Any number of ways also includes more tedious. There are already a lot of workers effectively performing QA on ML/AI verdicts (e.g., object recognition, malware analysis).
I can’t be the only one who finds the lack of power-user automation of web-applications in 2024 surprising.
When you compare Windows (and MacOS) desktop applications to web-applications, what stands-out is that web-apps get automation “for free” - whereas desktop apps needed significant investment in COM Automation / AppleScript to be user-automatable.
Obviously the problem is that the user’s agent (i.e. their web-browser) does not come with a Macros Explorer or VBA Editor, which hinders entry for novices (no, Browser Extensions aren’t the same). It is unfortunate that no browser vendor is economically incentivised to build-out this functionality - instead users get steered towards cloud(TM)-hosted, non-browser-based automation systems like PowerAutomate - or hit a brick-wall-learning-curve in building a browser extension.
So basically, machines doing your job for you is bad because a job is the only thing that is supposed to give your life meaning, and without that you won't be able to survive.
Well, do s/meaning/sustenance, and it's... not that inaccurate. Of course, there is also a fact that many kind of jobs exist, and that, empirically, machines eliminating some kinds of jobs generally lead to creation of new kinds of jobs (and increase of labour demand in the surviving kinds of jobs) so it's not as catastrophic as one might think.
Well, of course it's not hard to make such an argument because it has been made already multiple times throughout the modern history. So while it may very well be true, I am very sceptical about theoretical arguments on this topic: they've been consistently wrong so far.
Another lens would be that labor is the primary way that the broader population has meaning to the concentrations of capital and power in an economy. Society is broadly structured to reflect this, whether that is ethical or not from the individual perspective. For those people on the labor providing side of the equation, machines replacing their labor is an existential threat.
> Robots reduced the perceived meaningfulness of jobs across the board, irrespective of age, gender, skills and the type of work. In theory, machines can free up time for more interesting tasks; in practice, they seem to have had the opposite effect.
Because instead of moving us towards a <20 hours work week, it's "do more with less people" instead.
Just like in the old times every engineer would have a secretary to do mundane work. Nowadays we have computers but lost secretaries, so we are forced to do even more mundane works by ourselves, with the excuse that the computer makes it faster.
My experience is that the tech profession has morphed into connecting machines and algorithms that are exponentially harder to debug and secure, this is the biggest downside of the new era. In the old times one could write an algorithm and have a relatively good sense of how it interacted with data. Nowadays any important system is composed of dozens of machines and hundreds of disparate libraries strung together and interacting in different ways. Finding a problem and isolating it has become similar to finding a bug in a GOTO-ridden algorithm.
I can agree to an extent, but I think a lot of this can be mitigated by having tight, isolated connections between parts of a system where each result can be validated. Then you can address each step/service in isolation based on the inputs and outputs of each. This is sometimes harder to achieve, but having specific inputs and outputs at each step makes it easier to identify which step is "faulty". It's not always perfect, for example in the case of a validation library that is erroneously allowing incorrect data though, but you can hopefully identify the surrounding steps and narrow your search.
This is the theory, and it can work if done properly. However, just like in a GOTO-ridden language, modern distributed systems are extremely difficult to setup in this optimal way. The more components we have in our distributed system, the harder it becomes to have any sense of safety and correctness. No wonder we started to see systems imploding worldwide as it happened last week.
While connecting say two systems of 99% reliability will be better than say connecting two systems of 50% reliability I don't think that's the real issue.
I find the real issue is that there is no error propagation nor audit trail. Somebodys system notices a problem and does like `throw new IllegalStateException("Invalid Arg")` and good luck figuring out what happened. A lot of system/code aren't designed to be tested easily and they just accumulate bugs because nobody can fix them.
Those interested in this topic highly recommend reading "Ironies of automation" by Lisanne Bainbridge, its from 1983 but very relevant and provides a more thoughtful and nuanced discussion on the interplay of automation and work.
In their research Ms Nikolova and her co-authors found that people did not perceive a loss of autonomy if they were working with computers, where they have more control of the machine than the other way round.
This is why I strongly advocate for open source to replace the Big Tech closed server farms, the learned helplessness that people have when they give their entire audience over to Twitter and hope they dont get deplatformed etc.
Another angle is that machines and software are extracting the value of the work. In the past, you were employed to do some work that you did by yourself. Nowadays, a lot of this work will be done by computers, but it is not free. The seller of the system wants to be paid monthly for this help. So a lot of the money that would go to the employer or the employee is extracted by a third party.
>>"A technology that cuts down on boring tasks is fine; one that threatens your sense of identity is not."
The shareholder class sees no distinction between the two. This is a good reminder not to make your labor your identity and that technology implemented in the workplace is not for the benefit of the worker, but for the benefit of the owners. It doesn't matter if I toil with 1990's technology or 2020's technology, I call it quits at 5pm regardless.
In my personal life I get to choose (for the most part) the amount of technology that I interface with. The only computer in my car controls the spark plugs and fuel injectors. My bicycle shifts with friction. My stove/range only has valves and igniters. My TV is a monitor hooked up to a small PC. I'm not against technology, but I am ruthless in determining the exact amount that makes me happy.
At my job, AI creates a lot of garbage data that punishes and reward our drivers seemingly at random (poor job parsing images for correct delivery). I get to audit it in a super tedious manner. It's just plopped on top of my previous work, with no reduction in workload (I'm sure it will balance out eventually...)
where the AI do not interface with robots or elaborate contrivances... they become puppetmasters to humans. Much cheaper to do the actual work in meatspace!
27 comments
[ 6.6 ms ] story [ 67.1 ms ] threadWhen you compare Windows (and MacOS) desktop applications to web-applications, what stands-out is that web-apps get automation “for free” - whereas desktop apps needed significant investment in COM Automation / AppleScript to be user-automatable.
Obviously the problem is that the user’s agent (i.e. their web-browser) does not come with a Macros Explorer or VBA Editor, which hinders entry for novices (no, Browser Extensions aren’t the same). It is unfortunate that no browser vendor is economically incentivised to build-out this functionality - instead users get steered towards cloud(TM)-hosted, non-browser-based automation systems like PowerAutomate - or hit a brick-wall-learning-curve in building a browser extension.
Regular Economist nonsense.
Because instead of moving us towards a <20 hours work week, it's "do more with less people" instead.
I find the real issue is that there is no error propagation nor audit trail. Somebodys system notices a problem and does like `throw new IllegalStateException("Invalid Arg")` and good luck figuring out what happened. A lot of system/code aren't designed to be tested easily and they just accumulate bugs because nobody can fix them.
https://ckrybus.com/static/papers/Bainbridge_1983_Automatica...
This is why I strongly advocate for open source to replace the Big Tech closed server farms, the learned helplessness that people have when they give their entire audience over to Twitter and hope they dont get deplatformed etc.
It is why I work on https://github.com/Qbix/Platform
And why I started https://engageusers.ai/ecosystem.pdf
The shareholder class sees no distinction between the two. This is a good reminder not to make your labor your identity and that technology implemented in the workplace is not for the benefit of the worker, but for the benefit of the owners. It doesn't matter if I toil with 1990's technology or 2020's technology, I call it quits at 5pm regardless.
In my personal life I get to choose (for the most part) the amount of technology that I interface with. The only computer in my car controls the spark plugs and fuel injectors. My bicycle shifts with friction. My stove/range only has valves and igniters. My TV is a monitor hooked up to a small PC. I'm not against technology, but I am ruthless in determining the exact amount that makes me happy.
https://marshallbrain.com/manna1
where the AI do not interface with robots or elaborate contrivances... they become puppetmasters to humans. Much cheaper to do the actual work in meatspace!
They already did. See Microsoft.