My Favorite AI Use Cases As an AI Skeptic

As with Metaverse and NFTs before it, it’s almost impossible to follow tech news these days without hearing about artificial intelligence. I understand that this can be annoying: the technology is still in development, and everyone is still figuring out the ethics of making AI art or how to make it not suck . Then there are tech execs who promise big changes to your workflows before turning around and introducing tools that rewrite your emails to make them sound more robotic . It’s tempting to write AI off entirely.

That would be a mistake. Even as an AI skeptic, I have discovered several key use cases where AI performs better than any other alternative. Even if AI doesn’t become the magic bullet the tech industry is hoping for, in the right situation, AI can be a savior. As an AI skeptic, I believe this is where AI is truly most useful.

When to use AI

I tend to think of AI the same way I think of most automation software—as a fatigue killer. While I think AI image generation can often feel like taking existing illustrations and changing them slightly without the teacher noticing, or generative search usually ends up feeling like a rebranded “I’m Feeling Lucky” button, there are plenty of places. where machine learning has already shown its capabilities.

Even before “AI” became a buzzword, you’d already seen similar features in tools like Photoshop’s Magic Wand, which saves artists time by allowing them to select a photo subject with one click instead of having to zoom in and carefully use the tool manual selection. In these situations, no creative decisions are made, the artist does not lose control, and no copyright infringement occurs. Instead, machine learning does what a search engine or human can’t easily replicate, and at a much faster rate. This feels truly new.

This is the type of AI that interests me. If you’re asking yourself, “Why would I want a robot to write my emails,” here are a few problems for which AI might be the best answer.

Use AI to change the aspect ratio of an image

Credit: Adobe

Consider the following scenario: you take a photo with the wrong aspect ratio and now you need to crop it to fit on your website or slideshow. It sucks to feel like you’re wasting content, and that’s where AI-powered generative padding comes into play. Available in a growing number of photo editing tools, including Adobe Photoshop and Microsoft Cocreator in Paint , generative fill uses artificial intelligence to fill in the missing parts of your photo so you can make it any aspect ratio you want. Depending on your tool, you may be able to do this using a tooltip or by stretching an existing background. You might even have access to both options. But the basic principle is the same: unobtrusive AI that keeps more non-AI information in the frame.

This isn’t always successful, but if the background is simple enough it’s hard to tell it was ever used. However, like any artificial intelligence, it needs some human supervision, especially when it comes to photos of people. Take this example for example: a tech expert sent a selfie to a conference for promotional use, but it was extended by AI without her consent, resulting in the photo being more revealing than the original. The conference didn’t intend to change its appearance—they just wanted to change the photo’s aspect ratio—but the result was still misleading.

AI scaling

1 credit

AI scaling actually predates the recent trend towards generative AI, as most gamers will tell you they were using it before ChatGPT became a household name . But as companies embed neural chips into more and more devices, scaling AI is becoming more common than ever.

When used correctly, this method allows movies and photos to look sharper and even games to run faster. It uses machine learning to analyze images and add details that were not present in the original photo. This is useful either when processing graphics in real time, or when working with material that cannot be re-rendered from scratch, such as old photographs or compressed retro video game graphics where the original files are missing.

For example, let’s say you have a photo of your grandmother taken over 50 years ago. It’s as detailed as it looks, but with tools like Super Resolution in Microsoft Photos , you can use AI to increase resolution, forcing the machine learning model to try to fill in the gaps while sticking as closely as possible to the style of the original image.

This is not without controversy, as it can still change the mood of a photo, and when applied to things like video game remasters, it can change the artistic direction if not sufficiently controlled. There is also the ability to scale the AI ​​to hallucinate or render dreamlike images if the original image is too low resolution.

But with the right human curation and sufficiently rigorous training datasets, many professional products have leveraged AI at scale to great effect, updating hundreds or thousands of assets at a much more reasonable time and budget cost than if the project had been done entirely manually.

Perhaps the most common use of AI scaling by the average person is to improve video game performance. Tools like AMD’s FSR, Nvidia’s DLSS, and Sony’s upcoming PSSR use AI upscaling to run games at lower native resolutions but still look good enough when displayed on a high-definition TV, which in turn allows they provide higher frame rates. As techniques like real-time ray tracing become more common, AI may be the best way to take advantage of the fancy new graphics techniques being sold on consoles while keeping games fluid enough that they feel like you’re not sacrificing performance to see all these details. .

When something is on the tip of your tongue

I often find that using an AI chatbot to find information is less useful than a search engine. I feel like it robs me of the ability to see search results myself and instead acts as a middleman, as if I’m not old enough to understand Google pages. But there’s one situation where search engines can’t help: You’re not even entirely sure what you’re looking for.

In other words, AI is great when you have something on the tip of your tongue. Let’s say there’s an old movie that you remember a scene from, but you don’t know anything else about it. In this case, even coming up with keywords to place on Google can be difficult. There is a subreddit here where you can get real human help , but the problem is that you have to hope that your post will be noticed and someone will respond in a reasonable amount of time.

While you wait, maybe AI will answer your question instead. For example, a few weeks ago I was looking for a particular Batman comic that I heard about on YouTube but couldn’t remember the name of. It was supposed to have weird visuals and dark tones, but since those aren’t exactly unusual themes for Batman , Google only showed me the more famous comic with those themes. Meanwhile, just Googling “Batman comic with psychedelic art” brought up fan art and, ironically, a lot of AI-generated images, but not the comic I was looking for. (This also happened if I excluded a result from the search that I didn’t need.)

Meanwhile, both ChatGPT and Microsoft Copilot guessed the comic I wanted on the first try. I simply told them, “I’m looking for a Batman comic with trippy, psychedelic graphics and dark tones. This is not Arkham Asylum: A Serious House on a Serious Earth . I think it was published in the 1980s.” Bam: I was immediately directed to Batman: Cult , the comic I was looking for, without having to consult an actual comic fan or dig into my YouTube history to find the video where I first heard about the comic.

How AI can help people with disabilities

While I may not want an AI to write emails for me, not everyone is in the same boat. One of the reasons for this? For some people, writing an email can be physically taxing.

Let’s say you have a physical impairment that prevents you from typing words easily on a keyboard, but you otherwise have no problem understanding what you want to say. In this case, instead of painstakingly sitting at a desk and going through a few paragraphs, you might prefer to give the AI ​​a few sentences of what you want to say and then just look at what it has drawn.

This is just one example that Fredrik Ruben, head of the Dynavox Group , recently gave me. Dynavox makes assistive software and hardware for people with physical and mental disabilities, and while he told me AI poses some risks, it also gives his company a much better opportunity to help its patients.

Take AI voice cloning. This is also available inApple instruments and allows those at risk of losing their voice to train an electronic replacement to sound more like them. Previously, this required writing down almost every word in your language, but Ruben says this can now be done in just 50 sentences. This not only saves significant time, but also means that past recordings can now be used to clone voices, making the tool accessible to those who have already lost their voice.

Or let’s say you need an electronic voice machine to help you speak. AI can learn your speech patterns as you use them, providing more useful autocomplete options and allowing you to communicate with those around you faster than if you had to manually type in everything you wanted to say.

These are all areas Ruben is excited to continue exploring, and beyond our conversation, it’s worth noting that other companies are using AI-generated live captioning to help people with hearing loss. However, Ruben noted one area where using AI to help a patient with a disability poses a risk.

“AI is quite dangerous if the user has cognitive impairment,” Ruben said. This is because people with certain cognitive impairments may not know what to put in a prompt or how to proofread the AI’s draft to make sure it says what they want it to say. Instead, they may simply agree with whatever the AI ​​writes for them, even if it doesn’t reflect their true feelings. In such situations, it will be easy for the AI ​​to “impersonate” the user by misrepresenting it without the user knowing what is going on. Their friends or family will essentially only be talking to ChatGPT, without much input from the patient, and the interaction may leave the wrong impression.

Tools like AI voice cloning also come with risks like deepfaking , and it remains to be seen whether AI-composed or edited emails will be as effective as hand-crafted ones on a large scale. But on an individual level, these AI tools can help level the playing field when used responsibly, even if you don’t see much benefit in them yourself.

Easily enter data into a spreadsheet or transcribe audio.

Data entry is one of the most tedious parts of any job. When all you do is copy and paste information from a document into a spreadsheet, there’s little your mind can support. And yet it can take so long. This goes double if you are transcribing audio.

So AI quickly became very good at both of these tasks. Depending on your tool, you may need an extended subscription, but simply giving your AI chatbot access to your document and asking it to turn it into a spreadsheet is surprisingly reliable at this point. There are even multiple ways to do this, from simply using a text prompt asking for a CSV file that you can import into your spreadsheet program, or using a third-party tool to automatically generate a PDF and convert it to Excel. sheet . If you only have a few data points or all the data you need is available online, you might even be able to get what you need right from the chatbot’s regular interface without having to download a document. Take this McDonald’s menu I created using Copilot. (Here’s the actual menu for comparison.) You’ll need to double-check your output, especially if you’re asking the AI ​​to pull information from the Internet, but you’ll finally be able to get a rough draft without having to reformat the data manually.

Credit: Microsoft Copilot

Likewise, programs like Otter.ai , and now iOS Voice Memos, can generate full transcriptions of audio recordings. For me, as someone who started out in journalism, spending hours stopping and replaying interviews my senior colleagues conducted to transcribe them is my savior. It’s not always completely accurate, but editing an AI-generated draft of a transcription is much faster than simply writing it from scratch.

Quickly create travel routes

Time for something fun. While AI can help relieve boredom at work, one of the times I use it outside of the office is when I’m planning a vacation. Knowing what to do when you go on a trip can require dozens or even hundreds of Google searches, and an AI chatbot can narrow that search down to a single interaction.

Of course, you’ll probably want to supplement or cross-check your AI’s suggestions with your own research, but if you don’t know anything, asking the AI ​​to plan a sample stay at a destination can help you know where to stay. start looking. This way, you can avoid idly scrolling through travel TikToks or seeing what a travel website’s algorithm wants you to see.

Take for example the trip to Japan that I asked my co-pilot to plan. While the results it gave me are a bit basic and I’ll have to do my own searching to find restaurants or book accommodations, in general terms they’re surprisingly similar to a trip I took in real life shortly after Japan began easing restrictions for travel due to the pandemic. We were forced to use a travel agency due to when we booked the room, and yet this free itinerary is not too different from what the paid agency offered.

Credit: Microsoft Copilot

Of course, something like the above is a start, but using AI to map out a rough route before diving into deeper research can reduce the overall amount of work you’ll have to do. It will also allow you to tailor your needs to your specific trip, which can make it a little more flexible than other tools.

Take stock of your inbox

It’s time for me to admit my secret shame. According to my phone’s mail app, I have over 89,000 unread emails. I gave up: I simply do not have the ability to get to mailbox zero on my own.

Except now I don’t have to be alone. Outlook , Gmail and Apple’s Mail app are either working or about to roll out AI-powered help to clean up your inbox, with the latter set to debut this week . Consider automatic mail categorization (for Apple Mail), sentence-long article summaries, and other tools to help you get up to speed quickly. The AI ​​will even help you craft your answer.

As I said earlier in this article, I don’t necessarily need it all. And I’m still skeptical about whether I should let the AI ​​try to relay an email from my boss rather than just read it directly. But compared to being paralyzed by the size of my inbox and just missing most of it? Yes, this should make my browsing more productive.

I especially see this being useful when you’re returning to the office after a long commute or if you’re tired of your phone pinging you during the workday. For example, Apple Intelligence mail summaries are also combined with notification summaries, as well as the ability to set a focus that only bothers you when the AI ​​determines that a notification is urgent. Then again, I get a lot of emails, so while I’d be nervous using this focus mode when I’m waiting for an important call, I can also see it as a viable alternative to turning off the phone when I find myself getting an annoying beep every five to ten minutes.

This is where I, as someone who still longs for the days before I knew what GPT was, am most inclined to use AI. As someone who still worries about the ethical implications of where AI tools collect their data sets and how much power they consume , I tend to use them quite sparingly, preferring to only use them when they either save significant time or get something done what I need. I couldn’t do otherwise myself. I feel like I’m not alone there. For better or worse, it looks like AI isn’t going anywhere anytime soon .

More…

Leave a Reply