raynr

1. The funnel is even steeper than the linked post suggests.

There's a huge range of users between "casual and infrequent users" and "professional use of frontier-grade LLMs in math & code".

I think I'm in between there somewhere. I'm far from a software dev, my experience is playing with delphi and C when I was a kid. But I've been using LLMs to vibe code personal and work projects that I could never have done even a few years ago. They're genuinely helping work, because there's always some place where a simple little script or other utility can help. (Hell, isn't this partly why so many people start up startups?)

I imagine that many departments in many companies that are not "tech" companies face similar challenges - there's no budget to hire a software dev or even engage a third party to do things, but now because everyone's aboard the AI train, they have access to something (chatgpt, gemini, copilot, claude), and that's helped users implement their own solutions. Somewhere in the world, someone's excel-file-based accounting system has gotten an upgrade with help from an LLM.

2. The funnel is also driven by the the fact that users need products, and an LLM itself isn't a product, it's a capability.

Software devs are at the front lines because they can build their own products that can integrate LLM's capabilities. You already know how to automate testing, orchestrate, work together in a team. You're writing / configuring harnesses and flows that work for you. You're sharing / hoarding skills files.

The rest of us, like me, don't have that skill. Copilot the product is a mostly useless chatbot that takes up space in MS word. I guess its excel cleanup capability is semi useful if your company runs on excel and you've never used a filter before.

There is no work I can put an LLM to other than a "writing aid" or "better search", because the only AI product we have is the chatbot and maybe someone's equivalent of an agent studio, none of which we are trained to use.

Until the products come, there'll be a funnel.

The price is high because there is sustained demand.

Hah! No but seriously. For a variety of reasons, owning a car in Singapore is a step change in QoL.

Infrastructure reasons:

- abundant parking at destinations (e.g. malls), home (HDB flats, condominiums) due to fairly generous car park lot mandates. IIRC, HDBs, where 85% of Singaporeans live, have until fairly recently had a 1:1 unit:lot ratio. Similarly for condominiums. A lot of these are progressively being scaled down especially near train stations, but buildings and giant multi storey parking lots are fixed infrastructure that will be around for decades. Abundant parking will be available for decades still.

- roads are wide and well maintained and well connected.

The experience:

- because the number of cars on the road are controlled through the COE system, the experience of driving is good. We get slow traffic but it is extremely rare that we get the kind of multi-hour jams that other countries experience.

- the door-to-door experience saves tons of time. In public transport, I walk (7m), wait for the train (3m), sit on the train (36m), get out at the interchange and walk to another platform (2m), wait for the train (2m), sit on the train (12m), exit and walk to my destination (11m). Total time: 1h14m. In a car, I walk to my car (2m), drive to my destination (35m), walk from the car park to my destination (1m). Total time: 38m. x2 for the return trip, and the car saves you a whole hour a day.

- the public transport experience is not so good if you have a family or are disabled. Don't even mention cyclists, culturally we spit on them here.

The status: - it's true, the car is a status symbol. You don't know it until you get into conversations where driving a "mere BMW" as opposed to a mercedes is looked down on, or when someone mentions that driving a hatch is "not appropriate" (to station). Sedans are still popular here because they're "executive". Yes, these are all real conversation points I've been in.

Combine all of the above and owning a car in Singapore means you save tons of time, is technically "affordable" after loans, and you get to feel like you made it. So there's huge pressure to get a car. And once you're hooked, it is very difficult to give it up.

Then you have other reasons, like Taxis and Grab compete with individuals for the COE, and parents who "made it" in the boom years buying cars for their kids, and foreign money coming in. But I don't know how big these effects are.

I am currently surrounded by hyperventilating corporate and country level leadership who are fully swept up and captured by the AI craze.

I have seen some wild failure modes from people who are outsourcing their thinking to LLM. Amendments to clauses that don't make English sense. Multi-paragraph long replies in emails that say nothing specific to the issue at hand. Responding to questions with "AI says this" (but I asked you, not the LLM). Leadership wants us to embrace AI but there is no product for the layperson, it feels like everything is front end + generic prompts + $LLM_API_key. People want to make customer service bots that have access to personal data.

I feel like software devs are so lucky in that at least people in your field can see an LLM for what it is and harness (no pun intended) it appropriately. As a lay user, no such luck. Leadership and purse strings are far removed from IC work and don't understand why an AI product wouldn't work, they've heard otherwise in their circles, you had better make it work so that they can claim to have delivered an AI transformation this year.

Enter Zitron.

Zitron is a woo-pushing grifter (his product: his stance on AI). Even without examining his reasoning or the accuracy of his predictions, Zitron is hard to listen to. Mostly, he shouts out a constant barrage of bare assertions that his research is thorough and irrefutable and the doom is coming and ever "AI booster" who disagrees with him is an idiot, all the while without actually spending time arguing his point.

But Zitron feels like one of the few people actually pushing back against this craze.

I would very much like a better argument for a position that I support, please and thank you.

I'm surprised this happens between workers within the same knowledge domain. If you're a software dev asking another software dev for an opinion on code, can't you just... talk shop?

In my experience this sort of thing happens all the time when something has to cross between different knowledge domains and the sender doesn't apply their mind in order to co-operate on the issue at hand.

In the pre-LLM past, the sender would add no value, acting as a simple email forwarder, and they would blame you for the delays caused by your inevitable clarificatory questioning. But I always had a defence, which is that your email had no inputs and I had to ask questions to clarify.

Now, though? With LLMs, they just run whatever it is (contract, memo, policy) through whatever LLM they have available and paste the output in an email, giving them the appearance of having done work and added value to the project.

But their LLM outputs don't make sense, or don't apply to our organisation, or is a fluffy and abstract "right answer" with no connection to the specific concerns of the business. Parsing it is a chore, and takes time.

And since now I am the only one actually taking that time, I become the visible cause of the delay.

It's infuriating. Any tips on how to deal with this would be greatly appreciated.

There are also DCTs which IIRC don't use a torque converter. But chances are pretty good that an automatic transmission car uses a torque converter.

As a layman, this too strikes me as the problem underlying the "confidently wrong" problem.

The author proposes ways for an AI to signal when it is wrong and to learn from its mistakes. But that mechanism feeds back to the core next token matcher. Isn't this just replicating the problem with extra steps?

I feel like this is a framing problem. It's not that an LLM is mostly correct and just sometimes confabulates or is "confidently wrong". It's that an LLM is confabulating all the time, and all the techniques thrown at it do is increase the measured incidence of LLM confabulations matching expected benchmark answers.

I think that's the cart before the horse. People buy the big phones and so businesses cater for that.

> Congratulations: you have successfully turned your cool idea into a chore.

The article gave me a vague, off-topic sense of unease but your comment crystallised the feeling for me.

I really wish less emphasis is placed on this kind of blue-sky, "strategic" thinking, and more placed on the "chores". Legwork, maintenance, step-by-step execution of a plan, issue tracking, perspective shifting etc. are all, in my opinion, critically important and much more deserving of praise and respect than so-called "strategic" thinking.

Which, IME, most people can't do anyway! After they've talked their big talk you suggest that there's a practical, on-ground problem and they look at you accusingly, like you're sabotaging their picture. And I'm like, no, my friend; reality is sabotaging your picture, it's just the two of us here and you're not losing any face by me pointing that out, and also if you were an actual strategic thinker you'd have taken my on-ground problem into account already...

I played around with the website. I will take the author's advice to not take it seriously.

Many of the changes are active voice to passive voice. It might sound more polite, but in reality comes off as sneering or condescending.

For e.g., if I've already told you something, I usually change communication channels to a more casual one (pick up the phone, pop my head up, walk over, go on a chat channel) and say: "hey, that question you had, it's already addressed in my email, see this bit over here...".

I would not, as the website suggests, say any of this:

- "There seems to be a disconnect here as this information has already been provided."

- "The information has not changed since the last time it was communicated."

- "As indicated prior."

If you did this to me, I would think you are a jerk.

> Something went very wrong in negotiations here. That's all I can say. Who allowed this deal to be penned in its exact fashion?

I'm not sure this is the right way to look at it. It's more likely that nobody "allowed" the situation to happen, and the publishers just had much stronger bargaining power. So libraries either pay to participate and adjust their budget accordingly, or don't have e-books to lend out.

Or, put more provocatively in a forum run by a startup incubator: capitalism allowed this to happen.

My local library does appear to have a large enough budget to cater for this new demand but I can understand if many other city libraries don't.

IAAL but likely not in your jurisdiction, and I agree with this, because the biggest "legal" concern I would have as the linked OP is the company coming back and blaming me for some software I wrote that was out of my job scope entirely.

I would add that the goal isn't just to convince the company to delete the software, but rather to acknowledge and accept that there is no support, no warranty, and if things go wrong it's on their hands.

> I propose that the differences in opinion are likely differences in understanding the ill-defined context of the question.

The whole point of the article is that *on large platforms* it is impossible to have policies on moderation, etc. that people agree on. Not small, interest-specific, geographically and linguistically focussed niche platforms.

The example is further perfectly illustrated by users here dropping their confident, just-so how-tos that would accomplish difficult tasks like moderation easily, and other users disagreeing with their equally confident, personally experienced difficulties in executing such just-so plans.

If you can define the question so narrowly as to make the answers obvious, you're not talking about the same problem anymore.