So…AI.
Twice in the past week people have shared personal experiences about AI being used (or allegedly being used) during a remote job interview.
And this got me thinking. My knee jerk response is that is not okay. It is deceitful, undermines the process and is grossly unfair to the other candidates. Basically, it’s not cool. And in #GMC speak – it feels to me a massive probity issue.
The excitement about AI across a range of industries is palpable. Working in healthcare and health tech as I do, I too am excited by what is promised. There is a huge amount potential for improving the way we do things, doing more things and generally trying to making things better.
One of the other things I enjoy doing is putting pen to paper, and I definitely use AI (hotpot.ai to be precise) to draw me pretty but random pictures for when I post things. And one of the main places I post things is on LinkedIn.
LinkedIn does many wonderful things. It connects people, allows content to be shared, creates, sustains and enhances entire careers. But anyone who uses LinkedIn will be able to see the teasing button offering you the chance to write a post with AI. With algorithm clues suggesting that regular posting “enhances your profile”, this feels too good an opportunity to turn down.
Or does it?
When thinking about this, I did some thinking and reading about digital carbon footprints. Which led me down various rabbit holes. Rather than teaching you to suck eggs, and talk about the environmental impact of #GenAI, I wanted to share something that made me smile. Ecoping published an article about the environmental impact of email traffic [spoiler alert – send less, don’t use a signature, use links not attachments]. But the nugget of information in this article was the phrase “around 20 farts worth of CO2e per email”. Which is frankly delightful and worth digging deeper into.
My search engine algorithm then took a pretty random segway.
In 1861, a person called Ruge put a glass tube attached so someone’s through a hole in a chair to a collection vessel) and found that carbon dioxide was present (along with hydrogen and methane). This is very niche science, but subsequent research has built on this and we now know that by volume approximately 1/3 of fart volumes (34.7% +/- 14.7) are CO2.
Ruge, E. (1861). Beitrag zur kennuness der darmgase. Sitzber Kaiser Aka, 44, 739–762.
The methodological limitations here should be highlighted. The subjects were fed on a “standard-gas inducing regime” of pinto beans and lactulose to increase volume of flatulence and make measurement easier. Nonetheless, the volumes were found to vary with each event between 5 and 375 ml).
Another study, which for some of the study also used a gas inducing regime reported a median volume of 90ml of episode (33-125mls). So, has this got to do with LinkedIn?
Matt Tutt is a SEO and PPC consultant from the UK, and he reported that LinkedIn AI had invented a study on carbon emissions - specifically that the Carbon Trust had done a study about the carbon footprint of smartphones.
It turns out that this was an “AI collaborative article” he had stumbled across (https://www.linkedin.com/help/linkedin/answer/a1413111). This is where users are encouraged to add answers to AI-driven questions, which are then augmented presumably to improve readability and collate the answers given into a readable article. Amazing use of AI. Both in generating the questions, and sieving through the answers.
But as Ryan Law points out, the "top voices" badging system incentivises people to answer as many questions as possible, with pithy, concrete-sounding advice. there is no reward nuance. Irrespective of the validity of the question and the validity of the answers given.
A chap called Iain MacDougall did some maths estimates that each LinkedIn connection gives us an estimate of 0.009 kg CO2e.
Brett Landrum did some as well, and reckons 479,062 kg CO₂ is created from AI-prompted LinkedIn posts.
Now time for my bad maths.
If we assume based on the studies above, that 34.7% of farts are CO2, and the median volume is 90ml “per episode”, given that CO2 has a molar mass of 44.009 g/mol, this means that each fart produces a median mass of CO2 of 0.006g CO2 per fart (at RTP).
And this means, that each Connection on LinkedIn equates to 0.009 kg 1500 farts, and the AI-generated content on LinkedIn equates to nearly 80 million farts - mind-boggling 79,843,666 to be precise (if any of this is precise).
There is an interesting side-angle to this. Published in Nature, having undergone peer-review was a study that reported that “AI systems emit between 130 and 1500 times less CO2e per page of text generated compared to human writers”.
Does that mean we should “let the AI” do it for us?
Jonathan Gillham at Originality.ai estimates that 54% of LinkedIn posts analysed were likely-AI generated. And since Chat-GPT was released there has been an unsurprising huge surge in both amount of AI content, and the length of LinkedIn posts. I have no doubt that this applies to other online sites too.
If we circle back to where we started, writing stuff on LinkedIn using AI is not the same as using AI to answer interview questions. It might make it read better, and it might make a better picture for me than I would myself.
I would argue that just because Chat-GPT might use less CO2 than me doesn’t mean I should let it do the whole thing for me. Anyone can churn out AI-generated content and nudge their profile upwards. The amount of online is booming. But there is a cost. One is that for anyone producing content, the temptation is that it becomes about churning out quantity and not quality. Not just in terms of the ethos, but also in the fact that AI Posts receive 45% less engagement.
There is also the issue of integrity. Not just in job interviews, but I don’t think people should pass off other entities work as their own. If you are using AI, declare it. Be honest. Show you are forward thinking. Whether or not people care or not, is a different matter.
There is also the environmental impact of all of this – or the fart-equivalent ratio, because of the amount of content constantly being created with a few clicks, rather than carefully, curated and well thought-out creative work.
And as long as the algorithms do not promote quality, the rewards will always on quantity, and if my bad maths is anything to go by, this is capable of producing a lot of hot air [pun 100% intended].
Words by me. Picture by hotpot.ai (if it uploads...)
Comments