Comment by simplotek | Hacker News Reader

Comment by simplotek | Hacker News Reader

Parent Full thread

> Apple's AI is on your phone. It unobtrusive. Recently I needed to quickly find a company's QR code which I had photographed and Photos search found it in an instance.

I believe the iPhone automatically classifies photos based on who or what show up in them. Users can contribute to train classifiers, but the iPhone already works out of the box.

The iPhone also creates theme-based photo albums from the photos you took. I recall it creating Christmas photo books, photo books featuring a pair of people, persons and pets, etc.

This might be low-key AI, but it's the useful kind.

> It even finds text inside images.

I'm not sure OCR counts as AI. But yeah, the iPhone indeed does that too. We can take a photo of a telephone number or even a credit card and automatically fill in those numbers.

I worked on a similar feature in the past for an unrelated project, but that was not AI though. Marketing changes though.

> I'm not sure OCR counts as AI

This is a great example of the AI effect where people would call something AI when it was a daunting research problem but give it another label once it’s working:

https://en.wikipedia.org/wiki/AI_effect

In this case, Apple’s modern OCR is a complex neural network system which I think most people would class as an AI tool. It’s notably better than traditional approaches which were optimized for business documents.

> It’s notably better than traditional approaches which were optimized for business documents.

I'm not sure I agree, primarily because "better" is subjective.

A pipeline with template matching is extremely effective at extracting fixed form text in a standard layout, such as telephone numbers or credit cards, and computationally cheap as well.

But I presume a drop-in black box model which isn't bound to a low computational budget, can output plenty of false negatives and false positives, and can run on a single pipeline might be preferable at least from a product management point of view.

Also, neural networks looks good on resume while template matching doesn't. Just like statician/image analyst looks lukewarm but AI engineer looks superb.

I’m just going by the quality I perceive as a user. It handles basically every CAPTCHA, difficult scans of printed documents, etc. better than Tesseract. I’m sure there is lots of hard work beyond the pure ML component but from a user’s perspective it’s impressive.

> (…) better than Tesseract

Isn’t Tesseract also neural network-based?

https://github.com/tesseract-ocr/tesseract

They added that somewhat recently - the history dates back long before that was practical.

Reply on news.ycombinator.com