I want to believe the author, but it's difficult to read that there are hundreds of people 'enthusiastic about AI' and the best they could come up with was... a menu checker and a spam...
I want to believe the author, but it's difficult to read that there are hundreds of people 'enthusiastic about AI' and the best they could come up with was... a menu checker and a spam detector...?
No one thought to maybe, I dunno, connect your calendar and have an auto-OOO reply setup for when you go on holiday?
Or to feed it important emails you've sent, so that it can emulate your voice and you can just ramble at the LLM and it'll put it into a coherent client-friendly email for you?
.... You're telling me no one in their entire IT department is lazy enough to come up with that? Because ours sure did. And I mean that as an absolute compliment - our IT department IMMEDIATELY used LLMs to automate their ticket answering when it's stupid shit. They've got suggested replies, they have the correct teams auto-tagged and they even auto-clock every minute they spend to our time clocking software.
I am beyond curious how the hell OPs industry is database related, and... no one tried anything with... databases?
I got Claude to create an entire year's worth of fake seasonal campaigns across 30+ fake franchises, including fake campaign names for Google Ads, Meta, Pinterest, LinkedIn and generate statistics for each.
1000 rows of usable test data generated within less time than my lunch break. That would've taken me hours and hours to simulate myself from scratch with randomisation and excel.
Sorry for the long ramble, I'm just genuinely scratching my head here. I'm not saying LLMs are the end-all to everything, they're absolutely not and they're also pretty stupid and useless in a lot of applications...
But I have never seen one of my colleagues or my clients ask the LLM how... it's doing... as a feature... So... There's that.
The examples remind me of some complaints I see about Duolingo sentences sometimes. Sure, some sentences seem random sometimes ("The horse ended his relationship with the fish."), which makes...
The examples remind me of some complaints I see about Duolingo sentences sometimes. Sure, some sentences seem random sometimes ("The horse ended his relationship with the fish."), which makes people say that the app is teaching nonsense. It seems like people often don't have vision about how we can extend the things we learn. We're not specifically trying to learn how to talk about inter-species dating habits. If we already know the words for "woman" and "man" well, we might as well drill some animal vocabulary along with verbs and sentence structure.
If the demonstration is to ask the schedule for what's for lunch or interpretation on a simple email, it's to simplify that we can ask questions about a more relevant data set or a more complex document. Today I asked copilot to explain why a certain project was delayed in favor of another project, and it gave a good answer, providing relevant sources from files I have access to for verification (either I didn't attend the meeting when this was announced or I forgot). A coworker sent me a cryptic message with three parts I didn't understand (very brief response, probably because she was busy in a meeting). So, I asked copilot to explain, and it was able to interpret each statement and associate them to internal documents that helped me to learn about a process and advance the project without having to wait for a clarifying response or bog down a busy coworker with dumb questions.
So yeah. To echo, it's far from perfect, but it's also far from useless.
The author doesn't mention it, but I sure hope they're looking for a new job. Don't expect things to ever get better.
Seriously. The opening paragraphs were nothing but "Run from this company ASAP".
I want to believe the author, but it's difficult to read that there are hundreds of people 'enthusiastic about AI' and the best they could come up with was... a menu checker and a spam detector...?
No one thought to maybe, I dunno, connect your calendar and have an auto-OOO reply setup for when you go on holiday?
Or to feed it important emails you've sent, so that it can emulate your voice and you can just ramble at the LLM and it'll put it into a coherent client-friendly email for you?
.... You're telling me no one in their entire IT department is lazy enough to come up with that? Because ours sure did. And I mean that as an absolute compliment - our IT department IMMEDIATELY used LLMs to automate their ticket answering when it's stupid shit. They've got suggested replies, they have the correct teams auto-tagged and they even auto-clock every minute they spend to our time clocking software.
I am beyond curious how the hell OPs industry is database related, and... no one tried anything with... databases?
I got Claude to create an entire year's worth of fake seasonal campaigns across 30+ fake franchises, including fake campaign names for Google Ads, Meta, Pinterest, LinkedIn and generate statistics for each.
1000 rows of usable test data generated within less time than my lunch break. That would've taken me hours and hours to simulate myself from scratch with randomisation and excel.
Sorry for the long ramble, I'm just genuinely scratching my head here. I'm not saying LLMs are the end-all to everything, they're absolutely not and they're also pretty stupid and useless in a lot of applications...
But I have never seen one of my colleagues or my clients ask the LLM how... it's doing... as a feature... So... There's that.
The examples remind me of some complaints I see about Duolingo sentences sometimes. Sure, some sentences seem random sometimes ("The horse ended his relationship with the fish."), which makes people say that the app is teaching nonsense. It seems like people often don't have vision about how we can extend the things we learn. We're not specifically trying to learn how to talk about inter-species dating habits. If we already know the words for "woman" and "man" well, we might as well drill some animal vocabulary along with verbs and sentence structure.
If the demonstration is to ask the schedule for what's for lunch or interpretation on a simple email, it's to simplify that we can ask questions about a more relevant data set or a more complex document. Today I asked copilot to explain why a certain project was delayed in favor of another project, and it gave a good answer, providing relevant sources from files I have access to for verification (either I didn't attend the meeting when this was announced or I forgot). A coworker sent me a cryptic message with three parts I didn't understand (very brief response, probably because she was busy in a meeting). So, I asked copilot to explain, and it was able to interpret each statement and associate them to internal documents that helped me to learn about a process and advance the project without having to wait for a clarifying response or bog down a busy coworker with dumb questions.
So yeah. To echo, it's far from perfect, but it's also far from useless.