It is nearing the end of 2023.
Looking back on 2023, it seems that among the year's technical keywords, "AI" was undoubtedly the hottest.
Everyone seems to have used ChatGPT at least once and tried to utilize it. Especially since I started teaching at the university this semester, I could easily see students actively using it for assignments and reports.
I, too, am making good use of AI-based services, such as the YouTube captioning feature and Adobe Photoshop 2024's Generative Fill feature.
This feature was officially added as an internal default feature starting with Photoshop 2024, and I am finding it extremely useful when working on YouTube thumbnails or performing simple photo editing.
If I could go back to a time when this feature didn't exist, I can't even imagine it. If I accidentally run an older version of Photoshop, it immediately switches back to the 2024 version to run. Because of the Generative Fill feature..
Aside from that, although I am not currently using them, I am also interested in upcoming services. Among them, I recently had quite an interesting experience with a service called Soulmachine, where I conversed with an AI.
In this context, I am living with great expectations about how much the upcoming AI-based services will help my work environment and life.
However, at the same time, as I gain experience, I feel there are parts I should be wary of when using artificial intelligence. This is because I have recently experienced some inconvenience myself.
The Hidden Traps of AI-Based Services.
My Chost CMS, where I am writing this post, allows me to add a feature that translates posted articles using an external service called Weglot.
Of course, you have to pay to use it beyond a certain limit, but since I tried the trial version and was quite satisfied, I looked for a service with good value for money among similar services.
After actually applying it, I found that by simply adding a few lines of code to the header of the Ghost blog, translation was easily applied, making it very easy to build a service that provides the blog in English.
This was a very convenient service that used artificial intelligence technology to automatically translate pages and even helped with Google SEO.
Even though I have to pay around $15 a month.. I thought it was definitely worth paying for if the translation is this convenient and has such good features.
However, my satisfaction was completely shattered after just one day.
Because it was stated that the startup tier I purchased and used supported 15,000 words of translation.. but this amount was not the monthly recharge amount. It was the total number of words that could be translated while the account was maintained.
So, after translating my blog for about a day, I had already exceeded 15,000 words.
And the service was stopped. (Huh?)
Surprisingly, even words that had already been translated were no longer provided by the translation service. They were simply in a suspended state.
Of course, since upgrading the tier would resolve the issue, Conveythis kept emailing me to pay additional fees.
After a moment of hesitation, I decided to pay the additional fee to use a more convenient service. I thought about it. And as soon as I paid the additional fee, I started translating incredibly and consuming my usage.
Of course, upon checking later, there was also a way to organize text to be translated versus text not to be translated within Conveythis, and there was an option placed somewhere ~~here to stop translating once the translation volume was reached.
However, there was no way to separately disable the important post that AI had already translated on its own, and if things continued like this, it seemed I would have to pay for the higher tier again to use the service.
In other words, it cost much more than I expected.
So, I decided to give up using the service.
Since the additional upgrade tier I paid for was barely used, I was able to cancel it immediately, but there was no way to cancel the original payment.
So, I decided to inquire about cancellation through customer support.
Wait, there's AI even where I'm inquiring?
It seemed that they were utilizing a separate AI-based CS service for service operations. Well... as long as inquiries were getting through, I thought it would be fine, so I requested subscription cancellation and a refund.
But even after waiting a day, there is no reply.. and refunds aren't even processed..
While looking for the Conveythis page, I did find a page to send an email directly, but seeing an AI-like image there, I started to feel a sense of distrust.
So, I decided to look for a more direct refund method, and upon checking my payment history, I confirmed that the payment was not made through Conveythis, but through another service called paddle.net.
So, when I contacted paddle.net, a human—not an AI—answered immediately, saying they would forward the content to Conveythis, and within a few hours, an email arrived directly from Conveythis.
"I'm going to cancel my subscription."
as the user confirmed. So, when they said they wanted to cancel their subscription, an email arrived immediately stating that the cancellation was completed.
However, it didn't end there.
Even though the subscription was canceled and the account disappeared.. (I had them delete it completely). The refund was not processed properly. Since there was no news about the refund after waiting two weeks, I contacted Paddle.net again. The representative said they would check and the refund was completed the very next day.
As always, it's more convenient and faster with humans..
AI-based services are not necessarily bad.
In the case of the issue I experienced, it is not that the Conveythis service is problematic or bad. As mentioned above, upon checking the account details later, I found that there were detailed options to translate more efficiently and that there were methods to impose usage limits.
Therefore, it is possible that I am the problem for paying and using it without properly understanding it.
Also, since I was able to neatly complete the subscription cancellation and refund, it seems that there are methods available to resolve issues when they occur.
However, I want to emphasize that when using AI-based services that seem to work perfectly once you pay, there are many points we need to carefully consider.
It is said that a house built on sand collapses.
This phrase is often used to liken something that looks flashy on the outside but is actually built on a shoddy foundation, like a building built on sand.
This is the sentiment I feel after going through this experience.
Ultimately, the essence of using and utilizing services is not significantly different. However, building and providing such services requires careful consideration and involves a considerable number of use cases.
Unfortunately, it still seems difficult to automate and resolve all of these things based on artificial intelligence. Of course, if the user themselves has a high level of understanding and can smartly handle the cases, current AI services will be very convenient and useful, and in some cases, costs will be significantly reduced.
However, as they advertise and as consumers like us expect, for all service processes to be provided truly conveniently and easily, there still seems to be much left to think about and build.
And just as I was surprised by the extra cost incurred when trying to easily provide translations through automated services, it seems that companies providing AI-based services are also likely to be surprised by unexpected costs due to their overly high productivity.
When looking at recent news about artificial intelligence, it is easy to find stories about efforts to reduce costs; however, this trend likely hides an important story behind the scenes of these amazing AI-based services.
Ultimately, what matters to us is not AI, but service quality
There was a time I visited a company that claimed to provide AI-based review services, but I saw more than 50 employees manually conducting reviews inside the company.
"We are proceeding with the basis of AI, but we cannot take responsibility for that. So, people are only utilizing AI in the direction of having humans review and tune it first. That is much cheaper."
Yes, ultimately if quality is not secured and problems arise, it is still much more intuitive and cost-effective for a person to step in and resolve the issue directly.
It feels like we are living in a world where, just as I experienced, when AI fails to respond, people step in to resolve the issue.
Well, this issue likely varies depending on the service field or level, but ultimately, if it cannot be resolved, consumers using it will suffer losses in time and money. Therefore, consumers need to properly understand and choose to use it.
It still seems that people need to evaluate, choose, and use things a bit more carefully.
Since AI might not even reply, that's fine too..