It also seems like because of this pattern a lot of the tools wouldn't actually be that useful in the field. I've played around with a few now that 1. it was almost impossible to test and debug behavior because you can't tell which elements of the prompt cause it to do what and 2. you can rarely get it to do the same thing each time it's triggered with 100% confidence, which make them pretty useless as workflow tools. This definitely feels like bandwagoning rather than actually coming at it from the perspective of what anyone would find useful.
I don't even get the sense the questions/answers have been written by a human? I have no idea how you could land on the three available options for the cultural environment question.
Which questions do you mean? They were definitely written by a human (me). However, as I am a non-native English speaker, they might not be clear enough.
I could have been more constructive her in fairness! It's not so much the language as the presentation of the options in that specific question - they're kind of presented as a scale but the middle options deals with a different variable to the outer ones - western vs non-western and traditional vs modern. It might be worth breaking these kinds of questions into Likert scaled for more tightly defined variables.
I had some fun trying to get this to work with prompts based on specific brands etc. It felt like it had a pretty good attempt at a Fender Jazz Bass and a Moog Mother, but struggled with a Juno 6. Is that pure coincidence or would the model be able to understand the semantics of specific instruments and things as well as more generic terms?
Good question. In my experience combining generic descriptors is what works best. This is probably due to the text captions used during training mostly consist of generic instrument names, genre names and adjectives.
For arrangements/history/background/variations on folk songs, especially Irish, Scottish, and American ballads, mudcat.org is an incredible resource that's still rocking the early 2000s plain HTML look.
Can't say enough good things about sheldonbrown also. It's the ultimate resource for people who want to bodge a bike build from their parts bucket.
We set out to experiment with whether we can usefully automate low-code development tasks using RPA. While we were able to set it up to ship tools relatively easily, the constraints we encountered make it tricky to see many real-world applications for doing so. This article explores what we learned.
I've tried to create resources like this in the past for internal use in agencies and also mentoring clients and I think the thing that makes it really tough is that for any of this kind of information to be actionable the would-be SEO needs to get a grip on the creative and strategic elements. For example - I always found it tough communicating to colleagues/clients that it's one thing seeing a particular keyword on Semrush/Ahrefs that they'd like to after - it's another having enough experience to work out if it's something that's worth their effort based on things like their domain authority/core web vitals - and then again knowing what to do to go after it in RE search intent and how to out-compete the existing results.
Then there's the extra tricky bit that the kind of people reading these types of articles are often going to be small business owners/entrepreneurs who aren't necessarily going to want to put the time and effort into the above.
I don't mean this is a criticism of the original article to be clear. It's my own experience of trying to create resources like this that readers would be able to pick up and use to drive traffic (and ultimately make money). This article goes after more of a glossary-style approach, which is still going to provide value - knowing the lingo is really important for the kind of people who will want this kind of resource, if nothing else to help them understand what they're talking about if it comes to dealing with external stakeholders or agencies etc without getting ripped off. It's an inherently difficult topic to be comprehensive about in a blog post - if it wasn't, there probably wouldn't be so many people making good money as full-time SEOs.
I love this. I've definitely seen one of those TV chefs go to SE Asia somewhere and be served a broth that claimed to be >150 years old - probably Anthony Bourdain.
Tough one to get past the hygiene inspector I guess though.
This is the exact point in the market I try to look out for when I'm buying any kind of product - not just studio gear. I've heard someone referring to some kinds of products as the 'AK-47 option' - as in reliable, ubiquitous, mendable, versatile and (relatively) inexpensive. I think the other classic examples were telecasters, SM57s and the Toyota Hilux if memory serves.
Anyway, really cool to see a write up like this on an iconic but often unsung-hero of a piece of gear.
Is this always on or can it be opted out of? What if you don't want your data exposed to chatgpt