Speed of changes re AI at work

[Bunnycat101]

Has anyone else suddenly found their workplace is moving really fast on AI? I feel like I’ve almost been in a mini revolution. 3
months ago it was quite sporadic use with people using copilot as an extended search engine. Now, everyone is using it all the time to help draft content, write emails, get advice etc. The speed of change has been really interesting to me. Culturally there used to be a hesitancy at first that you might be cheating by getting copilot to write a letter or a briefing note whereas now it’s normal practice to use it at least in part. Notes are all done via transcription. I worry that junior roles look pretty vulnerable already. The things that I’d have got a new grad to do like minute taking, summarising documents can be replaced very easily by AI. My own skill set feels under threat as well.

AIBU to ask if others have found the same in their organisations? My other worry is that the school curriculum has not adapted at all. The world of work seems to be changing very rapidly and we’re still drilling 10 year olds on fronted adverbials or setting exams that are based on recall of facts.

I’m a GP and some of my GP colleagues have started using AI. It’s basically a recording device that is on the desk during the consultation (patients have to consent to this first), which then generates consultation notes afterwards.

I hate the way the notes are laid out. It’s done as bullet points, and not in a natural order. It also eliminates the things that aren’t strictly clinically relevant (eg the patient mentioning a forthcoming holiday, or updating on their child’s progress at school - the little things that I talk about with my regular patients).

I find the notes AI generates soulless as well as difficult to interpret.

My workplace has pivoted massively to AI. We’re all being told we have to use it, whether we want to or not. It’s not a good thing

Just you wait till it starts inventing diseases and parts of the body. Then the fun really begins.

Adjunctive AI is on its way which is meant to give a proper overview of sources and their reliability so that may help

Or making up clinical codes 😱

I also think AI is making Alexa look really stupid, I wonder when Alexa will incorporate ChatGPT or whatever the equivalent is

That doesn't stop it being a steaming pile of dog shit though.

Indeed.

I was first alerted in 2023 when my boss - quite a technophile - signed us up for a subscription to OpenAI (ChatGPT). Withing a few hours we realised it would simply make shit up when pushed into a corner. This is simply inventing magic commands you can use in Linux or undocumented "features" that didn't exist.

When you point that out it just goes "Gosh ! You're right". And then it's a 50/50 as to whether it will simply rephrase the incorrect answer and gaslight you into thinking it's correct. Or just advance with a less shit answer.

I have already seen the future of mankind. And it is going to be a perpetual treadmill of having to check "AI" output. With the added excitement that the people who can actually do that checking will have been the first to be fired.

Already there is a subtle schism opening up between people who are able to spot "AI" and act accordingly, and people who simply lack the ability to do so. Which already includes our politicians and civil servants.

We do indeed live in "interesting times".

That's like saying excel is dog shit if you want to write a paper. AI has its limitations but it's an excellent tool if used properly.

Digital Assistants weren't sold as "AI" originally.

Google Assistant has been replaced by Gemini. And even if a ChatGPT-Alexa Frankenstein is 10x less crap, I'd still look out the window after it told me the weather.

ChatGPT has been providing all source urls if you ask it to, including open source research papers, ie those that are not behind academic paywalls and subscriptions. You can also ask it for evaluative comments on the pros and cons of different schools of thought, so you can compare those ideas to your own beliefs.

people just need to get tough on AI and demand high quality information rather than placing all their trust and confidence in it. Unfortunately nobody has ever been given any guidance on proper use, so some will use it more effectively than other. How many times do we hear, well AI says .... as if it's the font of all knowledge.

its like a fantasist friend who tells you all sorts of nonsense, if you don't ask the challenging, searching questions they get away with it!

Oh quite. I'm quite happy to use a hammer to drive a nail in.

What I am not doing is asking my hammer to come up with new ways to layout the lounge.

ChatGPT has been providing all source urls if you ask it to,

Yes, but it can invent them too. So you need to actually follow and check them. (Unless of course you setup an "AI" bot to do that for you). And by the time you have done that - guess what ? You've spent not saved time.

@SerendipityJane the AI models released by ChatGPT in 2023 bear no resemblance to what it now offers.

As recently as 2024, they were very inaccurate but this year they've evolved significantly (See GPT-5 for more details) A day is a long time in the life of an AI LLM 😊

I rate ChatGPT highly against Copilot. However. MS Designer is great for image creation (although it is renowned for being incapable of spelling even if you purposefully give it a word, it cannot render it accurate on an image! It's creating rage on the MS forum).

Well of course not. But ai is going to become more and more ubiquitous so the choice is to get on board and learn how to make the most of it or be left behind moaning about it. I'd rather hone my skills, look at the different strengths and weaknesses of each option and use it to improve my productivity.

I've been following "AI" since the 80s when I did my degree. I still wouldn't trust it with my keys at a party.

E2A: To be fair when I saw "Watson" at IBMs Hursley labs, I was impressed. However that was doing a very specific task.

Work are doing training on how to use it as they want it used as much as possible

Then if you don't make time to check the sources you can't blame AI for that.

I normally limit the sources it provides and if I don't agree with its assertions I can check back to source. It does save time, it saves the time to do the initial research. The time you save on the mundane elements, you invest in other priorities such as refining and finessing the content into what you need it to be for the purpose you need to apply it (academic essay writing, professional CPD, corporate report or research for a project.). Some of what you find is throw-away, you can't expect 100% quality first time, but it does advance your thinking forward, rather than sitting there, puzzling it out and making no progress, or having to start things from scratch each time.

People are also been employed into roles not right because they've used AI for the job. My manager shouldn't have got her job she's producing a strategy and it's all been done by chat gpt...

I think if you work on the premise that what it gives you is 80% correct and then use your time to correct and amend the 20% rather than thinking it will give you a perfect document. It’s still quicker than starting from scratch!

Me too! I was taught how to build simulations of neural networks and the difference between supervised and unsupervised training models in excel spreadsheets. Great fun. Plus the theories of AI, Turin test, decision trees etc. The advancements between the 1950s until about 3 years ago were relatively unknown (apart from virtual reality, gaming). the AI we see today on every device, every web browser, took all those decades to become mainstream, has emerged in the past 3-5 years.

Which is a wider commentary on our education and qualification system ....

A growth area where jobs will come, is towards AI governance, controlled ethical use, which is something I'm passionate about. Ensuring AI is used to benefit humankind and not strip away our rights.

for example there is a high profile movement lobbying global governments against them allowing big Tech corporates to ingest all web based content into its large language models despite a proportion of that content being protected under copyright.

The original artists, composers, writers are not being asked for their permission, because governments see it as expedient to pander to Big Techs to satisfy their prosperity agenda. No thought to the fact those individual creators, with all their talent and creativity, are having their livelihood compromised by AI stealing their stock-in-trade and using it to create images, generate music "in the style of Elton John, Madonna" or a writing style based on Harry Potter, when the creators are still alive and they should be consulted and compensated. Daylight robbery.

ETA those who have made their fortunes from their talent like Elton and Madonna, are now fighting for the rights of up and coming artistes early in their performing career.

Yes, my org are pushing it big style, started 3-6 months ago. We have Co-Pilot embedded into Teams, Office 360 etc, and are getting the hard word from management to use the transcription functionality for all formal meetings and for drafting docs etc. I use it occasionally, but mostly just mess around with it... will probably have to get with the program at some point I suppose!

In my industry it’s moving incredibly fast - all the top firms want to be the first to be using it properly, and my firm even developed its own generative AI tool. I’m just going with the flow - might as well work with it rather than against !