Did you see Steve Troughton-Smith’s experiments using Midjourney to generate sample album art for his Broadcasts app or to generate app icons this week?

It’s hard to look at Steve’s examples and some of the incredible results that DALL-E and similar projects are generating and not be impressed. At least, I was!

AI is all around us already. Every time you get in your car, your iPhone knows where you want to go. Everyday interactions with Alexa and Siri are all powered by AI, and that’s before considering the emerging future of bots that can mimic human conversation, GPT-3, and these image generation APIs.

But it set me thinking. I had an adverse reaction to GitHub Copilot, and I haven’t changed my mind since writing that, especially on the issues around licensing and attribution. Why wasn’t my gut feeling about these image generation APIs the same?

I think it’s probably something to do with how transformed the raw material appears. With Copilot, I can imagine the code it suggests coming directly from one of the sources. It’s hard to know how accurate that impression is, but that’s how it appears. With GPT-3 or the image generation APIs, it’s harder to make that connection. Maybe it’s because I’m less familiar with those training data sets? It’s no different, though. Thousands and thousands of human artists and authors created the training data these systems run off.

I said I wanted to get on board with these new technologies when I first wrote about GitHub Copilot, and I feel the same way deep down. The world changes all the time, whether I like it or not. It’s been less than 50 years since a computer running at 80Mhz with 8MB (not GB) memory looked like this, and less than 100 years since this photo. I should probably get over myself and move with the inevitable progress these technologies will eventually bring.

The idea that a few giant companies will use this to make even more money from the work of countless thousands of unpaid humans that produced the training data doesn’t feel great, though. Let alone the legal issues that’ll inevitably land in courts over the next few years. They’re making it hard for me to take this grumpy old man hat off for now. 👴🏻

Dave Verwer  





Senior iOS Software Engineer @ Cambridge Mobile Telematics – Cambridge Mobile Telematics (CMT) is on a mission to make the world’s roads and drivers safer. We measure driving quality, incentivize safer driving, assist users in crashes in real-time, and improve safety for millions of drivers every day around the world. – Remote (within US timezones) or on-site (United States in MA)


Only one featured job this week? That means it's the perfect time to get your company's open positions noticed with a featured listing over at iOS Dev Jobs. What are you waiting for? 🚀


And finally...

How did we ever cope without this feature? 🌈