Aren’t these “new” AI features impractical? — A data scientist’s rant.

Google’s new AI features came out recently. Since then, my Twitter feed, newsletters and YouTube recommendations are all filled with the new AI features and how they are the “best”. I couldn’t help but feel irritated seeing these things pop up again. Even now, some people are debating the quality of Gemini’s new features and Apple Intelligence when neither of these services has been released yet. Don’t get me wrong; I was the most excited when GPT-3.5, GPT-4, DALL·E 3, and MidJourney V6, etc, were released. However, things trending in AI right now feel somehow artificially hyped up. They are not as interesting as they make it to be.

Many recent advancements in these services, like image understanding, live speech, etc, are deeply appreciated by technical and day-to-day users. The proofreading option in Apple Notes, the “What’s on my screen” from both Apple and Pixel phones, and call notes are impressive and are products of hard work from a technical viewpoint. While not everyone uses these features daily, they are valuable for niche groups, such as writers refining their work or sales teams enhancing their feedback processes. They remove a few steps in searching from services like Claude/ChatGPT.

But why do I need image generation integration in my photos app? The ‘Magic Eraser’ feature makes sense because it makes the photos more personal. So does moving people around after taking an image. But why would I want a random moment of my life in 8k? Especially when most of my displays are limited to at most 4k? Even the phone selling this has no use for 8k except for upselling cloud storage. Generating Emojis has some meaning; we use emojis a lot in our day-to-day lives.

In the end, how often do you open up ChatGPT daily? How many phone calls do you want to take note of? How many images do you generate daily? How many times do you zoom in on an image? The calculator is an interesting change, though I’m not sure how often people would use the “math notes” on it…

P.S. If you are an experienced developer, skip the following paragraph. You already know it, anyway. If you are starting in ML, give this a good read. But, if you are an influencer, someone posting things like “10 amazing things you could do with XXXX”, or “You are using XXXX wrong, here are ten ways to use it better”, etc., then read this 5–10 times. Ingrain it into your mind.

ChatGPT was a one-time event. AI has evolved quickly in the last decade, and these models are proof of it. However, not every advancement in AI is life-changing for a normal user. It took a lot of groundbreaking innovations for ChatGPT to come into existence. Some had to train an image classifier, others had to say that this approach is also suitable for text, some had to use it for translation and present the idea of “Attention” to developers, and others had to spend much time saying “Attention is all you need”. Then, people had to collect a lot of data, process it, verify it and work on it to release the first GPT. These took much time. The further versions, too, took a lot of time. This process also had many failures, including Microsoft’s Twitter experiment and Amazon’s AI recruitment system. Many developers had to have spent many sleepless nights trying to integrate these into browsers, operating systems, apps, etc. Things take time. Don’t hype everything up for your engagement farming. You risk being seen as just another bot, or perhaps you’ve already crossed that line.

These services and the integrations are amazing. We MUST appreciate these remarkable things and those we can achieve with them. But let’s not overhype cosmetic changes and reserve our appreciation for more practical and life-altering ones. All this hype only overwhelms genuine enthusiasts and discourages them early in their pursuit.

Thanks for reading till the end. If you like this article, follow for more.✌️.