AI is the big buzzword of computing right now – but what do you really need to know about it? We’ll take you through what it means, where you’ll find it and how to best use (or avoid) it.
What is AI anyway?
Mention AI and most of us will think of the science fiction depictions of AI, typically a mix of 2001’s HAL and 1984’s The Terminator. No surprise there; the depictions of AI in popular culture have looked for a long time at the dystopian side of the technology.
The reality for AI in the current era – it is an evolving space to be sure – is that it’s a lot simpler than our eventual subjugation by robot overlords might suggest. Call us picky, but we feel like that’s a future best avoided.
Current AI can be boiled down to the simple concept of computer algorithms that can learn from data, and the way that’s applied does vary quite a lot. At precisely the same time in history, AI is being used to advance serious medical research and creative ridiculous fake cat videos, and both are “AI”.
You’re more likely to come across the latter unless you work in some very specialised fields, though it’s quite feasible to get buried in the definitional weeds around what constitutes “intelligence” that’s beyond the scope of this guide.
Most consumer-based AI approaches work with large datasets that AI programs are essentially trained on. They repeat and repeat the same smaller inputs to work out a logical pattern that matches up with what you might expect, whether they’re generating text, images, program code or attempting to provide the basic facts around a given AI query.
At first any AI learning its model will deliver mostly junk results, but to simplify it massively, AI programs throw out the identified trash and work again with the good results to refine their ability to apply the “rules” they learn along the way.
If you recall back to your school day and learning “I before E except after C”, that’s an example of how you might train an AI, though it’s also a good example of where it might trip up as soon as it hits anything that is, for example, weird. Or foreign. Or interested in leisure, but again, this is why AI over time improves, because it’s feasible to then re-train with the exceptions as known additional rules.
This is also why the AI efforts of even a few years ago, replete with hands with dozens of fingers have improved over time, though additional digits in AI images aren’t exactly an unknown quantity right now.
A computer doesn’t entirely “understand” the full meaning of a sentence or the subject of an image, but it can iteratively be trained to deliver responses – sentences or images – that pass what we’d call an acceptable mark for quality, depending on the model used and the inputs you give it.
One important distinction to bear in mind with any kind of AI is whether the data it’s using is actually on your device (local) or using online resources in the cloud (remote).
For many complex operations remote processing will be your only option, but it is worth bearing in mind that once your data – whether that’s a text query or an image you want AI processed – leaves your device, some of your control over what happens to that data is lost also.
AI you might find every day
There’s no one company or brand that dominates the AI space right now, with most of the big players spending billions to integrate it into their products. AI today spans a wide range of tools, from AI chat tools like ChatGPT and Copilot that can answer questions, write content and solve problems in conversation, through to AI image generators that can create visuals from a text prompt. This isn’t an exhaustive list, but right now you can interact with AI models across the following ranges of products:
Windows PCs
Windows 11 leans heavily into the AI space via Microsoft’s Copilot AI, which offers a mix of on-device AI modelling (especially on Copilot+ PCs) and cloud services on a subscription basis, most prominently for consumers through a Microsoft 365 subscription.
Apple Devices (iPhone, iPad, Mac)
Apple was a relatively late entrant to the AI space with its “Apple Intelligence” product, which uses a mix of on-device intelligence, primarily through its Siri assistant, as well as wider AI features through ChatGPT. On the iPhone side side of the Apple Intelligence fence you’ll need an iPhone 15 Pro/Max or iPhone 16 or newer to access those features, while iPad users will need an iPad with an M1 or A17 Pro or better under the hood.
Android Devices & Google Products
If you’re rocking an Android phone or use Google’s search or other services, you’ll be interacting with Google’s Gemini platform.
Google provides this at different levels, from basic free stuff like the AI Summaries of searches you’ve almost certainly seen by now up to AI video generation models for paying Google AI customers, depending on your needs and precise platforms. Lower-end Android devices may not have the requisite processing power for much in the way of on-device processing and may not have all Gemini functions available to them, while more premium devices might not rely on cloud processing quite as much.
Some Android device manufacturers are also offering up their own spin on AI, most prominently here in Australia via Samsung and its “Galaxy AI” suite of features found on its premium Samsung branded Android phones and tablets, though for the most part Google’s Gemini is the platform you’ll find across most Android systems.
Do I have to use AI?
Broadly no, even if you do have a device that has inbuilt AI functions. However, if its use concerns you, it’s worth looking at your precise device settings in terms of what kinds of information are being collected that may relate to AI use and whether that information stays on your device or may be used for wider AI work.
These are typically decisions that are made when a device such as smartphone is set up, though in most cases you can access those permissions to revoke them without having to factory reset your device to do so.
It’s entirely feasible, by the way, that you’re using some form of AI without being entirely aware of it, too. If your smartphone has functionality that allows you to improve pictures automatically without needing specific tweaks to features like photo highlights or contrast, that’s typically using a form of AI because the way many of those automatic photo improvements work is via an AI engine that’s been trained over millions of photos to “know” what typically goes into a good-looking image.
You may also end up interacting with AI in some ways in your daily life without entirely being aware of it. If you’re applying for a job, for example, there’s a lot of evidence that many employers, especially in office-based positions are using AI agents to cut down on the analysis and time it takes to sift through employee applications – and in some cases that can extend all the way out to AI-led interview rounds to gather up employee information!
The reality here is that while there’s a lot of hype around AI, its usage is growing in specific ways, and it’s likely that it will become very difficult to entirely avoid it in the future, even though it’s still something of an emerging and evolving space.
Do I have to pay to use AI?
Any use of AI requires some level of computing resources, but whether or not you’re paying to use those resources depends largely on the services you use.
As an example, Google’s AI search summaries are built into how it works now, and that doesn’t come with a financial cost. However if you want to use the more advanced tools in Google’s AI subscriptions and access to its more powerful cloud-based AI models, that comes at a financial cost.
On the Apple side of the fence, while the current basic crop of Apple Intelligence tools don’t incur at cost at this time, that’s not guaranteed for the future, and because that product does include some ChatGPT tools for wider queries there’s the potential for payment there too.
Microsoft’s Windows Copilot has a mix of on-device AI features and some web-based features at no cost, but some wider features are tied into a broader subscription package. It all depends on your usage, essentially.
Can I trust AI?
Trust in AI is a big and complex issue, made more difficult by its increased use to generate text, images and videos that may have some level of deceptive content within.
That’s a human-led issue, ultimately, but there’s also the wider issue of using AI for analytical purposes, whether that’s to summarise a text or perform a search or even create some content, especially text that might rely on multiple complex sources or contrasting works with differing viewpoints.
How about we ask AI about its trustworthiness? Throwing the question at Google gets this result:
“AI cannot be fully trusted as a reliable source of truth or autonomous decision-maker, as it often produces “hallucinations” (fabricated data), errors, and potential biases. While useful for efficiency and processing information, AI lacks ethical reasoning and human judgment, requiring human oversight to verify outputs and manage risks.”
There’s something inherently amusing in an AI telling you that you can’t trust AI, but it’s fundamentally not wrong either.
One of the bigger issues around AI trust is in the area of “hallucinations”, where an AI lacks the discernment to entirely understand the meaning of a given text (for example) and tries to essentially pattern-match to what its programming has told it is likely to be an acceptable result, even if that result isn’t actually accurate.
This can extend out to areas where AI is being more prominently displayed, such as search results. If you’ve used Google recently, you’ve no doubt seen its AI search summaries or been prompted to use its live AI search functions. While many results may return an entirely accurate result, that’s not always the case, sometimes with amusing results – for a time, some search engines declared that you should eat at least one rock a day, or that there were no African countries starting with a K, which was news to Kenya – but it also raises some serious issues around trusting the results you find that way.
AI can be a useful tool for quick and simple research, but for anything complex it shouldn’t be the endgame; it’s well worth digging further into results and making sure that the outputs an AI query gives you are backed up with some solid sources, in proper (if somewhat more time-consuming) academic style.
Need Help with AI on your devices?
Need some help with an AI function on your devices, or how to set it up or reduce its use on your home or office systems? One of our friendly Geeks2U support technicians is ready to help you – give us a call on 1300 433 572.
