[p2detar]

> I worked as a developer at a company. I asked the business owner a question about a business task. He sent me a ChatGPT screenshot with the answer. I replied that it had nothing to do with my question and everything there was wrong. A minute later he sent me another ChatGPT screenshot. He didn’t even read the AI’s answer.

That's just rude and borderline psychotic behavior.

It's still a bit better at my workplace but irritating nonetheless - my boss would "research" a feature and prep notes in our wiki with some gemini chatbot exchanges attached. This is a of course no specification, but it's supposed to be a good base point to start working on the feature. Gemini already chose the coding libraries and concepts, so to the outsider it just seems like all that's needed is to code that into the product. Of course, it's not that simple and it mostly gets in the way rather than help. But now questions arise why is the feature not ready yet, when "the plan" is already there and so obvious.

[cjs_ac]

Just before LLMs became available to the general public, I worked for a small (fewer than ten employees) strategy consultancy. They had some industry-specific analysis tools that were in Excel, and my job was to turn them into a software product that customers could operate themselves. The owner had a mechanical engineering degree, but every time I asked him a technical question about the tools, he'd just give me a sales pitch for the thing I was trying to build for him. He was always pitching for new business, and seemed to struggle to get out of 'sales mode'. I have no doubt that if I were working for him now, he'd be pointing me to an LLM in response to any question.

[tardedmeme]

This is why executives think LLMs can replace everyone. They can see it can replace them, and project that onto everyone else. And they don't see the gaps in knowledge because they don't care about the facts, only the presentation.

[simpaticoder]

I have a theory that most people at most levels of work deal with the same amount of mental complexity. The difference is that as you move up the unit of abstraction gets larger, so your decisions and knowledge cover more scale. E.g. the engineer thinks about functions; the business owner thinks about products; the investor thinks about companies; the president thinks about nations; etc.

But people at the higher abstraction levels have a problem because they often never had (or lose) the ability to zoom in anywhere. And even the ones that can don't have time to learn to zoom in everywhere so they have to learn how to trust others, aka recognizing shibboleths. AI is great at sounding trustworthy and making reasonable looking output. In so doing, however, it's undermining the utility of shibboleths for large scale thinkers! That is the powerful are now deluding themselves that they have access to infinite reliable experts, and have gained the ability to zoom in everywhere, for only the cost of a data center. In a sense programmer experts like us are lucky because we have objective verification as a feedback loop to temper the exuberance. They do not.

If this is true, the kinds of error modes we'll see will be novel and catastrophic to a large fraction of businesses. If the feedback loops for correction are damaged or destroyed, we'll see firms gleefully, optimistically and energetically committing obvious mistakes until they die.

[cm11]

I think agree with the broad, but disagree with the specifics. I really don't think a lot of people at the product level think about products. I really don't think the investors think that much about companies. I think as you move up, the things that separate are (in no order): ability to solve OR mask problems (one of these requires more skill/effort, but to people above they can look a lot closer), ability to naturally or adaptively (this is a fair, but somewhat positive word for this) socialize with the people in that next level, nepotism or blurrier forms of getting your foot in the door, willingness to look the other way (or just be obtuse), willingness to lie (or just be obtuse).

After the IC-level, each level filters less for the duties of the higher role. Even PMs seem surprisingly par at product ideation, vision, insight amongst their cross functional project team members. So many products fail, so many startups fail, so many projects are late, but those things don't seem to be what dictate promotions or investment even when they are specifically that role's role.

[Balgair]

https://thebullshitmachines.com/

This is a bit dated, but I think the message out of UW is right.

To your point on bosses: Turns out, you can very much bullshit a bullshitter.

[nstart]

Also, I submit Ed zitron's the era of the business idiot.

https://www.wheresyoured.at/the-era-of-the-business-idiot/

Lots of little claims I disagree with there but the overall thesis has felt prescient these days

[alexanderh]

damn, this is a real zinger lol. Definitely seems this way.

[theturtlemoves]

Maybe unrelated but this memory pops up: I once worked with a mechanical engineer who could not visualize anything in his mind and had quite some trouble with word finding and explaining his ideas. He just couldn't say what needed to be built, he could only evaluate after it was built (often "argh, no not that").

It's as if his training had centered so much on 3d modeling and tangible tweaking as you go, that he hadn't learned to simulate anything with his thoughts before starting. He had to start building it to figure out what he meant. Incredibly creative person, out of the box thinker and big picture visionary, but with difficulty translating his ideas to verbal concrete steps. But nobody realized this, which resulted in a lot of frustration both ways:

"Why didn't you build what I said?" "I did. This is exactly what you said." "No it's not! I said x y z" "Yes you said x y z, this IS x y z" Silence. "Then that's not what I meant"

[gedy]

Same, but not sure it's his training and just the way his head works. Have met a lot of people in software who can't draw or understand (very) simple block diagrams of systems. Some people don't have an inner eye.

[alnwlsn]

Now you know why AI text to CAD works so poorly.

[ai_fry_ur_brain]

Its so much worse now.

I deal with this all the time these days. If I ask for some clarification on scope, I usually get these ambiguous answers with fantastical ideas usually concerning some dreamed up impossible to implement tool that people assume is now possible because of AI, because everything is possible to them now!

[Tangurena2]

There's an old computing maxim: GIGO. It used to mean "garbage in, garbage out" meaning that if you put garbage into a computer, then the output was garbage. Nowadays, GIGO means "garbage in, gospel out" meaning that no matter how bad the input was, the output was to be treated as The Word of God.

[progbits]

Since the tool's quality strongly depends on the abilities of the user, I think I prefer:

Garbage Invoker, Garbage Output

[thesamethrowawa]

> That's just rude and borderline psychotic behavior.

With some notable exceptions, this describes almost every business owner I've worked with.

[bitmasher9]

I use to get emails with some oddball questions slightly out of my field of expertise from business owners. I would answer, and they would forward my email to the person that asked them the question. They saw their role as routers.

[rightbyte]

Delegating questions to experts seem reasonable?

[ranyume]

Expropriating value is not.

[simpaticoder]

>That's just rude and borderline psychotic behavior.

No, that's just disrespectful and entitled behavior from someone in power toward someone without power. I'm sure the person (business owner) felt pleased with themselves for saving so much time and mental effort in the interaction, with additional points for "authenticity" and "honesty" because they didn't hide the fact that they used an LLM. It damaged the relationship slightly but only time will tell if that was meaningful (and the harsh truth is that it probably isn't given the short lifespans of companies).

[qwertox]

Maybe he was trying to say "look, when I can look this up, it means that you can too, i won't put any effort into it, because i pay you, but if i can use AI, so can you."

Maybe that was his rationale, "to educate" by example.

Absolutely nuts to treat employees this way, sending ChatGPT chats around.

[ramon156]

[flagged]

[542354234235]

[flagged]

[moomoo11]

so… they’re conmen

confidence man

[thesamethrowawa]

Flowers for Algernon has something to say about this.

[dbetteridge]

So Silicon Valley's big-head is based on an actual stereotype?

[abalashov]

Yeah, except that Big Head was, as far as I understood, just sleepwalking or floating through life. He more or less wholly lacked that unquenchable thirst for money that motivates MBA frat boys to yell at balding beta cucks to type the code faster before AI replaces them...

[justinclift]

> psychotic

Probably more under-developed than psychotic.

ie not really using their adult thinking any more

[notnullorvoid]

I think psychotic fits. They have lost touch with reality, and are treating AI as some all knowing entity.

[xp84]

People have no mental model for it. They know people don't know much, and are used to computers being deterministic. Now (setting aside complete rubes who believe anything) people are used to detecting human-generated BS and spam on the Internet.

- It's nearly always poorly written because dummies also usually can barely write

- It's poorly packaged (some random comments section, not a polished slide deck)

- It's biased in predictable ways (when you do know who wrote it, their self-interested takes are predictable).

Enter "AI" - it writes mechanically well (even though it overuses several tropes, it doesn't make grammatical errors). AI packages its output as exactly what you ask it to (e.g. finished coding projects, outlines, essays, etc.) And not being a person, it's not guaranteed to have a certain bias, but it does echo and validate all your own biases.

Given all this, people think AI sounds exactly like what they assume an oracle would sound like.

[snarf21]

At my company, non technical Managers (including marketing folks) are posting Claude and ChatGPT responses in bug tickets with the "fix" for the bug. No explanation of what the expected vs actual behavior is, just AI slop telling me how to fix it. Soooo helpful.

[batu1509]

The 'AI-generated pseudo-spec' is becoming a massive productivity killer. Non-technical managers prompt an LLM, get a highly confident 5-page document with chosen libraries, and think they've done 80% of the work. But as a developer, you now have to spend hours debugging and debunking hallucinated architectures just to prove why the AI's approach doesn't fit the existing codebase. It completely shifts the burden of actual product management into AI-babysitting for the dev team.

[timbaboon]

I was discussing a change we wanted to make to some tables in one of the DBs, but it would require a lot of effort because of the sheer size. Anyway, after saying how much effort would be required the response I got was to “please ask ChatGPT.” It gave a slightly different response, but changed it after I presented it with my solution and agreed that was the better approach. Just… sigh

[xp84]

"Apologies, you're absolutely right. My initial strategy indeed would have caused a massive database bottleneck because it would lock the entire table. This could result in downtime for the application. Let me know if you'd like me to make a step-by-step plan to implement your solution."