Artificial Intelligence (AI) feels like one of those sci-fi film concepts of “the future”… but the reality is that the future is now, and AI’s impact is being felt not just in the business and computing worlds, but also in the ways that everyday consumers can use smarter machines in their everyday life.
It’s vital to point out that what a lot of people think of when they consider “AI” are machines that think like humans do. We’re not there… yet, though there’s a lot of research and promising tools that can give a human-like appearance to artificial intelligence. For now, what we’ve got that’s largely facing consumers are instead computers that examine billions of examples of writing, art, photos and other human endeavours to provide rapid fire synthesised content built upon what it can find.
Is that actual human creativity? While there’s no doubt that everyone creating something builds on what came before, that’s probably a question best left to the philosophers. In terms of what AI can do for you as an everyday consumer, there’s a number of approaches that you can already use – and at least one you’re probably already using on a daily basis without being aware of it.
ChatGPT and Bard: Can they really write essays and emails for me?
You’ve probably heard of ChatGPT, an open AI model designed for conversational text generation. It’s the Star Trek computer, or 2001’s HAL (though hopefully more benevolent), able to answer queries and write text for you based on a simple prompt.
You can indeed use ChatGPT to write documents, whether they’re emails or essays, and it will produce mostly-reasonable responses. I say mostly reasonable because it very much does depend on your inputs and the length you ask ChatGPT to go to.
At a crude level, you could think of ChatGPT essentially using a statistical analysis of all the words it’s ever ingested (and some of the feedback driven into it based on what it spits out) to perform an effective “best guess” as to what the next word in a sentence might be. Early versions of the ChatGPT engine were quite bad at this, and while it’s improved in newer versions, it’s still quite prone to creating references out of thin air, as well as relying on what it reads absolutely online as being factual.
You can probably see the issues there, which is why it’s not a good idea to use it for that university essay or business proposal, even though it’s getting smarter and better over time. The same is true of Bard(https://bard.google.com/?hl=en), Google’s rather similar natural language chatbot; while it can rest a little more on Google’s search smarts, it’s still got its distinct limitations.
There’s also a lot of controversy around the ways that these models have been trained, especially around the potential use of copyrighted works and what that means for the rights of authors and creators. It’s equally a little murky – to put it simply, and I’ll point out I’m not a lawyer here – around whether the copyright on an AI-built work exists at all, though in some countries the line has been rather more clearly drawn, with AI creations lacking any kind of protection.
Artistic Endeavours: Can a computer paint the Mona Lisa?
The other big area where you can experiment with AI is in image creation and modification. These types of AI operate in a similar fashion to the natural language models of ChatGPT and their ilk, taking in a simple prompt and creating an image for you based on what you asked for. It’s like Photoshop, but the computer’s doing it all for you, right?
Not quite – and for what it’s worth, Photoshop has its fair share of AI features through an Adobe platform called Firefly – but the results you get can be quite interesting. Again, there is considerable controversy around style and image rights with these programs, and they’re not always the best artists when it comes to more photorealistic work, especially around faces, fingers and the finer parts of human anatomy.
Probably the best-known generative AI art program is DALL-E (yes, the pun is deliberate), from the same OpenAI group behind ChatGPT, but there are others and it’s an area that’s being integrated into a lot of software without specifically referencing which AI it’s using. The popular online image tool Canva, for example, has a host of “Magic” AI features, while Microsoft has just pushed out updates to Windows 11 which include its CoPilot AI that integrates both ChatGPT and Dall-E features as well as some proprietary Microsoft AI into the Windows interface. As such, once that update comes to your Windows 11 PC, you’ll see the older “Cortana” assistant replaced with CoPilot, which you can merrily ask to create the Mona Lisa in the style of The Simpsons to your heart’s content.
You’re already using AI (without realising it)
If you’ve got a reasonably recent smartphone from just about any brand that cost you more than the price of a cup of coffee, then you’re already using AI. No, I’m not talking about the inbuilt Siri or Google Assistant features, because you can turn those off or ignore them if you want.
Where you’re using AI is when you take a photo of just about anything. Modern smartphones rely on a lot of AI for image processing, so that when you take that adorable snap of your cat, it recognises that it’s a cat and adjusts the image output into something that should be as optimally pleasing to the eye as possible, based on millions of other photos of cats. In this case, it’s not so much heading out to the wider internet (for the most part) to compare Tiddles to every other cat, but instead using a built-up model of what people present in cat photos to adjust your picture.
If you’ve ever taken a photo with your phone and jumped into your photo library really quickly, you might notice the image change a little before your eyes. That’s the underlying AI making those changes based on what type of photo it thinks you’ve taken. You’re still in charge of taking the actual photo, the framing and so on, and it can’t make you into the next great pro shooter, for sure. They’re tools for the masses, and the quality they deliver can vary a lot – but then so too can our appreciation of what makes one photo “pop” and another one “too garish”.
What about AI fakes?
One of the bigger concerns – and it’s quite legitimate – with the use of generative AI lies around its ability to create convincing fake images, sounds or video. Sometimes these are rather obvious and designed purely for their humorous appeal, a la the rather well known (but fake!) image of The Pope in a puffer jacket. But they do have a darker side, depending on intent and how and where they’re shared.
That’s when you’re talking about what’s usually covered by the term “deepfake”. A deepfake can be anything from explicit material to material designed to actively spread misinformation. Both approaches typically try to swap in convincing-looking visual images, such as swapping faces to put people into particular situations, or modifying audio to create speeches that never happened, or both at once.
So how can you protect yourself from this kind of deep fakery? In terms of understanding images and video, it’s wise to do a little searching around those images; if a celebrity is suddenly pitching an “investment opportunity the banks don’t want you to know about” that’s probably a red flag right there, but just in case, search up that product and that celebrity to see if it’s an actual sponsorship, or a rip-off deepfake.
Actually, for all of those they’re scams, but the concept applies across any of that kind of media you might find. It’s also wise to look at images closely, because while AI generation is improving over time, it’s still got a distinct bias towards slightly unnatural features, like perfect symmetry, or oddly fused fingers in hands, or very weird lighting.