Honestly, what’s there now does better note-taking and data building than humans do. You have a shit ton of gen xers that wont interact with it or know how too. So its not being utilized except where its obvious to them “lets use it in our call center” … when there are a lot of billing and data tasks that its even better at. Ive been using a lot of Power Automate and agents in my flows both for turning certain emails and PDFs into data fields. It’s working, not without error, but so far less than what id get from most employees.
It never gets better though, it never understands what you’re or it is doing. So it’s pretty limited in scope. But I found that if you break down tasks into steps then basically write a full detailed how-to for it to follow it does alright and if you do that over and over you get some big efficiency gains on some tedious task sets. But you end up spending an hour or so a week after maintaining the wires between everything and updating the prompts, the JSON etc.
Its not a revolution but it does enable automating parts of once tedious Inference tasks and cuts down on the need for expensive bespoke SaaS applications to do the same.
HughPajooped on
It’s great for a lot of things I do in my job. From finding specifications of things on the web to something like a formula issue in Excel.
probablymagic on
If AI doesn’t get much better than this it will be wildly successful. It’s already really useful fit many tasks where it has very little adoption amongst knowledge workers, and is already indispensable for certain tasks, eg coding. We’ve barely begun the implementation phase, which [will take decades](https://www.amazon.com/Technological-Revolutions-Financial-Capital-Dynamics/dp/1843763311).
These tools aren’t very useful to people who generate words for a living themselves, so is natural New Yorker writers aren’t very impressed. Knowledge workers who aren’t professional writers should take this tech much more seriously.
As an aside, the idea of quoting Ed Zitron as an expert on technology is absolutely insane. When the pickings are that slim, your thesis is certainly weak.
amperage3164 on
Uh, what? How can anyone compare GPT-5 and 4 side-by-side and conclude AI progress has “stalled”? The chain-of-thought stuff is a game changer for many tasks.
4 Comments
Honestly, what’s there now does better note-taking and data building than humans do. You have a shit ton of gen xers that wont interact with it or know how too. So its not being utilized except where its obvious to them “lets use it in our call center” … when there are a lot of billing and data tasks that its even better at. Ive been using a lot of Power Automate and agents in my flows both for turning certain emails and PDFs into data fields. It’s working, not without error, but so far less than what id get from most employees.
It never gets better though, it never understands what you’re or it is doing. So it’s pretty limited in scope. But I found that if you break down tasks into steps then basically write a full detailed how-to for it to follow it does alright and if you do that over and over you get some big efficiency gains on some tedious task sets. But you end up spending an hour or so a week after maintaining the wires between everything and updating the prompts, the JSON etc.
Its not a revolution but it does enable automating parts of once tedious Inference tasks and cuts down on the need for expensive bespoke SaaS applications to do the same.
It’s great for a lot of things I do in my job. From finding specifications of things on the web to something like a formula issue in Excel.
If AI doesn’t get much better than this it will be wildly successful. It’s already really useful fit many tasks where it has very little adoption amongst knowledge workers, and is already indispensable for certain tasks, eg coding. We’ve barely begun the implementation phase, which [will take decades](https://www.amazon.com/Technological-Revolutions-Financial-Capital-Dynamics/dp/1843763311).
These tools aren’t very useful to people who generate words for a living themselves, so is natural New Yorker writers aren’t very impressed. Knowledge workers who aren’t professional writers should take this tech much more seriously.
As an aside, the idea of quoting Ed Zitron as an expert on technology is absolutely insane. When the pickings are that slim, your thesis is certainly weak.
Uh, what? How can anyone compare GPT-5 and 4 side-by-side and conclude AI progress has “stalled”? The chain-of-thought stuff is a game changer for many tasks.