As nearly always, completely agree chap. I think particularly in some areas of Comms, given the speed at which things are needed/expected, you do need to recognise 'look this is good enough, let's just get on with'.
My personal bugbear here was always press releases. Where with some companies you had to go through about 20 iterations/drafts as multiple people weighed in with a point of view about how it wasn't yet 'perfect'. And 95% of the time, the final version looked more like the one you had originally drafted than the one that got changed and amended by various people who *just had to have an opinion*. Utter waste of everyone's bloody time...
I am going to pick you up on your start point for this article! ;). Because the reason people liked AI poetry was because it was 'too perfect'. So therefore perfection is important to a lay-person, no?
A minor point chap!
TBH my biggest issue with AI at the moment is that it is *nowhere near* perfect (apart from, apparently, on poetry), and that it invents, obfuscates and is downright wrong in some of the stuff it delivers. I have spent (and think everyone in Comms should) many hours explaining and drawing down the guiderails I expect my Gen AI partner to work under. No BS, no making stuff up to 'please me', no un-sourced data points.
This also brings in the 'verification tax' and the ongoing need for humans in the loop. Research shows that for every saving you get from AI, you need to put in at least 40% of that saving back into review, verification and fact-checking. Because if you don't, then you are putting out AI-slop that is potentially 'fake news' and can damage your brand.
Now that is not 'good enough'.
Should we seek 'perfection' in Comms? No of course not, it is actually impossible and all it does is slow down a function that needs to move at speed. But, we must remain wary of what good enough actually is and not let our guard down...
Thanks for the comment, pal, and agree with your points.
I think it’s really important right now that we all know AI’s output is far from perfect, even when prompts have been carefully crafted to be as specific as possible. Taking what AI produces as an additional source of input can certainly be valuable, but using it verbatim and without further verification is the quickest route to a poor result (and a potentially damaging one).
It brings me back to the quote I’ve been using far too much recently, that the people getting most out of generative AI are those who could do the work without it. You have to have the knowledge to spot the inaccuracies, or at least question the information. A poor writer isn’t going to be able to ascertain whether the copy produced by AI is good writing (or is likely to assume it is, simply because it appears to be grammatically correct and not full of spelling mistakes).
Which sort of relates to the point about the poetry. I took that to mean that if, to the inexpert eye, a piece of prose perfectly adheres exactly to what people think a poem is, they’d regard it as a good one. But as we all know (and as Gallagher was getting at) genuine perfection - in any form - is full of imperfections.
Ha! And here is the beautiful irony of AI in a nutshell. "the people getting most out of generative AI are those who could do the work without it."
100% true... for now. And my point about the irony. Because the AI learns most from the people who can work without it, and refines its models based on the humans with the highest skillsets to challenge it and not need it, but recognise the time and productivity gains it can bring even to the best. In whatever profession you are in.
And herein lies the future quandary - are we training the AI to replace us, by helping it perfect and have the guardrails it needs to have in place for 'good enough' Comms?
It ain't learning that from people who a poor writers, poor thinkers... I just ponder this shit a lot right now matey, just as you do!
As nearly always, completely agree chap. I think particularly in some areas of Comms, given the speed at which things are needed/expected, you do need to recognise 'look this is good enough, let's just get on with'.
My personal bugbear here was always press releases. Where with some companies you had to go through about 20 iterations/drafts as multiple people weighed in with a point of view about how it wasn't yet 'perfect'. And 95% of the time, the final version looked more like the one you had originally drafted than the one that got changed and amended by various people who *just had to have an opinion*. Utter waste of everyone's bloody time...
I am going to pick you up on your start point for this article! ;). Because the reason people liked AI poetry was because it was 'too perfect'. So therefore perfection is important to a lay-person, no?
A minor point chap!
TBH my biggest issue with AI at the moment is that it is *nowhere near* perfect (apart from, apparently, on poetry), and that it invents, obfuscates and is downright wrong in some of the stuff it delivers. I have spent (and think everyone in Comms should) many hours explaining and drawing down the guiderails I expect my Gen AI partner to work under. No BS, no making stuff up to 'please me', no un-sourced data points.
This also brings in the 'verification tax' and the ongoing need for humans in the loop. Research shows that for every saving you get from AI, you need to put in at least 40% of that saving back into review, verification and fact-checking. Because if you don't, then you are putting out AI-slop that is potentially 'fake news' and can damage your brand.
Now that is not 'good enough'.
Should we seek 'perfection' in Comms? No of course not, it is actually impossible and all it does is slow down a function that needs to move at speed. But, we must remain wary of what good enough actually is and not let our guard down...
Thanks for the comment, pal, and agree with your points.
I think it’s really important right now that we all know AI’s output is far from perfect, even when prompts have been carefully crafted to be as specific as possible. Taking what AI produces as an additional source of input can certainly be valuable, but using it verbatim and without further verification is the quickest route to a poor result (and a potentially damaging one).
It brings me back to the quote I’ve been using far too much recently, that the people getting most out of generative AI are those who could do the work without it. You have to have the knowledge to spot the inaccuracies, or at least question the information. A poor writer isn’t going to be able to ascertain whether the copy produced by AI is good writing (or is likely to assume it is, simply because it appears to be grammatically correct and not full of spelling mistakes).
Which sort of relates to the point about the poetry. I took that to mean that if, to the inexpert eye, a piece of prose perfectly adheres exactly to what people think a poem is, they’d regard it as a good one. But as we all know (and as Gallagher was getting at) genuine perfection - in any form - is full of imperfections.
Ha! And here is the beautiful irony of AI in a nutshell. "the people getting most out of generative AI are those who could do the work without it."
100% true... for now. And my point about the irony. Because the AI learns most from the people who can work without it, and refines its models based on the humans with the highest skillsets to challenge it and not need it, but recognise the time and productivity gains it can bring even to the best. In whatever profession you are in.
And herein lies the future quandary - are we training the AI to replace us, by helping it perfect and have the guardrails it needs to have in place for 'good enough' Comms?
It ain't learning that from people who a poor writers, poor thinkers... I just ponder this shit a lot right now matey, just as you do!