upvote
Agreed. It seems everyone has forgotten who we’re building software for in the first place.

The idea of removing humans from the loop entirely feels like some sort of AI apocalypse.

reply
We don't build modern software for humans. We build it for profit extraction. It's why the modern cycle is:

  1) First-mover wins: burn VC capital to get a product out free
  2) Establish brand and market dominance
  3) Integrate advertising to keep the product free and grow the user base
  4) Crush competition and capture users due to "mind share" and "network effects"
  5) Enshittification until no one is happy with your product and making a replacement is impossible
reply
I basically agree with you, but much less cynically.

At the end of the day, in a well-policed society "profit extraction" == "what you can convince people to willingly spend their money on"

Turns out people are generally willing to spend their money on ads to grow their own businesses, and are generally not willing to spend their money on un-enshittified software replacements for what they're used to.

You might not agree with their choices but that's just how people are. I don't think it would be a better world to force them to choose differently.

reply
Replace people with companies. And realize that company decisions are increasingly AI driven. Today, often with a CEO that asks AI what to do, but eventually, I think the CEO will also get out of that loop.
reply
Can you show me this well-policed society? We've had quite a few decades, almost a century of this new kind of capitalism and afaik in most places it's just become feudalism 2.0.
reply
1. Personally I feel most Western countries have gaps but are well policed enough that I feel comfortable saying at least ~95% of spend represents the revealed preferences of people under their own free will. People could do better with their choices but they aren't coerced or outright defrauded in most cases.

2. I (like most people I feel) am generally very on board with political action to reduce fraud & coercion. I'm not on board with "let's just overthrow capitalism", which just introduces coercion on people's preferences, rather than remove it.

reply
Then again, often when I specifically prompt my agent that "this works like shit, make it better" - it makes it better. There are for sure best practices that agents are well equipped to handle. Often the solution to make UX better is to just make the process simpler (for the user). Agents are also well equipped to do that.
reply
so put "make it better" in an infinite loop = profit?
reply
I don't see any reason all the things you mentioned cannot be taught to an LLM using reinforcement learning. The newer models are already much better at this stuff, I see no reason that trajectory will change.
reply
Hence I wrote: "Structured data of actual usage would be needed for AI to ingest, in a way that it can derive intent from every action and identify friction."

The challenge is more about providing actual, objective data to learn this.

The LLM will not be able to decide what is "good UX" and "bad UX", it will have to ingest this information in vast detail from judgement provided by humans, which is also varying alot.

Case in point: Coming from that field, I personally don't have a good user experience using iOS, a majority of end-users will strongly disagree.

An LLM would need to be trained to handle both cases to judge the same human interface, so it would need alot of data to understand both views...

reply
The human UI designer has the benefit of being and to run experiments on themselves, but only n=1. for any problem where that is insufficient or even misleading that benefit disappears. I’d argue that UI is an area where you need n>>1.
reply
That's what UX user studies are for, which is a process of humans observing other humans using the product and taking notes. It's far more common than people seem to believe.

(also: UX is not UI, but I guess this goes without saying...)

reply
Not so. You can time how long it takes, how many clicks to goal, how much of the viewport is shown. There are so many metrics around UX and it can be approached in an objective way. So one UI can be evaluated on another for a given goal. Sick Fable on it and see.
reply
I frankly assumed the example of vending machine vs video editing software illustrated how this is not the answer.

A perfectly scoring vending machine UX applied on Adobe Premiere will tank the UX for Premiere users.

reply
Unrelated but as someone who has used several NLE Adobe Premiere is not what I would pick for good UX.
reply
Those are ways of optimizing patterns we already use. New UX patterns from the simple like object rollover states, to the complex like the iPod click wheel, take human innovation.

We respond with delight or disdain based on how they feel.

The only way for AI to know how humans feel about something is to be trained on what we’ve already said about it.

reply
I do so suspect that harebrained ideas like this is how we got “fully optimised, completely garbage” UX experiences that are present in some major corporate services. Google/youtube being a good example of this.
reply
Of course discretion is needed. It's just another tool at our disposal.
reply
Sometimes, the best way to use a tool like that is by disposal.
reply
> AI doesn't know what "good UX" is

Neither do ~80% (or more) of software developers.

reply
Doesn't make me any keener to use AI. In fact, the opposite.
reply
Of course, because they aren't UX designers.

~80% (or more) of UX designers don't know what "good backend design" is.

Even worse: ~80% (or more) of frontend developers don't know what "good backend design" is.

Not sure what's the point here...

reply
"they aren't UX designers" but they are doing the UX design in 99% of the places I've seen.
reply
Why is it always a race to the bottom?

"AI does X kinda bad" "Well most humans do X even worse so shut up and eat it up"

reply
It's not a race to the bottom. It is a statement that AI is already better than an average person or even an "expert" in certain areas. Probably most, at that. You can see it as "AI is advanced" or simply as "humans are pretty stupid, overall".
reply