Except when they're not. I have repeatedly, explicitly told ChatGPT I don't want a bloody cheerleader; I want a devil's advocate that will tell me when I'm full of shit, when my story idea is nonsense, when I have plot holes you could drive the millennium falcon through, etc. If something works, tell me why. If something sucks, tell me why.

What do I get? Every damned response showers me with praise. If I give it some of my story to critique and it tells me what a genius I am, then I notice a detail I got wrong and point it out, it pivots and tells me how right I am for changing that detail.

"Chef's kiss" is the absolute worst. I have baked into my permanent customization never to use that phrase. I even said "Every time you use that expression, I'm going to club a baby seal to death." Not only does it ignore that instruction, it deliberately flouts it and jokes about it: "Chef's kiss! I know, but that seal had it coming."

I've read from other users that OpenAI has deliberately made it more obsequious so as not to alienate paying customers.

Yes! AI chatbots really hate it when you anthropomorphize them.

Be careful. AI chatbots really hate it when you anthropomorphize them.

I'll leave others in this conversation to argue over what constitutes a "majority" and focus on the demographics and attitudes of those who voted for him. I place the blame entirely on Biden and the Democratic Party for this. They had the election in the bag, with most people - even most Republicans - unwilling to sink into the Trump debacle again, and somehow they still managed to fumble it. Why?

First, because they didn't quietly but firmly push Biden out of the race from the get-go. If they had started off with a real primary with viable candidates, just about anybody could have beat Trump. But Harris was unprepared and very obviously out of her depth.

How do I know this? Because even people that had every reason to vote for her didn't. It's so easy to write off Trump voters as being too stupid, racist, or self-serving to vote for someone better. But you need to look at the exit polls and demographics to see the truth: a tiny sliver of a minority of self-described liberals (7%) and Democrats (4%), atheists (27%), various marginalized groups, and a whopping superminority of Latino voters (46%) went with Trump. They consistently cited concerns for the economy, the border, and foreign policy in their reasons. If the Democrats had pulled anybody at all who could talk knowledgeably about those issues and propose good policies, we never would have been saddled with Trump 2.0.

I told my lefty friends the same thing this election that I told my righty friends last time around: If you just complain that your side lost because of stupidity/hate/corruption, you haven't learned your lesson and you won't do better next time. Listen to what the voters say made them vote the way they did. Take their concerns seriously and address those concerns, and you'll win.

Everyone in IT knows that AI is just a Clever Hans pony trick, an illusion to fool the rubes who don't know better. It doesn't think, it doesn't reason; it just digests and regurgitates without any conscious understanding of what it's doing. At my company we are encouraged to use Copilot for increasing productivity, but it slows me down. I ask it for API endpoints, and it just makes crap up. I ask it for Powershell code, and it invents nonexistent functionality. Whether it's AI-generated art, or prose, or code, or ad copy, or music - you're going to get garbage you have to manually correct.

Yeah. I see it kinda like the apostle Paul. I'll do everything I can to make the most out of this life and try to enjoy it as much as possible, but I have zero desire or intention to prolong it once my body or mind start to give out. I won't take active steps to shorten my life, but I will indeed "go gentle into that good night", and be glad it's over when my time comes.

I wish more people understood this. ChatGPT et al are really just glorified ELIZA except with a bigger vocabulary. If your pool of possible complete sentences and keyword recognition is in the trillions instead a couple dozen, it's still just a call-and-response rote ritual.

Whenever a new AI service comes out, I put it through its paces and inevitably see the same pattern emerge:

For the first few days, it blows me away with seemingly unique, on-target responses to my prompts. This is true whether it's a chatbot for creative writing, an image generator, or a music generator.

Then, over time, I start to see that it's rehashing the same concept over and over. In an extended roleplay conversation, it repeats the same stock phrases no matter what the context.

What we're seeing is ELIZA all over again. A one-trick pony that's amazing the first time you encounter it, but before long you realize it's all just mimicry following preset guidelines. The only difference between ELIZA and modern AI is that the sample set is many orders of magnitude larger, so it takes longer for the system to start repeating itself.

My company blocked access to all AI chatbots through our firewall - except Copilot. We're encouraged to use Copilot to solve our coding and engineering problems.

A typical use case:
Me: I'm writing a powershell script to solve such-and-so need in our on-prem Azure Devops Server instance. What is the right API endpoint to get this information?
Co: Sure, just use {nonexistent powershell function} to go to your.base.url/_apis/non-existent-endpoint
Me: Wait, {nonexistent powershell function}? That's not documented anywhere, and powershell crashes when it hits that command.
Co: Oopsies! I meant {other nonexistent function}
Me: That didn't work either. I googled it and found {correct function}
Co: Goodness gracious me, that's right.
Me: I tried that endpoint and nonexistent-endpoint isn't real
Co: Gosh, I don't know what happened! I do apologize. Try {other nonexistent-endpoint} instead

Yesterday was classic Copilot
Me: I'm running this pipeline in our on-prem Visual Studio Server instance (no matter how many times I say that, as often as not it gives me answers that only apply to the cloud Services). There are two pipeline agents installed on server {blurp}. When the pipeline uses agent A, it works correctly. But when it uses agent B, it throws this error saying it can't find the right version of Visual Studio. As I mentioned, they are both on the same server. They also both use the same service account, so they both have the same environment settings, the same versions of VStudio installed, identical agent versions, identical .NET assemblies, etc. If I open the agent configuration screen in the Azure agent pool setup screen, both of them have identical capabilities. How is it that agent A runs it correctly but B fails?
Co: Compare the versions of Visual Studio and make sure it is correct for both of them.
Me: Are you unable to read? I just told you, they are both on the same machine with the same VS available.
Co: Righto, try looking at the agent's capabilities and make sure they are the same.
Me: What part of "capabilities are identical" do you not understand?
Co: Look in the agents' config folder.
Me: In C:\agents\A and C:\agents\B, there is no folder called config
Co: My mistake, I meant to look at config.json.
Me: There is no file called config.json.
Co: Compare the versions of Visual Studio and make sure it is correct for both of them.

I lose more time trying to get a straight answer out of Copilot than if I had just googled instead. And considering my low opinion of the quality of google search results, that's a pretty low bar.

So, my fellow coders, beware: don't fall into the trap of letting Copilot do your thinking for you. That way lies madness (and really crappy unusable code)

Postscript for those wondering how I resolved the problem: I had to uninstall the "bad" agent and reinstall it to get the pipeline working correctly.

Post-postscript: Yes, the above dialogue is paraphrased. Copilot does apologise profusely every time I call out its mistakes, but it doesn't use so much colloquialism. However, my angry sarcastic responses are verbatim.

The responses to parent comment illustrate exactly what I've seen from my friends on both sides the past few elections. My lefty friends did it in 2024, and my righty friends did it in 2016: Call the people who voted for the candidate they didn't like stupid.

They are so consumed by hatred for the opposing team's candidate, they cannot conceive of a world where smart, informed voters deliberately chose to vote for that person. So they rationalize their side's loss by saying that all those other voters are too stupid to make the right choice, or they hate America and want to see it fail, or they're Nazis, or Communists, or the election was stolen, or or or... anything but rational adults.

Let me be clear that this is no defense of either candidate. 2024 was a rehash of 2020 and 2016, with choices that were sleazy at best and more leaning towards horrific. I'm just saying, everyone (well, ok, most people - I realize that there really are stupid people who vote based on what their favorite talk radio hosts tell them to think, or name recognition, or popularity/personality of the candidates) who voted had their reasons. They looked at what they believe needs to be fixed, looked at what each candidate offered towards solving those problems.

I keep telling my friends this when they fall into the "only stupid people voted for the other candidate" mindset: If you can't get past that and look at the exit polls, read the issues that most concerned voters, analyze why your candidate failed to convince the majority that he or she was the better choice to fix those issues... you're never going to learn what it takes to produce a winning candidate. Get past your thick-headed prejudice and look to winning hearts and minds with a superior strategy.

What I got out of my bachelor's degree in computer science: COBOL syntax, FORTRAN Honeywell CP-6 basic job control commands, UCSD Pascal with IBM graphics extensions, submitting jobs via punch card reader. Guess how much of that I still use?

What I got out of on-the-job training and personal study that still serves me 40 years later: Structured programming, troubleshooting, error trapping, reusability, future-proofing my code, dBase, Clipper, VBscript, C++, Powershell, MS-SQL, git, Azure Devops administration, bash, REST API design... probably more but I ran out of fingers to count on.

My daughter got some good hands-on training in PC repair, but... that was 10+ years ago and technology moved on without her.

We really need a return to trade schools. You don't need a college degree to do most of the work that keeps the world turning - whether it's plumbing or engine repair or even laptop repair or Azure Devops administration. Getting a bachelor's degree for that stuff is... B.S. rimshot

Boomer here*, over 30 years in one job because (for the most part) I love what I do and I really hate job interviews. Or at least I think I do. How the hell can I remember a job interview I did in the last century?

I did, and do, work 50-60 hour weeks on a regular basis. I'm actually working more hours and taking less vacation now than I did before I started working from home. Why? Because management thinks we owe them the time no longer spent on commuting, so they schedule meetings before the official start of the work day. And of course we're on salary so we do all that for free.

My younger colleagues burn out pretty quickly and go hunting elsewhere. But me? Not so much. Who's going to spend the time and money hiring and training a new employee who's likely to retire in 5-7 years, and at a starting salary competitive to what I currently make? So I hang on to what I've got and I am thankful for it, because it's better than filing for unemployment this close to collecting Social Security that may or may not exist in 5 years.

Now pull up your pants and turn that damn stereo down. And quit eating Tide Pods; they're not good for ya.

* Yo, check out my 6-digit userID** for street creds

** Yes, yes, I know there are over 600,000 users who were here before I ever heard of /. but my point is this ain't my first rodeo.

Circa 1988 I was getting into raytracing with the newly ported-to-x86 DKBTrace, later called POV-Ray, and for support I hung out on SIGGRAPH, the computer graphics forum on CompuServe (kids, ask grandpa about CompuServe. It was nifty!)

I read a post from a user also looking for advice on the subject, and I was surprised to see the name: Ward Christensen. I asked him "Are you THE Ward Christensen, father of XModem???" He was genuinely surprised that anyone had heard of him, much less was a fan.

You're not as at-risk as the people you point them at. Can folks in the path of your doxx-o-vision opt out of having their data sent to Meta?

Yeah. Remember the Google "master plan" graphic that was the darling of memes a hot minute ago? A joke business plan that included "orbital mind control", "robots enforce terms & conditions", "crop circles", takeover of "banks", "convenience stores", "airlines", "space travel", etc.

It was a joke back then. Now? Hard evidence in court.