The modern office - where we read out what copilot tells us to think

[LorenzoCalzone]

I attended a meeting today with senior people in my company. People were clearly just reading out their AI drafted thought pieces. I thought "Is this the future of office work now? Listening to someone's copilot results, maybe countering it with another AI generated point of view". These perfectly scripted perspectives are all over linkedin too.

Of course I use it too. The toothpaste is out of the tube!

I wonder if mathematicians felt like this when the pocket calculator came along. Its cheating!

This 100% did not happen, no one makes eye contact online it's impossible the camera is not in someone's eye.

Secondly if either one of the candidates is neuro diverse and calls this out your manager will be facing issues.

Sorry I'm calling fabrication.

I used AI to help with my CV but I still re-wrote it so that it was in coherent English, and did use it to write some bs personal development targets for work which are vague enough so I can tick them off without putting in any extra effort.

I tried chatGP at home to redesign an area of my garden and it was useless. Gave me a plan based on a supposedly square grid made of rectangles with measurements that were totally made up.

If you can use AI, you should use AI.
Companies want people to be AI savvy and know where and when to use it.
You can always tweak the results afterwards.
Go with the flow.

I don’t recognise this. Civil Service and attend loads of meetings. Copilot is brilliant for summarising meetings. Doing simple data analysis. Finding emails about x subject from two years ago. And it’s great at producing a summary of what I’ve been doing for the performance meetings.
Mainly causing problems in recruitment where the sift is harder and the interviews more stilted. We get people to confirm they are not using AI at the start of the interview now. Can tell when they’re reading from AI generated notes though.
But I work in an operational area where you actually have to do the thing. Not just talk about it.

@maxslice I don't think the poets are in any danger from AI currently. It's efforts at poetry are absolutely hilarious 😂.

I work in journalism. We don’t use it at all. Our readers would spot it straight away and rightly rip us to pieces in the Facebook comments.

I think the tech people use it for some the website functionality across the company and some other bits, but as an industry we’re largely avoiding it.

AI summaries on Google, including the use of our copyrighted work that’s behind paywalls, is having a catastrophic effect on local journalism so most reporters won’t use it as a point of principle.

The biggest effect it’s having on my day to day work is the number of AI-generated ‘legal’ complaints we’re getting. Pages and pages of pseudo-legal nonsense that we now largely ignore because it’s complete waffle. AI is absolutely dreadful at accurately interpreting UK law.

No doubt. And to be fair, there are reams of terrible and embarrassing poetry written by actual people. Still, every AI document I’ve ever read seems…thin, hollow. It may be factually accurate (or sometimes not) but for the alert reader they are without depth or nuance or individual voice. Elements that make reading a text worthwhile.

Yes, but mostly meetings are wasting time with people talking nonsense anyway.

I find AI very helpful indeed at work, however we should still use our own brains in considering what is and isn't appropriate use of it.

One of the things that pushed me towards leaving my old job was when my boss (who had no events management experience) told me (10+ years events management experience) to follow her template for planning events rather than use my own. However, hers had very obviously been churned out by Chat GPT.

It was insulting. I felt like I was being managed by a robot with no real understanding of what it was doing, with my boss mindlessly relaying the robot's instructions to me.

What was the point of employing someone with relevant experience if I was then not expected to draw on that experience or use my own judgement but instead mindlessly follow instructions from ChatGPT?

Why?

It's horrific

I am going through a massive fight right now where a charitable organdisiron I was attending as a service user (and later volunteering having told I would get paid soo, stopped when it was apparent I would not be given any paid role despite praise after praise and lots of roles available) used AI to understand my little known mental health condition.

AI told them it caused me to become delusional and psychotic so in the middle of a safeguarding enquiry, despite knowing me for seven years, the absolute fuckers went and told a mental health trust that I was suicidal when I wasn't, and that my illness made me delusional and psychotic. We were in the middle of our foster care application which has been thrown out and I am a witness in a court case that will be able to access my records.

I now have these falsehoods, along with many other blatant lies on my permenant medical records and it's a fucking slog to get them removed.

Because in the right circumstances it can save you time and/or boost your output.
This is why companies are paying AI subscriptions.
Why spend 4 hours creating a document when AI can do it in 5 minutes and you can spend 30 mins enhancing it?

AI is everywhere. It is already part of interview questions ‘Explain how you used AI to a positive effect’ etc etc.
Not using AI puts you at a disadvantage against someone who is proficient.

Would you not use a calculator to work out a maths problem?

Agree with this!

I think there’s a huge lack of understanding on how to use AI properly, that’s why people get rubbish results. Rubbish in, rubbish out.

What a nightmare, that's awful :(

I’m a primary school teacher and I’m horrified by how SLT are touting it as the solution to all our problems. Less experienced teachers are using it for lesson planning, writing models to share with children, composing reading comprehension questions etc. The results are abysmal but SLT just can’t see it.

A calculator doesn't use massive amounts of water, contributing to the environmental disaster we're experiencing so the two things are hardly comparable. Plus the issue with AI hallucinating things so you can't trust anything it does anyway so you might as well do it yourself.

I use AI, but as a starting point, or to tweak things. I cannot stand when people use it for work purposes indiscriminately. I recently received a pretty much essay length email from a client pushing back on various points on a legal document. It probably took them 30 seconds to churn it out in AI, and about an hour for me to wade through and refute all the inaccuracies.

As for AI generated newsletters, sometimes with the creator proudly announcing they have used an LLM. Why would I read something someone has put pretty much no thought into.

I do think that AI can be useful when used by someone intelligent but so many people treat it as some kind of god that knows everything and it really doesn’t.

I know

The really stupid thing is if you put does X condition caused psychosis and delusions AI will actually tell you very clearly it does not. So they didn't even take the time to double check (using AI) the information it was spewing out to them.

This is, in her other role, the head of safeguarding for a secondary school. I shit you not, those poor kids.

A data center doesn’t use massive amounts of water either, they are closed loop, like a car radiator or the home central heating.
Streaming videos is comparable to AI processing, no one is calling for a ban on YouTube or TikTok though.
AI hallucinating is when you ask it to read the internet for stuff. If you are giving the information to AI yourself and asking it to create a document how can it hallucinate?

Hallucinations are real, but if you know how to use AI that is cut down dramatically.
There are no solid arguments not to use AI in most office scenarios. It’s usually older people who struggle to grasp new technology complaining.

I work in Pharma, the AI and LLM used by the scientists and researchers is dramatically reducing our time to market, so actually improving and saving lives.

If I put 2+2 in a calculator I know with 99.9999% confidence I'm going to get 4.

If I ask a LLM AI a question then its accuracy rate is more like 80%.

Plus a pocket calculator will use a few milliwatts of energy to give me the right answer whereas an AI will use thousands of times more to give me the wrong one.

lol by which he means “if you’re not capable of original thought, and can’t explain your experience coherently, you’re no good to us”.

AI is perfectly capable of inaccurately repeating information you've given it, let alone inaccurately summarising or analysing. I asked it to create an infographic for me recently and ran out of attempts trying to get it to correct itself. Simple task, but I don't have graphic design skills and it was a choice of AI or Microsoft Word. It did not need to do anything except change the format and make it more attractive to read. About eight bullet points of info. It could not. Including with me asking it to generate its own prompt to ensure that it was doing the job correctly.

I frankly never thought I'd find anything that made me think "oh, this would be easier to do by drawing a picture in Word".

I use AI at work, it has its place, but the readily available models are next to useless for any tasks where accuracy is important.

I do like it for Excel - both for creating formulas and for understanding other people's.

The internet got off to a shaky start as well, and look where we are now.

Setting the environental impact aside...

Lots of AI tech companies have made a lot of money selling the narrative that AI will change the world and its important to invest to avoid being less behind; and most places dont have the knowledge to judge for themselves.

AI as it is today has some uses undoubedtly- doing stuff looking across data in seconds (as long as the data is accurate...), sunmarising meeting transcripts etc, but the rest is corporate slop, you can tell an AI email or speech and its never better than a human on top of being depressing people have so quickly defaulted to this crap.

I think it will develop, but running before it can walk, scary how unaware some companies are of the data processing element too.