In this blog, I won’t be expressing discontent with Amazon as
a whole; they have done a pretty stellar job of regularly discounting my novel without
affecting me personally. No complaints, there. But I am increasingly concerned
with their continued over-reliance on Ai for generating inaccurate summaries about
my product, especially when the margin for making atrocious errors is clear… errors
that both harmful to your business, and misleading to the customer. This is not
exclusive to Amazon either, with other product sites using their own version of
Ai - one of which states an inference - amounting to my horror novel being suitable for children because… it features children.
And if you don’t believe me, read it right here.
Good grief. If Ai just takes a brief look at the front
cover, it will realise this is horror, not the Hungry Caterpillar.
Amazon’s latest Ai generated summary of my novel (what amounts
to the bread and butter for an author) is, in my case, not only catastrophically
inaccurate, but bordering on a lie. It is a composite of UK reviews, that it has
resoundingly misinterpreted/taken out of context… while creating a brand new
expression to support its own value-judgement! I’m not exaggerating either (unlike
the latter). We will get onto “Fabulously Spoo” as a negatively interpretated
Ai comment shortly…
Above all, trying to contact a human who can help through either
KDP select, Amazon Author Central, customer service etc, to draw attention
to/rectify the issue, is nigh on impossible. Round and round in circles… click,
click, click… and then, back to the beginning. Sure, you can post on the Amazon
Author community forum, but all who raise similar queries to my own seems to receive
the proverbial GET-OVER-IT-GUY as their first reply; someone who purposefully
ignores a well-conceived query and goes straight in waving a can of gasoline. Not
even slightly helpful. It’s all really… rather weird. And speaking of weird,
what exactly do I mean by hallucinations in this blog title? (And yes, you did
read that correctly; I purposefully intended to incorporate this.)
So, after going round and round in circles, for the first
time ever, I logged onto Chat GTP earlier this morning, and asked it some
questions. I’ll stop writing for a bit, and just post the summaries. (Apologies
for the typos in the Q’s asked; I was getting increasingly frustrated.) These
aren’t my responses btw, these are Chat GTP’s…
Any images you see of our own planet earth are composite
images, created from numerous, blended digital photos to produce an enhanced
panorama. It’s important you have plenty to make a realistic picture, otherwise
you’re filling in the gaps with your imagination. And here’s the point, you
must have comprehensive evidence at hand, and the ability to interpret
correctly (and with clarity), otherwise you’re going full Black Adder – pretending
to paint no man’s land - in that episode of Black Adder goes forth.
Darling:
Are you sure this is what you saw Blackadder?
Blackadder:
Absolutely. I mean there may have been a few more armament factories, and not
quite as many elephants, but...
I have no problem with Ai, if it is frequently monitored,
quality-tested, and deemed appropriate-and fit for platform. I am no luddite;
but there’s somewhat of an irony in the example I’m using here: no man’s
land. I haven’t asked Ai to pass judgement on my Amazon page, and yet it
does, and in an increasingly confounding way. I need it to enhance my product,
not misconstrue the content and deliver to the masses as truth. For now,
I am happy to consign Ai into the no man’s land of my personal preferences/influence
on my life, at least until it’s fully formed, and a true reflection of what it was
intended to be. I wouldn’t purposefully head for a blood test, and request a
trainee nurse to fruitlessly search for a vein in my arm and then repeatedly attempt
to insert a cannula… because that’s damn painful: I would request a
professional who knows what they’re doing.
My novel’s Ai update yesterday fixed upon (and then added) a
single, detrimental phrase to the overall headline… citing evidence from a
glowing five-star customer review. Not only did it do this, based upon a single
word (lull) - taken out of context - but it wholly ignored the rest of one my
favourite reviews, using a five letter word as a fulcrum to embellish the entire
summary, and produce the strangest overview.
But it doesn’t end there. In terms of my book’s readability,
it gets weirder. Ai stated there are 3 positive comments, and one negative. All
four, are in fact, positive… but for some reason, ‘fabulously spooky’ is a negative
when it comes to readability, because Ai has only pulled two thirds of the
letters from the word spooky, rendering us with a brand new expression, “Fabulously
spoo”
I mean, what the Dicky Davies?
(Please Ai, don’t read and attempt to interpret that
expression!)
Does it believe the customer (again, five star) actually meant
to write: spew, poo, or poos? This novel… fabulously poos! Or is it, loose
spoon, louse poo… or sly spoonful (which personally, I’d take as a compliment.)
When all’s said and done, I can’t remotely fathom any of this. 😊
No data pattern, protocol, or algorithm could produce anything this bizarre. This
could well be an example of what Chat GTP told me about an Ai hallucination.
So, is it time to do something?
Is it time to go full John Connor, and what does that realistically
mean? 😉
People sit back, and don’t do anything. This isn’t a
criticism – in fact - its human nature, in our increasingly busy existence, to seek
the path of least resistance, and wait for others to get involved. It’s far
more comfortable to remain impassive, and wait… and let others shoulder the
weight. Recently, on our street, a business decided it was going to extend its
premises, which would cause significant upheaval, noise, pollution - as well as
increased footfall - to all living in this quiet Victorian cul-de-sac. The
advocates got wind of these proposals, combined their efforts, and stopped the
planned extension in its tracks via a cogent and sustained appeal to the
council. I’m very proud of those well thought-through letters. Others on our
street, closer to the intended site, who would experience the same negative
effects… did nothing. I can guarantee their exasperation if this thing went
ahead. They had a chance to make their voice heard with a single paragraph… but
didn’t.
So, here’s what I’ll do (as part of this whole John Connor
trope). I will push with all my might to get answers. I will write, advocate,
push, push, push… until I can I get these answers. I will seek to see Ai
summaries are changed/restricted/refined to be more reflective of their product.
I am one voice, but I will try. I will try for you, good author, as much as I
will try for myself. Feel free to share, comment, guide people to this blog,
and above all, become a second, third, fourth voice etc, in breaking through the
complex tiers of complexity to reach the people who matter, and let this ring
in their ears, until something is done about this.
It’s not only a ridiculous scenario, it’s fundamentally unfair.
JSC
No comments:
Post a Comment