[LoginTech] Apple AI Became More Convenient but Lacked a Surprising 'One-Shot' Feature

Analysis and Search with Just a Camera Point
Photo Cleanup and Writing Assistant Features
Mistakenly Identified Samsung Laptop as "Looks Like a MacBook"

Last weekend at dusk, I pointed my iPhone at the Han River from a building in the northern part of Mapo. When I pressed and held the right camera control for about 2 seconds, Apple’s artificial intelligence (AI) feature, 'Apple Intelligence,' activated its visual intelligence function. The AI recognized the scene captured by the camera lens and ChatGPT analyzed it by saying, "This is the Han River in Seoul, with high-rise buildings and a Han River bridge visible in the distance."

I tried Apple Intelligence, which Apple launched in Korean on the 1st of this month. The iPhone, enhanced with generative AI features, has become much smarter than before, but users familiar with various AI services like ChatGPT might find it hard to discover any groundbreaking innovations.

What stood out the most was the 'visual intelligence' feature that recognizes and analyzes scenes captured by the camera lens. Visual intelligence is divided into 'Ask' and 'Search' functions. For example, when I took a picture of an exercise machine I didn’t know the name of at the gym and pressed 'Ask,' ChatGPT accurately identified it as a 'step box' or 'lat pulldown.' When I took a picture of an unknown flower, it provided the flower’s name and its meaning. When I scanned an English product manual, it naturally translated it into Korean. When I took a photo of a poster for an IT event I needed to attend for reporting purposes, a window titled 'Create Event' popped up, recognizing the date and location on the poster and adding it to my calendar.

This time, I took a picture of the laptop screen and part of the keyboard I was working on. Seeing the weather information on the monitor, it said, "You’re working on a sunny day," which surprised me. However, it mistakenly said, "Judging by the keyboard layout, it looks like a MacBook." The laptop was actually a Samsung Electronics product. ChatGPT’s responses could be copied with a single tap but were limited to about 5 to 6 lines at most and were not saved automatically. The search function in this visual intelligence feature works by searching the subject on Google Images. Since it was similar to the 'Circle to Search' feature shown on the Samsung Galaxy S24 released in January last year, it didn’t feel particularly new.

Comparison before and after deleting the price tag in the photo using the cleanup feature.

The photo editing feature introduced a new 'Cleanup' tool. It flawlessly removed small elements that distracted from the main subject of the photo. It could be used on any photo in the photo library, including those taken with other devices. Cleanup worked even when not connected to the internet. However, when removing large objects in photos, traces like edges remained, resulting in unnatural outcomes.

Apple’s ambitiously launched 'Writing Tool' transformed roughly written text into complete content. Popular functions such as 'friendly,' 'professional,' and 'concise' were organized into a separate section, allowing users to convert text with a single tap. Additionally, users could freely change the tone, for example, by saying "change to SNS style," and use ChatGPT’s writing capabilities to add content. Text could also be converted into tables or reformatted to highlight only key points.

The Siri integrated with ChatGPT allowed for more natural and comfortable conversations than before. It understood the context of speech and could deliver the desired results even when mistakes were made, such as "Set an alarm. Not 7 AM, 8 AM." However, seamless integration between applications (apps) still seemed challenging. When I asked Siri, "How’s the weather in Seoul today?" and received an answer, then said, "Write the Seoul weather in the notepad," it only entered "Seoul weather" instead of the actual response, causing some disappointment.