Branding in the real world: acronym branding (2016) | Hacker News Reader

Branding in the real world: acronym branding (2016)

brandingmag.com

27 pointswalterbell11 comments

Sometimes it's enough that it's a catchy word with a free .com or .io domain. Doesn't need to be a meaningful acronym always, think Google, Sony, Xerox, ....

I made a name generator once that tries to do that: https://catchybrand.vercel.app

If you're interested in computer-assisted backronym composition, please check out this interactive tool that I made, based on GPT-NeoX/Pythia models: https://github.com/dwrensha/acronymy-assistant

3-minute video explanation: https://youtu.be/LjOHnXRIp4Y

Check the Google when you make a new acronym brand. If there are no significant search results and you can pronounce it, and get the domain, congratulations, you have a brand. Might have to be 5 letters or longer though.

>We waste precious time and resourcess explaining what the name means, instead of touting the benefits to customers.

This would happen with any name that isn't a description of products, so almost every company really.

CCC exists and stands for Chaos Computer Club.

I don't know why but the story of CCC reminds me of MMM (aka Minnesota Mining, aka 3M).

You can give ChatGPT a word and ask it for an acronym that spells out that word, and that the words match a certain brand/product/idea

Why use ChatGPT when you can pay 7 figures for a world leading branding agency to do the same exercise?

Isn't that called a backronym?

Do you have any prompt tips for this sort of task? I’ve tried it a bit ago and didn’t manage to get very good results even with a lot of iteration. Could have been the niche subject matter though

I don't think there are good prompts for this. ChatGPT sees the world in tokens, not letters. That makes it an uphill battle to make it reason about individual letters. It's not impossible (it can tell you the first letter or the number of letters of words with decent accuracy) but it certainly has a harder time than a human raised on words composed of letters has.

This would be a good usecase for networks trained on letters instead of tokens, but I'm not sure if any big ones exist, since for 99% of tasks tokens are just better

Reply on news.ycombinator.com