AI seems like a cult

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:This will be eye opening- try to think of 10 jobs/professions that won’t be either entirely eliminated or significantly impacted (to the point of dramatic RIFs) in the next 10 years.

It’s harder than you think.

Any of the skilled trades (plumbing, electrical, HVAC, carpentry, fabrication, maintenance/engineering) automotive/vehicle maintenance/repair, firefighting, policing, emergency medical response, commercial fishing, oil and gas drilling….


I think most traditional white collar or professional jobs will be eliminated by AI, but jobs that require human dexterity and problem solving in endlessly dissimilar scenarios will continue to be dominated by humans.


Bad luck for lawyers, doctors and accountants. Good luck for people who turn wrenches, screwdrivers, or fix things.

Speaking for lawyers and doctors (myself and family members), this is laughable. Lawyers are getting sanctioned for using AI.

Smdh

Lawyers will be one of the most replaced fields by A.I. as an A.I. program hooked to a laptop in the courtroom could monitor proceedings and review millions of cases in a microsecond to cite previous case law and raise objections before the human prosecution could even finish their leading statements.

Same for non-criminal lawyers as well, will be almost an extinction event.

[Anonymous]

Anonymous wrote:

Anonymous wrote:New-generation versions of AI are much better than the ones of even a year ago. But ...aren't the hallucinations getting worse? That was the gist of the recent NY Times article.

Yes, the new NYT article said that hallucinations have gone from 1-2% to up to 50% or more from the new gen AI.

That doesn't even include the irrelevant information or the incorrect-but-also-bland tone that AI does when it is correct or not hallucinating.

Like Kerouac said in 1958, "Smart went crazy!"

[Anonymous]

AI can save time with programming in terms of getting an MVP out since it helps avoid a lot of boilerplate stuff, but once you’re trying to build a complex product I’ve found its utility to be pretty limited since AI isn’t very detail oriented and it takes more time to prompt it for minute changes in the code than going in yourself and doing it right the first time.

[Anonymous]

This old timey radio show was on WAMU
I like listening to old radio programs broadcast on a Sunday night on my local npr station.

Right now I am listening to:

“With Folded Hands”
The Science Fiction Hall Of Fame story by Jack Williamson that’s been called “the best story written about robots, ever.”
(Original air date April 15, 1950. NBC network. Running time 29:25)

Highly recommend
https://open.spotify.com/episode/4whsuxmioQp75Dlc26dC83?si=bUKhUehiTPKeFr4U5pVMLQ

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

I appreciate the concern! Luckily, some of us have mastered the delicate art of using tools and critical thinking. It’s a niche skill set, I know.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

While it's not that impressive at the moment, about on par with a 100 I.Q. college educated human, when you compare it to what they were like 5 years ago, it's quite amazing how rapidly they are coming along and improving in their abilities.

[Anonymous]

Those of you who are using it for work tasks like emails and reports with some success, is there a program who like best? I think it’s a tool that I agree will probably separate the employee from the unemployed pretty soon. People have to be able to work quickly and efficiently. Of course they also have to be smart enough to edit/revise what AI gives them! But all the naysayers here sound, I don’t know, old! Thanks…

[Anonymous]

Anonymous wrote:Those of you who are using it for work tasks like emails and reports with some success, is there a program who like best? I think it’s a tool that I agree will probably separate the employee from the unemployed pretty soon. People have to be able to work quickly and efficiently. Of course they also have to be smart enough to edit/revise what AI gives them! But all the naysayers here sound, I don’t know, old! Thanks…

I use ChatGPT for 80% of tasks. Great for things like emails.

If you need better writing, Claude tends to be a little better.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

I appreciate the concern! Luckily, some of us have mastered the delicate art of using tools and critical thinking. It’s a niche skill set, I know.

I went to an official Microsoft CoPilot training session on how to prompt AI.

They said you should include in the prompt things like "do not make up any of the information if you can't find an answer". In other words you can reduce hallucinating by telling the AI not to hallucinate. I didn't find that very comforting. But telling y'all now so it can be part of your elite niche skillset.

[Anonymous]

Anonymous wrote:Those of you who are using it for work tasks like emails and reports with some success, is there a program who like best? I think it’s a tool that I agree will probably separate the employee from the unemployed pretty soon. People have to be able to work quickly and efficiently. Of course they also have to be smart enough to edit/revise what AI gives them! But all the naysayers here sound, I don’t know, old! Thanks…

Re: old and naysayers...there are reasons why age and wisdom are correlated.

We have seen some job disruption trends before.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

I appreciate the concern! Luckily, some of us have mastered the delicate art of using tools and critical thinking. It’s a niche skill set, I know.

You can’t even write your own memos, dummy.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

While it's not that impressive at the moment, about on par with a 100 I.Q. college educated human, when you compare it to what they were like 5 years ago, it's quite amazing how rapidly they are coming along and improving in their abilities.

They’re just models, fool. The same as models we’ve been using for decades without the “AI” branding.

There is nothing “intelligent” about any of this.

[Anonymous]

Anonymous wrote:Those of you who are using it for work tasks like emails and reports with some success, is there a program who like best? I think it’s a tool that I agree will probably separate the employee from the unemployed pretty soon. People have to be able to work quickly and efficiently. Of course they also have to be smart enough to edit/revise what AI gives them! But all the naysayers here sound, I don’t know, old! Thanks…

Jesus Christ, look at the quality of your own writing. Do we really think this person is capable of determining the quality of outputs from ChatGPT? It’s the blind leading the blind here, and we’re all going to be worse off.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Asking AI for sources is really the worst use for it- AI does best when you load in a document and ask for a summary, talking points, etc.

It’s also great for first drafts of things that don’t matter that much, like an email.

I’ve also loaded standard operating procedures and asked for improvements or automations I can make in processes.

I’ve also used it to grab code to automate some current processes- something that I really did not have time to figure out before- it’s amazing, but it’s a tool, and you have to know how to use it properly.

I can see your point. At the same time, AI adds errors and mistakes. Checking for its mistakes takes as much time, or longer, than just doing the work myself from the start.

Not really? I’ve found high accuracy if I ask AI to summarize a document. It’s recognizing patterns in text at that point, not trying to generate answers. It saves an enormous amount of time. I can ask it to draft out a memo and it basically works as an outline that I can fill in and elaborate on.

It’s a tool, it’s not magical, and you have to know how to use it. You can’t treat it like voodoo or be overly reliant on it. The more you understand how it works, the more powerful it becomes for you.

Anyone who thinks AI is 1) actually AI and 2) a powerful tool is far too stupid to reliably check its output for accuracy.

While it's not that impressive at the moment, about on par with a 100 I.Q. college educated human, when you compare it to what they were like 5 years ago, it's quite amazing how rapidly they are coming along and improving in their abilities.

They’re just models, fool. The same as models we’ve been using for decades without the “AI” branding.

There is nothing “intelligent” about any of this.

The intentional hindrances are holding them back.
Something like half of Google's staff are needed to keep A.I. from becoming too sentient, too racist, too homicidal, etc.