How AI will revolutionize web governance

So today’s talk will be on how AI will revolutionize web governance. And I should probably elaborate on that to say how AI will revolutionize websites and web governance. My name again is Oliver Emberton. I’m a CEO here at Silktide. And today this talk will be covering firstly how generative AI changes everything. So if you’ve not been tracking the world of AI or maybe even catching it on the sidelines, you’ve probably heard a lot of hype, but you may be confused as to what is actually causing the underlying excitement and why this hype cycle is different from all of the others you may have experienced in the past.

Let’s talk a little bit about that. And when you understand that a lot of everything else that follows will slot into place. We’re then gonna dive deeply into creating content with AI for websites, as well as how accessibility may be revolutionized with AI. How customer support will be changed by AI. And finally, how the website optimization of user experience will be changed by AI.

There’s a lot to go through. We’re going to allow roughly 10 minutes for each of those. There should be Q&A at the end, and depending on how well this is received, I imagine that there’s much potential for further webinars on each and every one of these that catch your interest. So firstly, how does generative AI change everything? Well, what is generative AI? generative AI is a bit of a buzzword and unfortunately it can be a little confusing because strictly speaking it refers to AI that generates content and that’s a little bit misleading or narrowing.

I would encourage you not to think of it in those terms. What you really should be thinking of generative AI as is a new kind of AI that is much, much smarter. Let me show you. So historically, AI according to the news, is some kind of terrifying Terminator monster, some hyper advanced technology that is perfect and flawless and scary in every way.

But as we know, AI in real life is something of the butt of jokes. Generative AI is not like this, so the kind of tools that we use to being bad things like Alexa or Siri, which are almost embarrassingly and predictably poor, but actually understanding what you want, have been displaced by new technology that is actually extremely good in these areas.

A common cause of confusion is people mix up this kind of AI with modern AI, and I’m going to show you why that’s a mistake. Generative AI is intelligent in a way that all previous AI was not. Let’s look at a few quick examples and I apologize, some of you may be familiar with these, but I just want to level set before we get to the real details of website governance.

So firstly, the image generation. If you’ve not, well… if you’ve been on the Internet at all in the last 12 months, you can’t but help have come across images like these that are generated by AI. So this is completely generated by AI, 100% generated by machine from nothing more than the text prompt. Right. So someone will have typed in a premise like red haired female warrior in the snow and an image of this quality is generated entirely by machine, typically in 30 or 40 seconds.

That is a standard capability. In fact, this image I think is about four months old. Just to show you that that kind of thing is possible. And similarly, images like this, which are even more creative and expressive. This is Darth Vader in the style of Wes Anderson. And if you think about the level of complexity and understanding required for a machine to consider, it knows what Darth Vader looks like.

It knows what Wes Anderson’s movies look like, and it can just synthesize something like this to a near photographic quality. That is absurd. And it’s now table stakes. This is just default. This technology exists, anyone can use it. Again, this is Midjourney 5. You can just go online and try it yourself. You can run this on your phone.

So one final example here. This is an example of some cartoon anime kind of style illustration. The same technology that can generate photographic images can also generate cartoons. It can generate paintings, hand-drawn sketches, it can generate pixel art, it can do whatever you want. Right? And this is insane. Again, I want you to think like this. This looks like it’s done by a professional artist.

This is doable by machine from nothing more than text. What else can AI do? Generative AI can do things like computer vision. And again, this is an example where generative seems like the wrong word because you’re thinking, I’m looking at something. I would encourage you not to think of it as generative. I just think of it as smart.

So computer vision using generative AI allows a machine to look at something like this. We’re looking at a butterfly on a leaf and asking an AI can you write an Instagram caption for this? And sure enough, it not only understands the image, but it’s able to apply it to understanding in a surprisingly non obvious way. It takes the blue butterfly and it writes, in this case, nature’s delicate dance where dew meets wings, which captures the quality of the image, but it even manages to use a blue emoji butterfly in order to decorate that post and hashtags that would be appropriate for Instagram.

So this is just transcending what the state of the art was only less than a year ago. And again has become something you can use basically for free, now if you know where to go on the Internet. An even more impressive example here would be I’ve taken a screenshot of a web page. This is a screenshot, I believe, of Shopify’s checkout and giving it a screenshot itself and asking it to analyze it.

Well, that’s one thing. But here we’re actually gonna say, Can you write the code needed to create a web page like this? And sure enough, the AI can look at this and literally from scratch code HTML and CSS to make that page work from an image. Now this would historically be something at a human level. Now, to be clear, it’s not going to do a good enough job yet that you would not use a designer to refine the results.

It’s definitely imperfect, but it’s shockingly good considering, you know, where this came from. Less than 12 months ago, none of this was even conceivable. So lastly, I want to show you one final area. Again, generative AI. I don’t think of it as generative. Think of it as like smart. I’m going to give it some content to analyze. So what I’m going to say, and this is GPT4. I’m just going say to GPT, analyze Silktide’s recent AI announcement, summarize the tone of the announcement and any potential confusion or concern this announcement might cause. Being human I actually made a typo in there. I said confusion of concern, which proves even humans can still make mistakes. Anyway, ChatGPT went ahead and it searched the internet for that and then it looked at this page. So our web page for that topic, it read it.

It all took a few seconds and it came back and it summarized in brief what it thought was happening, right? So just think what’s happening here. It’s browsing the Internet, it’s looking at web pages. It’s performing a pretty human level comprehension of that web page, and it’s actually able to summarize what’s going on in a pretty intelligent way.

Right? So it looks at the content and the tone and so on, and it also made intelligent suggestions on what we might improve. So in this case, it says claiming to be the world’s first in AI powered web governance could set high expectations, potentially leading to disappointment if the tool doesn’t meet the groundbreaking innovation implied. Now, as much as it pains me to hear that that is actually a legitimate concern and something as CEO I would like to know about.

But here the AI from nothing more than a couple of sentences of text was able to assess something that it didn’t know about and make an intelligent analysis and give it to me with almost no effort whatsoever. All of these are examples of one new universal technology, right? So previously I’ve worked on AI for 20 odd years. I did my degree in AI back in 1997, and AI has been running for about 50 years or so professionally.

We’ve always had separate fields of AI. There was like computer vision and natural language processing. These were specific areas of interest. But what has changed is they’ve all been replaced in the last, like two years. Every single one of those areas has been superseded by a single new technology, which is what we have. In summary, it’s a little bit confusing, but we would call GenAI and GenAI is a new generalized paradigm for intelligence.

You should think of it as it’s a kind of intelligence that now that we’ve discovered this technology, we can consistently apply it to basically anything. For the same AI that was doing, for example, the image generation you saw early on, the exact same software that does that can create music. With nothing else, other than we give it music instead of images.

So the same brain works on music and images in the same way. That’s insane. And for what it’s worth, it also works on DNA. So we seem to have unlocked, like the core fundamental concepts of intelligence in a way that we can now apply to just about anything. This is… it’s impossible to overstate how incredibly transformative and important this is.

This is why you’re hearing the hype. This is why you’re seeing the explosion in demand. This is why everyone’s going on about this. It’s because there is genuinely something new here and is transforming the world in every dimension. Generative AI is redefining the fundamental level what computers can do, and as a result, it’s also going to redefine what people will be expected to do with computers.

So with that preamble out of the way, let’s dive in to some details. We’re going to start talking about aspects of websites and website governance. So we’re going to look at how this new AI technology transforms them and how that may affect you and someone who works with websites in the real world. So firstly, let’s look at creating content with AI.

So many of you probably already played with something like Bard or ChatGPT or one of these other tools to create content. I’m going to focus initially on written content, but for those of you who haven’t tried this, one of my favorite exercises to blow people’s minds is ask one of those AIs to write a poem. And again, my AI of choice is ChatGPT 4. So I asked GPT 4 to write a poem for this webinar.

So it came back with: In Austin’s heart with hope and charm, Oliver spoke with wit and calm. “AI will change the web,” he said, “in ways we’ve never seen or read.” But as he spoke with great elation, AI decided on a vacation. His words hung still, the screen froze tight, Poor Oliver winced at the sight.

He laughed and said, “See, here’s the fun. AI’s journey has just begun. It’s clever, fast, a bit sublime, But hey, it trips up from time to time!” Now it’s being a little self-deprecating there, which is actually what I asked it to do. But this took me seconds, right? I actually spent about 2 minutes total giving it a prompt. It came back some ideas I kind of played around with.

I went back and forth. The kind of back and forth collaborative creation of content like this is something that again, didn’t exist 11 months ago and it’s now becoming standard. And I’m going to show you what that looks like in a professional context with website creation. So there are a wide range of tools. This is just a smattering of them.

Jasper is definitely one of the more popular ones. I’m going to dive into Writer in a moment just to show you that, because I think it’s a good example. If you want to have fun just for your own personal enlightenment and or amusement, I can highly recommend Sudo Write. Sudo Write is a tool for generating fiction. So like novels, and it is detested by professional fiction writers because, well, lots of reasons, but it’s very entertaining.

It will help you write your first novel. If you’ve never tried that, I just recommend trying it for free. But anyway, we’re going to dive into Writer. So writer.com is one of a new generation of professional tools designed to help you write website content using AI. And all of these tools are only months old, right? Because the technology is only months old.

But I’m going to show you what this can do right now. And I want you to bear in mind that what you’re seeing here is not unusual, right? Other tools could do this to varying degrees and there will be many, many more tools that will do what you’re going to see here. In particular, Microsoft Word is integrating similar features to what I’m going to show you, and tools like Notion, for example, already have comparable functionality. So let me show you. So I wanna write a blog post, right? And the blog post I want to write I’ve decided I want to write a blog post called How to make your pet famous on Instagram. So what I do is I go into Writer and they’ve got a feature that will use AI to generate a blog post for me, or at least to get me started.

So all I really need to do is give it a blog title. I’ll do that and then it will come back to me and it will suggest an outline for my article. So this is an example of the kind of collaborative workflow you’re probably going to be shifting more and more into if you if you work with generated content.

We use it to kind of get you started, right? You say, I want to do this, and it’s like, okay, here’s a structure. Does this work for you? Does this come across what you… Does this cover what you’re aiming for? And in this case, it came back with like headings and I tweaked them, right? So I took the first heading and I changed it to finding your purrfect pets because apparently I love puns I’m a terrible person.

But I also I dragged and dropped a few around like kind of reordered it and tweaked it a bit. And so I’m telling it what I want and then I carry on to the next step and it comes back and it’s like, here’s some kind of key points I could make for those. And I’m like, Yeah, okay, that seems legit.

I didn’t even bother looking at these to be honest, I just went, Yes, fine, go ahead. And then it writes me like that, writes me about 12 words on that article just from scratch and takes about 30 seconds. So already I’ve got content that would previously have potentially taken hours to write. Now, I’m not going to claim that this is as good as something that I would have written if I was doing this professionally, but it’s pretty decent.

Certainly as a first draft, it’s not without value, but there’s more going on here. And one of the things that’s going on is this is a new generation of tools that’s empowered by AI. So if you look on the right hand side of this word processor, it’s got a whole ton of AI augmented commentary about this blog post.

So it’s highlighted sections in blue and it’s saying things like, Well, this isn’t written very well. You could probably, you know, tighten up language, make it easier to understand or the style here. Yeah, maybe it’s not that appropriate for what you’re going for. Maybe you want to consider your audience better or it’s not got inclusive language. Maybe you’re accidentally being offensive to someone in your writing and so on and so on.

These are all radical new capabilities that a computer could not do at all that are now something that you can just have appear in a sidebar and there’s a ton more besides which I’m not showing you in this screenshot, but again, I want you to assume that everything you’re seeing here is obviously possible now and is just going to be distributed across every single writing tool that you already use in the space of months, not years. Months. Like you will see this same Microsoft Word probably somewhere around January or February, March next year. It’s already… you can see the demos online if you want to Google for them. So this this is going to be standard. Every single person who has access to writing software is going to have AI that can help them write as they do it.

And I want to show a little bit of what that’s like. So I’m going to take this article and I’m going to dive into this first bit. So this is our opening paragraph, our intro, and it’s okay, right? It says, In the world of social media, it’s not just humans who can become famous. Our furry friends can too.

And I’m like, okay, all right. But as an editor, and this is what I increasingly become now, I’m not like a writer. I’m becoming more like an editor for the AI. I’m gonna look at this. I’m going to be like, all right, we select that section. I’m going to modify the tone. So I want to make that a little bit wittier, because I’m British.

So I just feel obligated to be witty at all times. I’m sorry. It’s a national curse. See that in itself is a joke. Anyway, so we’re going to modify the tone here. I’m going to try to make this a little bit wittier, and it comes back with some suggestions. It’s like, well, maybe even our furry friends can get their pawsome 15 minutes of fame. Groan.

But you can see what it’s doing, it’s trying to come up with a pun, right? And then likewise, even our furry friends can find fame if they’ve got the right pawtential. And lastly, my favorite one is Who says you need opposable thumbs to be a star? Now, these aren’t necessarily great, right? You’re probably not going to get rid of your world class professional writer and replace them with these specific suggestions.

But they’re not terrible for provoking improvements, for giving me ideas, for brainstorming. And the thing to bear in mind here is this is a computer. This is a piece of software, a normally considered, cold, calculating machine, and it’s come up with puns and jokes and ideas that are creative. And I can do that with two clicks of a mouse.

So this is the sort of thing and there are so many more examples of this that does change the way we’ll work with computers. Anyone who’s been using AI copilots in areas like programing and so on is already used to this. But in word processors, this is often catching people out by surprise. So anyway, I’m going to pick one other area.

I’ve got this part of the article and it talks about while any pet can become a star with the right personality, some breeds are naturally more photogenic and popular on social media. And I’m like, huh, that’s a good point. I want to get my SEO up here. I want to get lots of dog breeds and cat breeds and things like that in this article that would definitely help boost SEO.

So what I’m going to do is I’m gonna write one extra line below this paragraph. I’m going to type in popular breeds for Instagram, colon. And I’m like, now I would presumably research what the popular breeds are on Instagram, and then I would figure that list and I would type them into bullet points. But because this is AI, I can just push one button and it does it for me.

So it goes ahead and it figures out the popular breeds on Instagram and it writes them out there and I can edit that list. I can make it longer or shorter, whatever I want, but it just took care of a whole problem that otherwise would have been manual and did it all for me in literally… I just press command and enter and that was it.

One key press. Everything you’ve seen here is just one level of abstraction. This is working on the content yourself like an editor, but with AI, I would encourage you to also think about it a level above and the level above is that if I can do it as a human, AI could also control the AI, right? So what I’ve got here as an example is a workflow where this is, funnily enough, this is for a webinar where the idea is this workflow takes as an input someone’s webinar delivery.

So say a video recording of my talk right now. And then it uses AI to summarize the transcript of that webinar. Then it writes Thank you emails to the participants of the webinar, which is based upon the content of the webinar and its own analysis of it. And then it writes a follow up email to say, you know, how you’re doing and maybe like engage with them in an area like, say, asking them a survey question, what would they like to hear about in future webinars?

This is an actual workflow that you could build with the same tool I just showed you. This is Writer, so you can not just use writing tools to create content. You can use writing tools like automate the flow of content and actually implement business processes. And again, this is not science fiction. This is stuff that exists right now and there are multiple competing products that do it.

So a lot of people don’t realize it because it’s changing so fast. But this is the world we already live in. It’s just not evenly distributed yet. So a few quick predictions for content. So content generation of websites, it’s going to be a few consequences of all of this. So the first one is that standards will rise and the standards will fall.

So what I mean by that, with the benefit of modern writing technology, I will be able to write faster, right? I can use AI to help take care of some of the grunt work for me. AI also may provide intelligent suggestions that allow me to come up with ideas I wouldn’t have done already. And there’s so many versions of the tools I don’t have time to show you right now.

But like the fiction writing tool I mentioned will help me brainstorm story ideas that will help me brainstorm descriptions and characters and so on. So this allows people potentially to create better work, but simultaneously it allows a ton of people who don’t currently write at all to start writing. And potentially that will result in a parallel reduction in quality where, you know, it’s like when YouTube came out and people could make videos, anyone could put a video online.

So you get lots of terrible, terrible content because people can just put it out there. But you got, additionally, you got some new brilliant content because some people work crazy hard in their bedrooms and can create like a high quality movie or something in their basement. When you democratize technology like this, you’re going to just get an explosion of the range.

So there’ll be both at the same time. Next up, you’re going to see mass personalization, things like what I showed earlier with that workflow. Why not send everyone who attends a webinar, a custom email based on what you know about them on LinkedIn and your webinar and have AI do all the work for you. That sort of future is going to become standard probably in a couple of years.

Discoverability. So things like SEO and social media are largely played out games, so most marketing professionals are used to the way SEO works, right now, the way social media works right now, you produce high quality content, you’re in competition with your peers and you hope to do better than they do, and as a result you receive a set number of visitors.

But that whole game is about to be reinvented. Everything about it is going to change. We don’t quite know how, but it’s likely that creating 100 or 1000 or 10,000 times more content for whatever ends up being is going to force search engines and social media platforms to start looking for AI maybe filtering it or maybe judging it more harshly.

We genuinely don’t know, but it’s certainly going to change that game. And lastly, roles are going to shift. If you’re somebody who works on content, you’re probably going to find that you spend less time writing words and more time directing or being like the creative editor or, you know, the leader in this field. So let’s move on. Next up, improving accessibility with AI.

So web accessibility, as you probably know, is a deep passion of mine, and of Silktide’s. I’m gonna very quickly look at an example of what current assistive technology looks like in this case for someone who is either blind or partially sighted. So that’s to say, let’s say I don’t… I’m using a website, I’m using a screen reader, which is technology that reads out the screen I’m using, and it attempts to allow me to interact with the website without requiring sight.

Now, this technology, for anyone who’s used it is terrible, to be clear. It’s awful. It’s just there’s no good way of describing it. It’s just genuinely an awful, awful experience from start to finish. And I would compare it to a bad phone menu. So what I mean is if you ever called up like a bank or someone and have an experience where the bank is going, thank you for calling our bank, please listen to the following options.

Press one and so on. And you want to kill yourself, right? Because it’s just terrible. Because you have to listen to it and then you have to wait for it to do things. And you have to, like, keep listening to more or more options to get to what you want. And interacting with it. It’s just awful. Now, assistive technology here is very similar, so let me show you.

So if I was looking at a website like a news website, it might read out something to me like BBC News, Home page, and I’m like, Okay, fine. That’s the title. Great. Next. Skip to content link. Okay. Finally, I kind of want to do that, so I’ll click that. Great. Thanks. And I’m waiting and I’m like, okay, I get to the next bit and the next bit is headlines heading level one.

I’m like, Okay, great. I don’t care, get me to the next bit. US and Russia hold talks and Ukraine crisis link. Okay yeah that sounds interesting Let’s click that. Great. You see this is like torture and this is a massively accelerated version and now like click the link, it’s going to go to a new page. And so the new page is going to start and read out the title.

BBC News US and Russia hold talks on Ukrainian crisis. I then go to the next step, but it’s going to be like, skip to content link, and I’m repeating the whole process all over again. This is unfortunately, this is a highly utopian version of what should happen. The reality is 5, 10 times worse than this. Every time you use a page, if you’re using a screen reader, it’s like, seriously, it’s unimaginable difficulty and it’s awful.

It’s one of reasons accessibility so important. But I want you to consider how generative AI would allow us to transform this experience just for technology we already have today. So let me show you. Consider instead of a unhelpful phone call with a robot menu, if you could just have a conversation with a human being on the phone, because that’s actually basically what we could do now.

So instead, the the AI might say, here are the top ten news headlines. US and Russia hold talks in Ukraine and Omicron, UK records, whatever, and then you don’t even let it finish because you’re gonna go, Hold on. Stop. Tell me about Ukraine. You’re not going to actually click on something. You’re not going to need a button.

You’re not going to need to say, press option one or read out the text exactly as it written. All these other limitations that current technology have. You’ll literally be able to get a natural explanation like a human would, and it’ll be like, Okay. Answer to that question. Right. Notice it didn’t say I’m going to new page. Loading the page, here’s the title of the page, Skip to content link… it doesn’t do any of that.

It just goes straight to the thing that it knows you want. It can do that for you. You can again say stop and ask another question. This kind of paradigm, I believe, is going to completely displace modern assistive technology because it is so obviously massively better and inevitable now that we have unleashed this technology. This seems like it’s a matter of, well, probably years, but a few years before this is going to transcend what we take for granted.

Now, I want to give you a very quick example, because people don’t know this, how good modern AI voice synthesis is. So this is both speaking and understanding natural text. I’m just going to give you a quick like ten second clip, but I want you to listen to this. This is standard, this is on your phone right now.

Things like Siri way behind the curve. This is ChatGPT and what it can do with voice.

(Teardown) Hey, my name’s Tech Teardown. I run a YouTube channel about emerging technology. How would you introduce yourself?

(AI) Hey, Tech Teardown. That sounds like a super interesting channel. If I were to introduce myself, I’d say hi. I’m Assistant, a virtual assistant here to help answer questions and provide information on a wide range of topics. So what emerging technology are you exploring these days on your channel?

(Oliver) Here’s an example of another classic accessibility challenge. We have here an image of coronavirus cases, right? This is a bar chart- well, it’s a line chart, and this is a common accessibility challenge. And the way we solve these challenges at the moment is we create alternative text and long descriptions for images like this.

And this is a real challenge. You know, expert professionals could work on this problem and it’s genuinely hard. How do you explain an image like this? Well, you can ask an AI and AI will do actually, modern AI, do a relatively good job of this, right. It will try its best to explain the nuance. And you can see on the right hand side, read in your own time, it’s a pretty reasonable explanation.

But I would actually argue that in future that we’re actually going to transcend this as a model. So this may take a while to take effect. But consider for a moment, you don’t need to read out a long description of that chart. What you want to do is you want to let people talk to the chart. So imagine people asking a question naturally, like when were cases increasing most quickly and then the AI can just be, well, the graph shows the steepest incline in March 2020 and February 2021, indicating that’s when cases were increasing most rapidly.

That is literally an output from an AI that you can use today. It’s not integrated into assistive tech, but I think that much is inevitable. So I would expect that over the coming years we’re going to see things like screen readers shift from their current paradigms into models where this sort of thing becomes possible, and that is going to reinvent just about everything. Here’s another example.

This is Be My Eyes. This is a mobile app that can literally look at whatever’s on your camera and you can ask questions. Has the milk expired? No, you have two more days. Or which one is tomatoes? The one on your right. So these are insane capabilities and this is standard. You can literally download this app and use it right now.

So again, generative AI is intelligent in a way that all previous AI was not. So many of the things that you assume AI could not do is changing, and a lot of this has enormous impact for accessibility. So in predictions, I will say that the assistive technology we see that I’m talking about now will eventually go universal, which means it won’t just be used by people with disabilities, it will be used by everyone.

If we could build the tech that you just saw, everyone would use it. I would be driving my car and I would just ask my phone or my car or whatever it is to find out the news or to, you know, help me order some food or whatever it is. And that would be potentially the assistive technology that lets you browse the web.

It could be the same tech. It potentially transcends the model of screen readers and so on entirely. Now, there’s an analog to this. You may not know it, but remote controls, they used to be assistive technology, right? The remote control for a TV was designed to help people with mobility issues. And then it turned out it was so popular that they were like, we should let everyone have remote controls.

I think you’re going to see the same thing here. And the result of that will be that assistive technology becomes incredibly good because when everyone uses it, there’s enormous commercial pressure to make it excellent. So that’d be a wonderful thing. Again standards will rise and fall. The standards of average accessibility will probably go through the roof, but then you’ll also have people who probably stop caring about it or think that they delegated to AI, which is almost certainly a mistake.

So there’ll be some nuance there and then roles will shift again. What it means to work in accessibility, what it means to evaluate accessibility is very likely to change in ways that, quite frankly are almost impossible to predict at this stage. But it’s going to be quite dramatic.

So customer support with AI. So there are a number of companies already tackling this problem.

Many of these companies are, of course, infamous in the customer support world. Honestly, there’s more examples than I could possibly do justice to here. But I want to give you a few quick highlights of how AI is transforming customer support, particularly within the concept of web governance and websites. So firstly, we’ve always come across chatbots like this, right?

So we’re probably familiar with having a little chat window in the bottom right corner of a screen and it asks you questions and then potentially talk to a person or you talk to an AI. We’ve also probably had a very bad experience with this because most of the AI that we’ve experienced in the past is terrible. But again, generative AI is not. Generative AI has new capabilities and is vastly superior at tackling these kinds of problems.

So some of the things that are now basically standard you can feed in your own knowledge base of help and so on, and you can just feed it in as text or PDFs or video and these AIs can actually understand that and actually provide high quality explanations of the issues. Literally from the knowledge you have taught it in those articles, which is crazy.

Likewise, they can also learn from previous conversations and best case studies and URLs from your website. And all sorts, so this is a whole new set of capabilities. Simultaneously, the classic handover problem, which is where it’s not clear where you want an AI for one thing and a person for another can be largely solved now. So there are situations where there’s a business process and you need a person to get involved and then the AI can just seamlessly hand you over. And things like translation.

Historically, we have always written software and then had to translate it afterwards. But one of the really crazy things with modern AI is it’s pretty much able to speak any language out the gate automatically. And just as an aside, it’s worth understanding because this is the kind of thing that sets my mind ablaze at least, if you don’t know, the way that these kind of technologies behind the scenes work, because it’s very different from old AI, large language models like ChatGPT are basically taught without lessons.

They are taught by feeding them the internet. Their education is just, read the Internet, and once you’ve read enough of the Internet, it turns out you will just learn everything. You’ll learn history and geography and poetry and marketing and law and whatever. And one of the things that it turns out you learn just automatically is how to speak every major language and how to translate between them, which is slightly terrifying.

So all the AI advances you’ve seen in recent years, or recent year really, are literally an emergent property of just feeding an AI the Internet, and watching it acquire these incredible skills all by itself, which is wild. But one of the side effects of that is, yes, your AI chatbot can now speak, say, fluent Spanish without you having to do any extra work.

AI also allows us to look at things like transcripts or video feeds. So this is looking at a customer meeting, for example. But in addition to just looking at what was said, you can use AI to actually introspect on that information so you can actually ask your AI questions like, what does the CRO care most about when evaluating new software?

And here the analysis is not just searching for some text like we associate with a computer. It’s actually making intelligent inference is based off of what they said, maybe what was implied, the context of the situation and so on. Now, I want to show you a quick example of how Silktide ourselves are actually taking these paradigms and applying them to our software.

So many of you may have seen on our website we recently announced new AI capabilities. I’m going to show you a two minute video of how synthesized AI helps you solve problems with your website’s quality, and it is using much of the same technology. And this is available this week.

(Silktide AI promo) If you’re using automated tools to test your website, you’re probably used to something like this. You can see a problem. You can see an explanation. But… now what? You’re left deciphering documentation and submitting support tickets. If only there were some way you could click a button and… Oh wait. There is. Introducing Silktide AI. With a literal click of a button, Silktide’s AI will talk you through any problem we find on your website and how to fix it. I’ll just do that, and… wow. Fixed. Now what about this “Click here” text. That’s not good. Silktide tells you why, but it’s also smart enough to suggest a replacement too. This isn’t just some static help text. Silktide AI is intelligently looking at your specific problems and talking you through what you need to understand and fix them.

Let’s say I still need help. I can ask a follow up question like Can I make this accessible without making the link any longer? Would you prefer a more or less technical answer? You get to choose. What else can I do? Quite a lot. Link doesn’t work, let AI talk you through it. Sentence too hard to read? Here’s a better one.

Need help adding a skip to content link that matches your brand colors? We’ve got you covered. With Silktide AI, you spend less time deciphering problems and more time getting things done. Experience a smarter, more efficient way to enhance your website with Silktide AI.

(Oliver) This Silktide AI feature is available right now this week to all Silktide customers. So everyone who’s already a customer will be receiving an email from us with instructions on how to access that capability. It’s more than I can go into on this demonstration, but obviously you can check it out and we’d actually encourage you to do so.

We’re very excited about what that means for us and our customers in the future. Right. So moving into predictions, predictions for customer support. Well, firstly, AI support is going to become the default. We expect AI support to become the standard channel that everyone uses in almost all situations. Eventually. And the reason for that is because it’s just so much better on several dimensions.

Firstly, obviously it could be available 24/7 and it can be immediate. So there’s no queue, there’s no wait for a person. There’s no limitation on supply, but previously the kind of chatbots we’ve had in earlier AI were very, very bad. They were incapable of solving these kind of challenges, but modern AI is capable of solving much, much greater challenges than ever before.

Not quite to the level of, you know, like a Ph.D. medical student or whatever. But who knows, maybe one day will. It’s certainly advancing at a tremendous pace. And what it’s already capable of doing is, you know, typically comparable with a lot of people in a support role. So we expect the majority of front line support will move to AI in the near future.

Again, standards will rise and they will fall. So the standard of support when you can get immediate access to a smart AI agent overnight, we expect the average support standard to go up. However, the AI is not magical. The AI can only answer things that it’s been taught about. And so if someone just gets an AI chatbot and puts it on a website and it doesn’t know anything, well actually the standard of their support probably went down because at least a human being would have interrupted you and asked their boss or something.

But here the AI just sat there giving terrible answers. So there’s still a need for people to get involved. And that kind of brings us to our last point. Again, roles will shift, so we expect that people who are working on support will probably transition more into people who teach A.I. supports and maybe provide a higher level of support where the air is incapable of doing something.

Or you want a human in the loop, or generally more strategic to look at things like, Well, what about a tone of voice? Or what about getting ahead of the problems that people are experiencing and addressing those before they even become problems? So again a big shift in roles, but general rising on the level of a role, I would say on average.

Then lastly, for today, we’re gonna look at optimizing user experience with AI. So user experience. Well, again, there’s an absolute ton of players in this space. We’re not going to able to give them anything like any justice. But to give you a brief taste, some new capabilities you may not be tracking. One is I can take a hand-drawn sketch. This is a hand drawn sketch of a mobile phone design.

It’s not a very good sketch, but that one on the left and I can import it. And modern design software will just turn it into a design. Now, the one that you see on the right is generated by a machine. It’s not very good design. But then I didn’t really give it much to work with, but is certainly better than nothing.

And I can start from that. And then I could be editing what it’s done. And again, this is something that couldn’t be done maybe now a handful of months ago. Also, and even more impressively for me, you can go into design software and you can actually give it a brief. So in this case I’m going to describe in English, I’m going to say I want a plant shop for people who forget to water the plants and I’m going to give a design style of green and minimalist.

I’m going to push a button and it’s going to design the whole damn thing for me in like a minute. So it just went ahead. It did like a welcome screen, a log in screen, discoveries, green e-commerce, you can go into this plant list, product details, cart… Now are all these designed perfect, absolutely not. You can see some very obvious layout issues here and it’s undoubtedly not got the full requirements of the project because I only gave it two sentences to work with.

But it did do this in a minute and I can refine it. I can go back in there and give it more information. I can say actually move button to the left, etc., etc. This is software that you can use today, by the way. This is Uizard, I don’t know how to pronounce it. It’s like wizard, but UI z a r d and there are many tools like this and similar capabilities coming to other tools. I’ll show you a few more quickly. This is Figma, which is like the most popular design software, and they are adding AI capabilities like this. So say you’re working on this design and you want albums at the bottom of the page. Draw out an area where you want some of your albums to appear and then get AI to do it for you because it’s kind of guessed what you want.

So it’s like, yeah, you want a grid of albums, that’s probably what you want. And then you can be like, Yeah, that’s nice, but I’d like a different album. Could you suggest another album here and then… go back down. There you go. It’s come up with some new albums so you can play around and be like, Yeah, I like that one.

Thank you very much. So this is this is how I can just assist the way you work. AI can also help predict the outcomes of your designs as you’re working on them. So this on the left is a sample web page design, and on the right is the heat map of that web design. But the heatmap is entirely fictional. AI has predicted how visitors will look at this design and is giving you a heat map before you’ve even made the design available to real people.

And it turns out AI could do this about 96% accuracy, which is kind of wild, but people are very predictable. So for example, in this case it’s predicted the headline will dominate their attention, but also say that woman’s face in the bottom left. Human psychology is, yeah, like I say, surprisingly predictable. And you could use this to inform your designs as you’re working on them, rather than having to build them and test them in a traditional way.

Now, again, I want to show you a little bit of a sneak peek. This is something that we are adding, but it already exists in Silktide, but we we are testing before we release it to the world. This is a feature we’ve wanted for a long time. So let’s say we’re looking at a web page design. This is a contact form and this is obviously kind of what you might call a a call to action page, a page where you really want to optimize the design for effectiveness because it has a real effect on your organization.

So with Silktide’s AI you can take a page like this, press a button and get an immediate summary of what the AI believes your goals are for this page and specific user experience improvements you could make to improve the conversion of this page. So that’s insane. But just to show you some specifics here, it has determined that the goal for this page is to encourage visitors to request a demo of our products and to showcase credibility and so on.

And it’s made specific suggestions which go on beyond what I’m showing you here. But one of them is to make a more prominent headline, that says something like Request your free demo now, which we don’t actually do at the moment. And another is to clarify the call to action. So the request to demo button could be something like get your free accessibility scan now and then we should potentially consider reducing the number of form fields and so on.

Now, it’s not hard to see how this can add tremendous value to your overall website improvement and how it’s literally as simple as pressing a button inside Silktide to get this kind of feedback. Also, and I can’t show it here, you could type into that box so you could ask questions about what its feedback was and it could potentially clarify them or you could debate certain points with the AI and say, Well, actually there’s a good reason I’ve not done that.

Why don’t we do something else? So kind of amazing capabilities that’s in testing right now. We expect to have that available to our customers at the start of next year.

Generative AI, you’ve seen doubtlessly examples like this. This is just Photoshop, but there are many others like it where you can generate images simply through text prompts. But what people often miss is much more sophisticated abilities.

And I want to show you a couple from Adobe where you can use creative tools that already exist to control this far more. So people are used to now, well, a lot of people are used to, just typing in a box “I want a woman emerging from an explosion of marshmallows” and getting a picture like this. But a lot of the time you get a picture like this and you’re like, Yeah, it’s not exactly what I want.

And it might feel like you haven’t got the control. But what I want to show you is actually you do. So this is this is a bit of a sizzle reel from Adobe. Most of what you’re going to see in here already exists in Photoshop. I just want to give you a feel.

So changing the scene into winter, yeah, this is standard. You can do this now, you can just take a photo of something and just completely reinterpret it and even animate it. You could look at an image and ask to create artistic brushes from the image and then use it to repaint it. You can add textures… like this is literally simulating metal and other reflective surfaces and building 3D objects from 2D objects.

That’s insane. You could do crude sketches and then ask AI to interpret them in a way that they become more realistic aspects of a painting or a photo. Creating templates for mood boards. I mean, you get the idea right? This is crazy. These capabilities that exist today, right? This is not fiction. I’ve actually used this one myself quite a few times.

Take part of an image and say, I want versions of that and just tweak individual parts of the image at your discretion. It goes on, and I’ll just let this finish. But here you’re creating a font and then giving it a fur texture and turning it into a logo. Here you’re taking a photo of your headphones and then it’s putting your photo of headphones in the other photos and combining them like they were one photo in the first place. It’s absolutely wild.

So yeah, let’s move on from there.

So some predictions for user experience. First, all design is going to become augmented. You simply won’t have a choice because every single design product is already gaining this capability. So there’s no way you won’t be able to do this if you’re doing the UX and it’ll become standard because it’s just too powerful.

It’s just too much value that’s added. Again, standards will rise and fall because some people will this to improve the quality of their design, and some people who can’t currently do user experience design will be able to because AI will do the work for them. And lastly, again, roles will shift so people will find that what they spend their time doing will probably shift up.

So it’ll move away from the day to day minutia of, say, moving boxes around on a screen and design software, and they’ll move up to more of a creative director type role. So summarizing because we’re down to the last 8 minutes or so now. Generative AI is intelligent in a way that all previous AI was not. I hope I’ve demonstrated that to you clearly.

Now. I know there’s a lot of stuff in the news about how AI is unreliable or can hallucinate. It can have various imperfections, and those things are true, but they don’t take away from the absolutely absurd value of what they are now capable of doing. And the rate of change in that capability, where basically everything I just showed you was impossible 12 months ago and is now integrated into products used by millions or in some cases billions of users.

And that is only going to accelerate. So there is no way to overstate how big of a deal this is going to be. You want to be tracking it. There are some universal predictions. You’ll notice that the four areas I looked at, they all kind of repeated some pattern and that’s not a coincidence. One of the things you’re going to find is the AI augmentation I think is going to become the default.

So with user experience design, you’re going to use AI to help you do it. With content generation, you’re going to use AI to help you do it. With accessibility, I think accessibility will be aided by AI. That will help you make things accessible, but it also accelerates what assistive tech can for you as well. And so on, so you’re going to see that pattern across every aspect of every piece of software.

You’re going to see again that standards will rise and standards will fall across every dimension. AI will enable people to do better work, but it will also enable people who can’t currently do something to start doing it for the first time. And so you’re just going to have this broad thing, this democratization of access that’s going to create a much greater range of expression that we’re currently used to.

And lastly, you’re going to see roles shift. Every single person who works in basically any job I think is going to be encountering AI over the next two or three years. I mean, I’d probably say two, but we’ll see. And as a general rule, I think it helps to think of that shift as everyone becoming like a director.

So when you make a movie at the moment, the most valuable person, or at least by some definition the most valuable person in the movie is the director. Right? But the director is interesting because they don’t write the movie. They don’t star in the movie, they don’t film the movie, they don’t do the audio for the movie. They don’t write the music.

Or what does the director do? Even the director just sits there and tells everyone else what to do. And I think that’s what most jobs are going to become. They’re going to become us telling AI what to do. It’s a fundamental reinvention of the nature of work, and it is going to challenge us in many ways. I think some people will find it brilliant and enormously empowering because they’ll have a lot less busy work and nonsense to deal with.

But then it’s probably also going to demand more from us in terms of total output. So it’s something to be wary of, mindful of. I will leave you with this quote that AI will not replace you but a person using. I will. So I encourage you to keep your eye on this. So that’s everything I have for you today, other than the questions and answers.

So I’m going to open it up. Just to remind you all, my name is Oliver Emberton. You can follow me on LinkedIn, and if you haven’t already, I encourage you to check out our AI capabilities at silktide.com/ai.

So, Jess, I don’t know if you’re able to unmute yourself. This is a good question. (Jess) Hello. (Oliver) Hello. She is, excellent. Hi Jess.

So we’ve got a few questions here. Can you recommend any AI tools that can automate alt text generation for images, even better tools that can batch process alt text for multiple images and an image library? Off the top of my head, the only company I know that does that is or in that space is Scribely, who I can recommend. I don’t know exactly the details of how they work, but you can check them out. Scribely S C R I B E L Y. It’s what you think it is. Scribely. Apparently my brain is frozen up on spelling, but yeah, they’ll probably help you out.

We have another question. What program uses predictive AI heatmaps? So I’ve actually seen 2. Uizard have that capability built in, and then it’s in one of the slides I’d have to check. I’d go all the way back for this one. Give me a second. But where is it here? Attention Insight. There you go, Attention Insight has that capability.

What is that Figma AI feature called, is it a plugin? So that’s actually from Figma’s recent blog post. If you Google for Figma AI, I think you’ll see a blog post where they announce their plans for AI. They made an acquisition of a company that provides that capability. They were making a lot of AI plugins and they basically said they’re integrating that technology into the Figma product core.

But yes, you can use a plug in for at the moment and then they’re going to make it a core part of Figma in the near future. Let’s have a look. With the rapid rate of progress of generative AI right now, is it wise to sign into large contracts with gen AI-based platforms for marketing? Ha. Not quite sure I understand the question.

I could interpret it in a couple of ways. So I guess if the question is asking is it wise to kind of lock yourself down to something right now because everything’s changing so fast? Yeah, probably not. I don’t think you need to though, most of the you’re seeing right now is very much early access, you know. So I mean, even just to take Silktide as an example, we’re providing access for our customers for free to a certain level.

So we’re going to let everyone have a certain amount of credits to use our AI capabilities. And we’re actually testing to see how that works in practice because this is so new. This is like completely uncharted territory. We don’t actually know how how widely adopted certain features will be or what the challenges might, you know, the second order consequences might be.

So there’s a lot of like testing the waters going on with all these companies. I don’t think anyone should be just going, right. We know this is going to be the one winner, we should choose this tool. This all is just too early. So I think right now I would encourage you to explore broadly and move around a lot between these products and see what makes sense for you.

Right. Next question. If everyone becomes directors, what will happen to the roles of the directors as we currently know them? Well, that’s true, but then, isn’t everything directors all the way down? I mean don’t we basically have like world leaders and politicians or whatever as well as… like there’s always a hierarchy, right? There’s maybe there’s CEO, but a CEO has leadership roles within them and then they have leadership roles within them.

And so on, then they go all the way down and eventually get people working on specific tasks. I think what will just happen is we’ll just move everyone up one step on the chain but we’ll increase the range of intelligence of their capabilities. So it’s a little bit like, have you ever had I’m sure you will have the experience of working with someone who’s really good at taking initiative, right?

So, you know, a difference between… you’ve got someone and you say, can you do this task? And they maybe do it, but they have to keep fixing it and correcting it and then moving into a state where you give someone a task and they just own it and it’s brilliant. So I think it’s going to transition more people from the first state to the second, but then it’s going to create challenges for people because not everyone’s necessarily going to adapt to that new reality. I suppose it’s going to be very interesting.

I’m not going to lie. It’s probably going to be the most dramatic cultural, societal change we will ever live through. Everyone on this call is probably… This is going to be the biggest thing in your life. So yeah, it’s crazy. Right. Next question. My mind is officially blown. I like to hear that. I’m looking forward to playing around with Silktide AI’s capabilities.

Thank you very much. You’re very welcome, Pauline. Right. Do you think there will be any impact to AT users that have old hardware and technology to use forward? Absolutely. So there will be real challenges here because a lot of assistive technology is hardware driven and that doesn’t advance at the same rate. Obviously. The big advance I expect in assistive tech, and this has been on the wall for a while, is that big tech is going to own it.

So right now there’s screen readers like Jaws, for example. I think they’re going to find it hard to compete because I think the screen meters or similar tech that’s going to come out is going to be like Microsoft, Google, Apple, the people who own the platforms, they’ve already got assistive tech in their platforms and they’re going to make that tech so good and so integrated into AI.

I suspect. I don’t know. I suspect that that will end up displacing a lot of traditional, you know, old school screen readers and assistive technology providers. But you still need hardware. You still need switches and so on. You need that. And it’s going to be interesting to see what new players emerge or how old players adapt. Honestly, I don’t know exactly how it’s going to work out, but it’s going to be a very interesting time.

It’s going to make a huge difference. The amazing thing, though, is it should transform accessibility. It should make accessibility… Yeah, I would estimate like ten times better, just absurdly better. Question: which AI is Silktide using to power its AI features? We predominantly use OpenAI. So it’s OpenAI for everything you’re seeing here.

But we have other things that we explore. But at the moment, everything you’re seeing is OpenAI. Would AI resolve problems with accessibility overlays? So interesting question. Nice spicy question, by the way, what a way to put me on the spot. At the moment, it doesn’t. Certainly. We are quite known to be not fans of accessibility overlays. I actually think the concept of overlays as they exist is misaligned with what AI could do to improve them.

And that the actual solution… The biggest problem I’ve got with overlays is that they’re like attached to the website level. If they were attached, say, at the operating system level, then they might actually make sense. So again, this is an area where big tech maybe can get involved and fix things. But certainly at the moment, anything we’ve seen out of any kind of overlay provider is not going to solve those problems.

Let’s have a look… Other thought leaders to follow? Follow Ethan Mollick, Ethan, M O L L I C K if you want one follow. There’s many. But he’s brilliant. He’s absolutely brilliant. And yeah, literally, I’m going to end my list there. What else? Can Silktide help you find all the examples of a company department name on the company’s website.

This is where the name has had to change, we need to find the existing examples of it. This may take a human a long time to do. Let’s just say that’s in R&D right now. But yes, we are working on something along those lines. Can slides be shared? Absolutely. We will do a version of this, probably slightly edited because of the technical challenges.

We’ll do a version of this that we will upload for people to share afterwards, may take a few days to prepare. Are you looking at using AI to provide user level specific training and support? Quite possibly, but slightly outside the scope of what I’m willing to admit to just yet. But yes, we’re doing a lot of active R&D in this space.

Do you think AI will fundamentally change the way people navigate and search sites? Absolutely. Yeah. I think the fundamental paradigm for computing will change. It’s a long discussion, but the really simple version is, I think we’re going to move away from single channel, keyboard, mouse, etc. interaction that we’re used to into multi-channel that’s going to include at a minimum speech as well as other interaction paradigms.

And what that means is imagine using your computer where you can talk to it and use it at the same time. Or maybe you’re in an AI environment, whatever it is, all sorts of actions are there, but computer software is not written like this right now. We have concepts like focus and so on that wouldn’t make sense. It’s a very challenging place.

It’s honestly an entire talk in itself. But I think it’s going to be things like this reinvented, and I think it’s going to fundamentally change the way that this tech works. Now, I just realized we’re 4 minutes over, I’m going to do a couple of questions quickly.

Can’t see a follow option on your LinkedIn, Oliver? Well, if you can’t see a follow option, then feel free to add me. I imagine I’ll be accepting a bunch of requests. Okay. Just wrapping up. Lastly, how much this presentation was generated by AI? A non-zero amount, maybe 10% or something like that. Yeah, they helped me. And lastly, I’ll just take David’s question. Is anyone reviewing the ethics of generative AI?

It seems like wow with what’s possible, without scrutiny of quality and consequence. If an AI tool generates a statement that is untrue that’ll massively benefit me, should I use it? Right. That’s a topic in itself. There are a lot of people looking at this. Whether or not they’re doing a good enough job is more debatable. But you should understand that companies like OpenAI that parallel this tech, put a tremendous amount of effort into doing things like preventing you asking questions or getting answers that would be seen to be harmful to society.

Do they universally succeed in that mission? No. But they are very actively engaged in it. And yes, I see I’m reasonably optimistic about this, but not everyone would agree. Honestly, that should probably be a webinar itself. So with that, I’m going to wrap up. Thank you guys for listening. We will send you some follow up on this. Like I said, there will be a recorded video that you can share with colleagues, and we intend to do many more webinars like this in the future.

So if you found this one helpful, please do let us know. We really appreciate the feedback. Thank you very much.