About 75,000 Deloitte staff have been given access to a generative artificial intelligence chatbot to create PowerPoint presentations and write emails and code in an attempt to boost productivity::About 75,000 staff to be given access to ‘PairD’ tool with advice to validate ‘accuracy and completeness’
Not sure how increasing the number of emails everyone has to read and PowerPoint presentations everyone has to sit through will boost productivity. 🙄
That’s the trick, you don’t read them.
Person A creates an email with AI, attaches a ppt created with AI, and send it to person B.
Person B gets AI to summarise the email and ppt, and create a response.
Repeat.
It’s an ouroboros of shit.
The game of telephone has itself become mutated. I can’t tell if that’s funny or sad.
I know you’re (partially) joking but, the way I’ve seen my coworkers swoon over chatgpt for code snippets, I’d be shocked if thousands don’t already use it for summarizing emails.
Yeah, it’s really not good, especially because the people I’ve seen use this stuff the most are the ones who would rarely ever question the quality of the work it does or bother to make sure it’s accurate.
And because managers and business owners see “productivity” increase, they expect that that’s the new standard, and AI will essentially just have made doing quality work harder for everyone involved.
Exactly. With “AIs”, people need to remember that coherence ≠ cogency.
For the common user, it’s almost like that one quote that gets tossed around equating sufficiently-advanced technology and magic. The less-informed (those who truly believe the “AI” moniker that has been tacked on these things) read the output and it’s astonishingly readable, so they assume that since the nerdy experts built this thing, it must be correct.
Then as you said, managers/execs don’t make any effort to understand the lower echelon because capitalist Darwinism has caused them to chase numbers instead of, I dunno, connecting with the humans who actually keep their business running. LLMs are just their newest toy for slashing the shit out of wages on their P&L for the next several quarters.
Hah, I didn’t mean to sound like a crazy person, and I definitely don’t have a solution for capitalism, but that’s just my frustrated two cents.
Job applications these days.
Ai writes the cover letter and resume.
Ai reviews it and confirms/rejects.
I really hope this goes through, because the pile of dogshit they sold my municipal government can’t fairly be called code.
It’s astonishing. I will never ever deal with that company if I have the opportunity in the future because of how utterly incompetent they were creating a “microservices architecture” for us which is in effect a series of AWS Lambdas running random Python that do not coordinate, suffer from race conditions, and in many places do not do the thing in the first place.
It’s the second worst code I’ve ever seen. The first was home grown at another public sector job.
The cost for this shit that could have been replaced by a simple Flask app? Or even a well structured, in-lined Python script? $1.5 million. Ongoing costs in the hundreds of thousands per year.
This should be a $20-50/month EC2 instance.
Absolute garbage.
No, I am not underestimating the complexity requirements nor the communication overhead/competing stakeholder interests and all other manner of externalities.
Deloitte belongs in the dust heap.
BUT THEY CHECKED ALL THE BUZZWORDS
Serverless. Lambda. Cloud. Python. Microservices. Blahblah.
The number of companies I’ve contracted for that just try to hit all the buzzwords and don’t actually architect for their need, nor simplicity and maintenance is ridiculous.
Example for one place I’m consulting for right now…the average simple change like enabling/disabling a button based on a required field in a form typically takes touching 10-14 different files.
Why? Because checkboxes need to be wrapped library components with their own NPM repo. They must also be super dynamic and flexible, so tons of inputs and outputs. They must also be themeable. Anywhere that uses them must be an isolated reusable form. Consumers of that form have logic to decide whether or not that form displays a certain way for the context. That component must also bubble up it’s state and the form state to its parent. It’s parent is a micro frontend (fuck micro frontends). That must communicate it’s state to a composable parent. That parent must rely on stupid NgRx/Redux to dictate button state.
I hate it all. I feel like engineers these days are convoluted just to be convoluted.
The new business model will be to hire humans to do the work to create enough data until the job can be automated by another facet of the AI.
AI equivalent of training your Offshore replacements
Exactly, this makes me very anxious. Feeling that we’re just cutting the branch we are sit on …
“Please send an email to IT explaining why our server is offline”
IT responds “Why are you attempting to put egg salad in your server?”
Yeah this is a great plan to ‘boost’ productivity…
This sounds like reddit using bots to boost site activity
Then they will have bots to read emails and presentations and at the end it will be bots emailing other bots and bots presenting to each other. Bots all the way down.
I’m not surprised. I’m sure there are many companies out there who are beginning to use generative AI
Doesn’t everyone with O365 have access to this now?
