is this has a really low energy footprint so can be put on embedded devices meaning you can easily create PoCs for embedded devices and start utilising your environment and helping it become smart.
							These are basically free to run in-process or you can host them, and you’d pay cloud compute costs.
						</aside>
					</section>
					<section data-auto-animate>
						<h2>Bespoke Models</h2>
						<p class="fragment fade-up">Explore with <a href="https://replicate.com/tmappdev/lang-segment-anything" target="_blank">Replicate</a></p>
						<aside class="notes">
							The next level up you might want to try is running some pre-trained bespoke models. There are hundreds out there and alot of experiments can be found opensource. 2 particular favourites of mine are Replicate and Hugging face. On these you pay per run, and that will be as much as the computer takes.
To give an example we’ve recently done some work with a client where they had thousands of tiny tasks that were done by designers. The tasks were dull. To give some context they were doing straightforward activities on photos of products that were taken in the natural environment (let’s just say). 
Things like 
Fixing brightness levels
Removing “junk” from the images
Making the weather brighter
Adding some “in-house” clouds to images
These tasks are menial for trained designers, they want to make cool products not clean up awful product images.
So here let me demo how you can find and use a few models.
https://replicate.com/tmappdev/lang-segment-anything
https://www.combinedshipping.co.uk/Graphics/Dirty%20Hot%20Tub(1).jpg
https://replicate.com/zf-kbot/inpaint-and-guess-prompt
						</aside>
					</section>
					<section data-background="black">
						Think small, optimise after
					</section>
				</section>
				<section data-background-image="assets/robot-art.png">
					<section data-state="focus-bg">
						<h1 class="standout">My Art Journey</h1>
						<aside class="notes">
							Back to my art, things were going well here too. I dedicated to sharing at least one image a day on social media. Here’s an early image I created.
						</aside>
					</section>
					<section data-auto-animate>
						<img src="assets/art-example.jpg" style="width: 50%; height: 50%" />
					</section>
					<section data-auto-animate>
						<img src="assets/art-example.jpg" style="width: 50%; height: 50%" />
						<p>2 likes</p>
					</section>
					<section data-auto-animate>
						<img src="assets/art-example2.jpg" style="width: 50%; height: 50%" />
					</section>
					<section data-auto-animate>
						<img src="assets/art-example2.jpg" style="width: 50%; height: 50%" />
						<p>11 likes</p>
					</section>
					<section data-auto-animate>
						<img src="assets/example-wes1.jpeg" />
					</section>
					<section data-auto-animate>
						<img src="assets/example-wes1.jpeg" width="80%" height="80%" />
						<p>417,865 likes</p>
					</section>
					<section>
						<img src="assets/the-russo-brothers-comment-1.png" width="60%" height="60%" />
					</section>
					<section>
						<img src="assets/the-russo-brothers.png" width="50%" height="50%" />
					</section>
					<section>
						<img src="assets/the-russo-brothers-comment-2.png" width="50%" height="50%" />
					</section>
					<section>
						<img src="assets/the-russo-brothers-comment-3.png" width="80%" height="80%" />
					</section>
					<section>
						<img src="assets/press.png" />
					</section>
					<section data-state="blur-bg">
						<img src="assets/avengers-combined.png" />
						<aside class="notes">
							It was very pleasent having 5 minutes of fame but it came with some unexpected consequences
						</aside>
					</section>
					<section data-auto-animate>
					    <img src="assets/anti-ai-art.jpg" width="80%" height="80%" />
						<aside class="notes">
							Turns out people really didn't like AI art - and I started recieving abuse.
						</aside>
					</section>
                    <section data-auto-animate>
                        <img src="assets/anti-ai-art.jpg" width="60%" height="60%" />
                        <p>25 DECEMBER 2022</p>
						<aside class="notes">
							Even on christmas day
						</aside>
					</section>
                    <section>
                        <img src="assets/ip-law-art.jpeg" width="80%" height="80%" />
						<aside class="notes">
							You'll hear this sentiment from around the internet. AI is polarising people's opinions
						</aside>
                    </section>
					<section>
						<style>
							 /* Styles for the new "Arguments" grid */
        .arguments-grid {
            grid-template-columns: repeat(2, 1fr); /* Two columns for Anti-AI vs Argument */
            text-align: center; /* Align text within argument cells */
            font-size: 0.6em; /* Slightly larger text for arguments */
            line-height: 1.4;
        .arguments-grid .grid-item {
            padding: 1.5rem; /* More padding for readability */
            display: flex; /* Use flex to vertically center text if needed */
            align-items: center; /* Vertically center content */
            min-height: 80px; /* Ensure consistent height for argument boxes */
			font-weight: bold;
        /* Specific styling for Anti-AI sentiment */
        .arguments-grid .anti-ai-sentiment {
            background-color: #721212; /* Dark red for anti-AI */
            color: #fce4e4;
        /* Specific styling for the argument */
        .arguments-grid .argument-response {
            background-color: #0c4382; /* Dark green for argument */
            color: #e4fce4;
        /* Remove hover effect for argument grid items */
        .arguments-grid .grid-item:hover {
            transform: none;
            z-index: auto;
						</style>
					<div class="grid-container arguments-grid">
						<div class="grid-item anti-ai-sentiment">
							Talentless typing, not art.
						</div>
						<div class="grid-item argument-response fragment fade-in" data-fragment-index="1">
							Ideas matter, talent inclusive.
						</div>
						<div class="grid-item anti-ai-sentiment fragment fade-in" data-fragment-index="2">
							Flooding market, indistinguishable from art.
						</div>
						<div class="grid-item argument-response fragment fade-in" data-fragment-index="3">
							Market shifts: sell artist, not art.
						</div>
						<div class="grid-item anti-ai-sentiment fragment fade-in" data-fragment-index="4">
							Theft from innocent artists.
						</div>
						<div class="grid-item argument-response fragment fade-in" data-fragment-index="5">
							Focus ideas, not direct recreation.
						</div>
						<div class="grid-item anti-ai-sentiment fragment fade-in" data-fragment-index="6">
							Environmental damage.
						</div>
						<div class="grid-item argument-response fragment fade-in" data-fragment-index="7">
							Energy impact acknowledged, use responsibly.
						</div>
					</div>
                    </section>
					<section data-background="black">
                        <p>The more you succeed, the more you get resistors</p>
                    </section>
				</section>
				<section data-background-image="assets/robot-bully.png">
					<section data-state="focus-bg">
						<h1 class="standout">Business Culture</h1>
					</section>
					<section data-auto-animate>
						<div class="grid-container arguments-grid">
							<div class="grid-item anti-ai-sentiment">
								Mindless prompting, not real work
							</div>
							<div class="grid-item argument-response fragment fade-in" data-fragment-index="1">
								Share the prompts, this is a shortcut for all
							</div>
							<div class="grid-item anti-ai-sentiment fragment fade-in" data-fragment-index="2">
								They're coming for our jobs
							</div>
							<div class="grid-item argument-response fragment fade-in" data-fragment-index="3">
								Help improve your job but focusing on what matters
							</div>
							<div class="grid-item anti-ai-sentiment fragment fade-in" data-fragment-index="4">
								The output is slop
							</div>
							<div class="grid-item argument-response fragment fade-in" data-fragment-index="5">
								Break it down and improve
							</div>
							<div class="grid-item anti-ai-sentiment fragment fade-in" data-fragment-index="6">
								There's too much to learn
							</div>
							<div class="grid-item argument-response fragment fade-in" data-fragment-index="7">
								Start small and work up
							</div>
						</div>
					</section>
					<section data-auto-animate>
						<h2>This will be defunct in 6 months</h2>
						<aside class="notes">
							The only true risk to your organisation is resisting AI because there'll be a better AI in 6 months
						</aside>
					</section>
					<section data-auto-animate>
						<h2>Being Burnt</h2>
						<p class="fragment fade-up"><img src="https://felloai.com/wp-content/uploads/2025/11/Google-Nano-Banana-AI-Image-Generator-All-You-Need-To-Know.jpg"></p>
					</section>
					<section data-auto-animate>
						<p><video data-autoplay src="https://storage.googleapis.com/gweb-uniblog-publish-prod/original_videos/ImageEditingGemini_Inline_XZuiDzE.mp4" /></p>
						<aside class="notes">
							But the problem with AI models and these networks that you will build is you’ll be obsessing with creating networks and testing them and tweaking them and improving them when suddenly a company will come along and change everything.
In this scenario we were a few days before delivery when google announced...
https://felloai.com/wp-content/uploads/2025/11/Google-Nano-Banana-AI-Image-Generator-All-You-Need-To-Know.jpg
What Nano Banana did is it suddenly became able to reason like we had just hacked.
https://storage.googleapis.com/gweb-uniblog-publish-prod/original_videos/ImageEditingGemini_Inline_XZuiDzE.mp4
This is where your world comes crashing down. Everything you’ve been developing has fallen away and you have to make the disturbing choice of canning the whole development and integrating the new model, or continuing to pour resources into this development because of the sunk cost
But do you know what?
The bet strategy when it comes to AI products
Make it easy to replace any individual component in your network of tools.
That way you can easily adopt better models, and pivot onto more exciting and fun ways of doing things, and running A/B tests to understand the quality / cost balance of each of the strategies on hand to you
						</aside>
					</section>
					<section data-auto-animate>
						<h2><a href="https://gemini.google.com/app">Demo</a></h2>
						<aside class="notes">
							Fill the hottub with crystal blue water and clean all the mold of the sides
							https://www.combinedshipping.co.uk/Graphics/Dirty%20Hot%20Tub(1).jpg
						</aside>
					</section>
					<section data-background="black">
						Think small, optimise after
					</section>
				</section>
				<section data-state="conclusion" data-background-image="assets/robot-sad.png">
					<section data-state="focus-bg">
						<h1 class="standout">The Viral Burnout</h1>
						<aside class="notes">
							Now I’d love to say going viral was the best thing to ever happen to me. 
							But it wasn’t. 
							It wasn't the Anti-AI art brigade hating on me
							Nor was it the journalists who quoted me but never spoke to me
							Sadly what happened is my moment of euphoria rewired my brain chemstry. 
							It had changed my hobby from a habit into an addiction. 
							I craved the feeling of that half a million likes and made me endlessly pursue a lofty goal of making the perfect artwork again
						</aside>
					</section>
					<section data-auto-animate >
						<div class="grid-container">
							<div class="grid-item">
								<img src="https://pbs.twimg.com/media/FmT18y-XwAADkT8?format=jpg&name=large" alt="Midjourney V1">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/FmVRSelXwAIvPyX?format=jpg&name=small" alt="Midjourney V2">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/Fl-gZdaWAAMdljO?format=jpg&name=small" alt="Midjourney V3">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/FnbHKkrWIAEF9qg?format=jpg&name=small" alt="Midjourney V4">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/FoZ0g92WcAITVTU?format=jpg&name=small" alt="Midjourney V5">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/FyE7q-iWAAAk-YS?format=jpg&name=small" alt="Midjourney V5.1">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/F1x0zXtWAAIzcRc?format=jpg&name=small" alt="Midjourney V5.2">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/F3ccr-wW0AAvJnk?format=jpg&name=900x900" alt="Midjourney V6">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/F4e6ymQW8AAgiSq?format=jpg&name=small" alt="Midjourney V6.1">
							</div>
							<div class="grid-item fragment fade-up">
								<img src="https://pbs.twimg.com/media/F5mXDvOWkAAgRQj?format=jpg&name=small" alt="Midjourney V7">
							</div>
						</div>
						<aside class="notes">
							As the likes grew i witnessed a change in the art I was producing. 
							<br>Click<br>
							It seemed to me the darker the art got
							<br>Click<br>
							The more likes it received
							<br>Click<br>
							This resulted in me getting darker 
							<br>Click<br>
							which attracted a darker audience. 
							<br>Click<br>
							I had created a reinforcing feedback loop 
							<br>Click<br>
							in a nightmare landscape.
							<br>Click<br>
							When a model learns from it's own output
							<br>Click<br>
							It has a term in machine learning 
							<br>Click<br>
							called "Model Collapse"
						</aside>
					</section>
					<section data-background="black">
						<aside class="notes">
							In humans it's called burnout
							I'm happy to share I'm back to coping, but this isn't a story of my return to health
							This is a warning. To all people exploring AI. Be careful when you become a product of the algorythems.
							I wanted to share beautiful art with the world. I ended up sharing horror
						</aside>
					</section>
				</section>
				<section data-state="conclusion" data-background-image="assets/robot-future.png">
					<section data-state="focus-bg">
						<h1 class="standout">Conclusion</h1>
					</section>
					<section data-auto-animate data-state="blur-bg">
						<h2>How should you approach AI usage?</h2>
							<li class="fragment fade-up">Vision</li>
							<li class="fragment fade-up">Think small</li>
							<li class="fragment fade-up">Optimise after</li>
							<li class="fragment fade-up">Test alternatives</li>
							<li class="fragment fade-up">Things move fast</li>
							<li class="fragment fade-up">Embrace the resistors</li>
							<li class="fragment fade-up">Beware of sunk cost fallacy</li>
							<li class="fragment fade-up">Prioritise outcomes over outputs</li>
						</ul>
					</section>
					<section data-background="black">
						<img src="assets/ai-not-replace.jpg" />
						<p>by mitch0zᵍᵐ</p>
					</section>
					<section data-background="black">
						<h1>Any questions?</h1>
					</section>
				</section>
				<section data-state="conclusion" data-background-image="assets/robot-future.png">
					<section data-state="focus-bg">
						<h1 class="standout">How does AI Art work?</h1>
					</section>
					<section>
						<h2 class="standout">Image Classifiers (2012)</h2>
						<img class="fragment fade-in" width="60%" height="60%" src="https://media.licdn.com/dms/image/v2/D5622AQHv-6ydy0DaOw/feedshare-shrink_1280/feedshare-shrink_1280/0/1709302115361?e=1760572800&v=beta&t=QmOXqWqECfU06FmxD-eEgkB0Vi5lreglHXF5xW0LtA4" />
						<aside class="notes">
							So how did we get to the point where we could generate images? Well You are actually all to blame 
It’s predicted that over two trillion photos have been taken, but as of 21st March 2024 136 billion images are indexed by Google Images. Often these images have a description of what is in them. It’s the perfect example of an input and an output. Image + Description.
These are a little chaotic, but you can also “buy” datasets that contains thousands of images and each image has a high quality description that has been applied by researchers. Examples include ImageNet, Cifar-10 and Open Images.
Also a number of companies used you humans to help them identify the items within an image. Unwittingly you were training AI to take an image and be able to determine a text description of what was in the image. This is known as an Image Classifier - examples include GoogleNet / Inception, VGGNet and ResNet.
						</aside>
					</section>
					<section>
						<h2 class="standout">Deepdream (2015)</h2>
						<img  width="40%" height="40%" class="fragment fade-in" src="https://upload.wikimedia.org/wikipedia/commons/d/d0/%22Mona_Lisa%22_with_DeepDream_effect_using_VGG16_network_trained_on_ImageNet.jpg" />
						<aside class="notes">
So you’ve created an image Classifer. It can take an image and tell you what is in the image.
But then someone thought “I wonder what will happen if I play this backwards?” So researchers basically reversed the algorithm pointing it at an image. So let’s say we want a dog, and we point the algorithm to a picture of the mona lisa.
What DeepDream did was explore the picture for things that felt like patterns that matched dog, and then it would nudge the pixels around to make them closer match what it knows as dog-like patterns in it’s dataset.
What is output is very unusual!
					</section>
					<section>
						<img src="https://compote.slate.com/images/09a84b78-17de-4030-a375-1fa8bee1fe96.jpg" />
						<aside class="notes">
But the problem with this architecture is that it needed an image to apply the output to. It just amplified images
</aside>					
					<section>
						<h2 class="standout">Image GANs (2017)</h2>
						<p class="fragment fade-in"><a target="_blank" href="https://thispersondoesnotexist.com/">ThisPersonDoesNotExist.com</a></p>
						<aside class="notes">
This was a new technique to generate better quality images from scratch and it did so by creating a pair of networks. The Generator would try to make new images based on the input prompt. The Discriminator would review the image and try to judge if the image was real or fake. These would work in tandem until the images generated had a very high quality of being deemed as “real”.
A great example of this is 
I used to use this website all the time to create images for Personnas so we could empathise with a real person in our teams.
Note these people don’t exist, they are entirely AI generated. Using StyleGan
The problem with Gans is they are great at producing randomness from a specific data-set, but you can’t control them. You can’t ask for any image. You are simply building thing like the training set - so novelty isn’t possible.
</aside>					
					<section>
						<h2 class="standout">Diffusion Models (2020)</h2>
						<img class="fragment fade-in-then-out" src="https://i0.wp.com/novita-blog.s3.ap-southeast-1.amazonaws.com/a-complete-guide-to-using-nai-diffusion-0%2AvzoCGilqsoLsVAFs.gif?w=512&ssl=1" />
						<aside class="notes">
In a diffusion model the took a curious approach. They had identified that images kind of need a hook to attach themselves to like in the DeepDream model. ANd they need a way of being generated without over complexity of the background like in the ImageGens.
Researchers discovered they could represent the act of turning an image into static noise as maths.
The same maths can be used in reverse. So we can start with “noise” like static on an old TV set. Then we pass across the image applying the reverse algorithm to denoise it. The diffusion model is just trying to clean up the image and the qualities represented by the prompts start to emerge.
						</aside>
					</section>
				</section>
				<section data-state="links" data-background-image="assets/robot-admin.png">
					<section data-auto-animate data-state="focus-bg">
						<h1 class="standout">Contact Me</h1>
					</section>
					<section>
						<div class="social-links">
								<a href="https://linkedin.com/in/thedigiguru" target="_blank">
									<i class="fa-brands fa-linkedin"></i>
									<span>thedigiguru</span>
								</a>
							</p>
							<a href="https://x.com/thedigiguru" target="_blank">
								<i class="fa-brands fa-x-twitter"></i>
								<span>thedigiguru</span>
							</a>
							</p>
							<a href="https://instagram.com/digiguru" target="_blank">
								<i class="fa-brands fa-instagram"></i>
								<span>digiguru</span>
							</a>
							</p>
							<a href="https://sigmoid.social/@digiguru" target="_blank">
								<i class="fa-brands fa-mastodon"></i>
								<span>digiguru</span>
							</a>
							</p>
						</div>
					</section>
				</section>
		<script src="dist/reveal.js"></script>
		<script src="js/gpt-component.js"></script>
		<script src="plugin/notes/notes.js"></script>
		<script src="plugin/markdown/markdown.js"></script>
		<script src="plugin/highlight/highlight.js"></script>
			Reveal.initialize({
				hash: true,
				// Learn about plugins: https://revealjs.com/plugins/
				//parallaxBackgroundImage: 'assets/adamhall.jpg', 
				//parallaxBackgroundSize: "1578px 1578px",
				//parallaxBackgroundHorizontal: 0,
				plugins: [ RevealMarkdown, RevealHighlight, RevealNotes ]
			Reveal.on( 'slidechanged', event => {
				const current = event.currentSlide;
				const background = event.currentSlide.slideBackgroundContentElement.closest(".stack").querySelector(".slide-background-content")
				console.log("SLIDE", event.currentSlide);
				if(event.currentSlide.dataset.state == "focus-bg") {
					//Reveal.dispatchEvent("focus-bg")
					background.classList.add("focus");
					background.classList.remove("blur");
				} else {// if(event.currentSlide.dataset.state == "blur-bg") {
					background.classList.add("blur");
					background.classList.remove("focus");
					//Reveal.dispatchEvent("blur-bg")
				// event.previousSlide, event.currentSlide, event.indexh, event.indexv
		</script>
Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FilesExpand file tree

thames-valley-ai.html

Latest commit

History

thames-valley-ai.html

File metadata and controls
