 |
| Movers carrying boxes. Image source. |
A few years ago, when my husband and I were moving to a new apartment, we heard about a moving service that said they would do everything for you. You don't have to do anything at all. You don't have to pack or do anything. They do it.
My first thought was "that's not possible."
In my experience, when I've had to pack up everything and move to a new apartment, the first task is getting over the mental hurdle of how overwhelming it is. Ugh, I'm gonna have to look at all my stuff and make decisions about how to sort it and what to keep and what to throw away. And what to pack first because I won't need it for a while, and what needs to be packed last in an extremely convenient way because I will need it immediately at the new apartment.
The idea that I could just not do any of that, that someone else would do it for me... Impossible. How could someone else work through my own feelings about deciding what little knickknacks I'm going to throw away, and how to organize stuff in the kitchen that I haven't used for a while but I would like to have convenient so in the future I won't feel like "oh it's too much work to get out all the baking stuff, I guess I just won't do it" and then never bake anything ever, and whether I should just gather up the random jewelry on my dresser and stick it in a plastic bag, or is that a bad idea because I'll probably forget it exists, also, do I have too many books, and so on and so on?
But then I realized, oh, this moving company offering the "full service"- they are not ~moving~ in the sense that I would define moving. They just put all your stuff in boxes and bring it to the new location. That's all.
To my mind, that is a very different thing from "moving." They're not going to make decisions about what to keep and what to throw away. They're not going to determine the mapping between the locations of every single item in the old apartment, and the locations in the new apartment- ideally we would want to map every object to the equivalent spot, but since the layouts of the 2 apartments are different, there might not be an obvious "equivalent spot," and then you need to determine a spot based on the preferences of the people who actually live there and use those things.
They're going to pack things in general categories- clothes will all be together, kitchen things will all be together- but there will be plenty of miscellaneous things that get categorized into boxes differently than I would have categorized them, and so it will take forever to find them when unpacking.
Really, this "full service" would create extra work for me- I'm going to have to do the process of "what should we keep and what should we throw away" during the unpacking, which is harder than doing it while packing- surrounded with so many boxes full of stuff, you can hardly walk across the room, wouldn't it be easier to just shove all these miscellaneous things in a drawer and then go take a break, rather than making decisions about whether to keep them and where they should go? Also, it's going to be so much extra work for me to locate specific items I need to use, somewhere in my boxes packed by somebody else.
But wait, what if I knew that, going into it? What if I wasn't expecting them to ~move~ as I would move, but I knew they were just going to put everything in boxes and bring it to the new location, and then I would have these additional tasks added on to that end? Ahhh. Okay, thinking about it this way, I *can* imagine situations where this kind of service would be helpful. There might indeed be situations when you just need your stuff moved, and you don't care that it won't be sorted in the way you would have sorted it. We shouldn't model this as 'a convenient alternative to doing the packing yourself'; it's for when there are reasons you can't do the packing yourself, and you accept that this less-ideal option of someone else doing it is better than not being able to move your stuff at all.
The idea that someone else can magically make the right decisions for me about packing all my stuff... that someone else knows what's right for me, in all these little personal details... imagine if you believed that. And then you used this moving service, and they brought all your boxes, and as you're unpacking it, you're gradually realizing "no, this is all wrong, this is all wrong" feeling so distressed at the idea that they are trying to force their categorization system on you. Expecting them to be an authority figure who is all-knowing and will make the right decisions for you, and getting more and more upset as you discover you don't want to live under their rules.
Alternatively, imagine if you believe yourself to be in charge of your own life. You believe that you know your own needs better than anyone else. You choose to use this moving service, knowing that they are simply going to put your stuff in boxes and bring it, and it shouldn't be interpreted as an authoritative command on The Correct Way To Organize Your Stuff. You need someone to bring your stuff; they bring your stuff- but you are in charge of your own life.
I've been thinking a lot about this kind of mindset where you believe someone else is in charge of you, someone else knows what you need, someone else will take care of you. And all you need to do is generally follow the rules of the system. Versus having a mindset where you are in charge of yourself, and you identify your own needs, and investigate what resources are available to help you meet those needs, and you make the decision yourself.
---
I'm a software engineer, and I've been using an LLM (Deepseek)- an AI tool- at my job sometimes, to help with debugging. I kind of like it and kind of hate it.
There are things that the AI is good at. If I have a very simple question, the sort of thing that can be easily answered by anyone who has experience using the programming language in question, then yes, the AI will probably give a good answer. Recently my husband decided we should make a javascript tool to help our son with studying Chinese. He told the AI basically what he wanted, it spit out some code, then we ran it, and we felt like, yeah, this is a good start. It wasn't totally what we wanted, but it was a good start. And as I've been modifying it, adding features, I often ask Deepseek how to do this or that in javascript. Because I'm not familiar with javascript, but I'm asking things that are easy enough. And Deepseek gives me an answer, and I can read it and understand it, and if it's what I want then I copy the parts that I need into my code.
But you shouldn't trust AI answers when you actually have a difficult problem. At my job I was working with some code related to Android OS, which I had never used before, and oh man, Deepseek really got me off-track. It's like... you type in your complicated problem, you copy/paste your error message into the chat box, and Deepseek will say "Yes! We are finally getting to the root of the problem!" And it will very confidently tell you the next step you should take.
Oh, I hate this so much. In many cases, it's totally wrong, but the tone of it is so confident, like it totally knows what's going on, and it just needs you to do one little step and that will solve your whole problem. And *I* don't understand what the problem is with my code, so it's extremely tempting to want to listen to someone who comes along and says they totally understand the problem. It's psychologically very difficult to take a step back and say, I've been taking Deepseek's advice on this and it hasn't helped at all- this is probably something that Deepseek is just not going to be able to figure out. I need to try something else to solve this problem.
The worst thing is when I read the first line of Deepseek's output- "Ah yes! You have finally identified the real issue!" and then I keep reading, to see what the explanation is, and I can't understand the explanation. My instinct in a situation like this is to read it more carefully, and even if I can't understand the reasoning, at least pick out a few parts that I can try out. This instinct comes from reading material written by humans, who only write such things if they truly believe themselves to have helpful insight and experience. !!! Don't spend time carefully re-reading an LLM's output. This way lies madness! No one actually spent any time writing it, so there's no point to spending extra time to re-read it. If I can read it once and understand it, okay that's a good sign, but if I read it once and I can't understand it, the most likely cause is that it is just wrong.
Learning to recognize that, and step back from it.
But oh man, the idea that someone else is in charge of me, someone else totally understands my situation, when I don't even understand it, and someone is giving me an answer which I can't understand but I just have to dutifully follow. Oh, this is why I often avoid telling Deepseek what my variable names are. !!! When you read some forum post or sample code or documentation written by a human, and you want to try it out and see if it solves your problem, you have to understand it enough to know how the variables it uses would correspond to the variable names you are using in your code. You copy and paste from it, and maybe you don't understand what exactly is going on, but at least you understand enough to know what thing is interacting with what thing. But Deepseek tells me to edit my code in such-and-such a way, and I'm like, I can't even figure out how that would apply to the code that I have- and the temptation is to copy my code into the chat box and say "Okay what would that look like in this code here?" and it will spit out some output, so seamlessly using all the correct variable names, and I won't even have to think at all, I just copy and paste, I won't even have to think about *where* to paste it, or how it relates to the rest of my code- Deepseek writes the whole thing. No need to think or understand, just copy and paste. Someone else is in charge of my life. And then of course I try it and it's completely wrong.
(And then I paste the error message into Deepseek and it will say "Ah, I understand the problem!" and tell me that you can't use this function when you're also using this other function, and then I will say "umm, the code that *you gave me* used both those functions" and then it will say "You're right to point this out!" and very confidently give me more wrong things to try.)
So I try not to let Deepseek see too much of my code. Deepseek, I want to make a button that the user can click, which will then call a function, which will save a csv file- but I'm not telling you the name of the button, the function, or what data we're saving in the csv file. You give me output that I can understand well enough to know where to substitute my own variable names. Okay?
I worry that people are viewing LLMs like they're all-knowing and they can be in charge of our lives in this way. The writing style is so smooth and so positive... like it totally knows what I need, it knows what's right for me, and if I don't understand it, that's because I'm just not qualified to be in charge of my own life. I should just do what it says, even if it doesn't make sense to me.
Whereas, the other way to view LLMs like Deepseek is: This is a tool. I look at my own situation and my own needs, and I decide if this tool might be helpful, and I try it, and then I evaluate the output that it gives, and see if it looks like it's worth using or not. I'm in charge of my life.
So, that's my AI rant. The #1 most important skill you need when using LLMs for software engineering purposes is to recognize when the LLM is just not going to be able to solve your problem, and stop wasting time on it. I haven't seen anyone talking about this- I haven't seen training about it. I've seen people saying "AI is terrible because it's often wrong" and people saying "let's add AI to everything" but who is talking about *how* to recognize when it's useful and when it's not? It's so deceptive how the AI will always tell you it totally understands your problem- you don't need to think, you don't need to understand, these are big technical concepts and you are just adrift in a sea of powerful forces you can never understand, but just be a good kid and copy and paste the code that Deepseek gave you, and everything will be fine.
---
When my daughter was a newborn baby, I knew everything going on with her life. At that age, you can pretty much sum up your baby's status- when did they last drink milk, when did they last sleep, when did they last get their diaper changed- and that's the entirety of what's going on with them. I was able to be the "all-knowing" force in charge of her and taking care of her.
But after some time, the baby will start to have their own personal life, which the adults don't really know about. For example, suppose you're looking for something and you can't find it, but you remember seeing your 18-month-old baby playing with it earlier. You can ask the baby where it is. And often times the baby will toddle off and find it and bring it to you. Yes, at 18 months they can do this- they have knowledge that the adults around them don't even have. They have their own thoughts and experiences. They've become complex enough that you can no longer know everything going on with them and be fully in charge of them.
And my older kid is in elementary school, and I don't know what he does all day. I just have to assume that whatever the school is teaching him is basically fine. I can't vet everything. You can't control every influence in your kid's life; instead, the strategy is to teach the kid how to make good decisions for themself.
He went on a class field trip- the first time the kids went on a field trip without their parents coming along. We packed a lunch for him and gave him some cash to buy a water bottle. And then he spent all the money on a toy, and he was thirsty all day. Every kid will make some bad decisions when they get a little bit of independence for the first time. Hopefully the bad decisions will be minor things that they then learn from, rather than anything serious.
It's unavoidable- I can't hover over him all the time and make sure nothing bad happens. The important thing is to teach him the skills he needs to be in charge of himself.
---
This has also changed how I view doctors. Before, I conceptualized it as, the doctor is in charge of me, the doctor totally knows my situation and will tell me what to do, and I just have to trust them and do it.
But now I see it as, this is a tool. If I have some kind of health issue, then I go see a doctor, with some expectations about what I want to get from them. It means I have to think about how to communicate my situation to them- they *don't* know my situation until I tell them, and they are just humans who might misinterpret what I say. I need to not just naively answer their questions, but also notice if the questions they are asking aren't really getting at what I feel are the most significant aspects of my medical issues.
Having a mindset like... having a general idea of what I want from the doctor. Rather than approaching it like, I am a simple little leaf, blown about by medical problems which are too big for me to understand, so I just need to lay out all my symptoms in front of a doctor, and they will do the work of summarizing what the situation is and telling me what to do.
A while ago, I was going to a therapist, and at one point he sent all his notes to me, and I saw that some of the notes he had written contained slight misinterpretations of things I had said. These misunderstandings didn't change the effectiveness of the therapy- I really did benefit a lot from it- but they did shatter the concept I had of what a therapist *is*. No, he's not in charge of me, he doesn't know the ins and outs of my entire personal life and what I need to do. He's just a person, who has skills and training that can be helpful to people with mental health problems. That's all. This is a tool; I can choose to use it if I feel it's helpful. It's not in charge of me.
I am in charge of my own health situation. The services that doctors offer are a very important tool for my management of my own life.
---
I wonder, is it a sign of privilege, to have access to the resources needed to take charge of your own life and not have to be helpless and dependent on big systems you don't understand? Or is it privileged to not have to think about that at all, to just assume the system will take care of you?
I've heard about people who don't have the money to pay their bills, who refuse to even look at the bills that come in the mail. Or refuse to look at their bank account balance. Sort of trying to ignore the problem and hoping everything will be fine. This has never made sense to me- obviously if you don't even acknowledge there's a problem and try to do something about it, the problem will get worse. But actually, it's a very different thing to say to yourself "let's collect the data on how much money I have in each account, my average monthly income and expenses, and let's make a spreadsheet and plan it out" when you actually have enough income that you *can* make a budget that feels good, compared to if your income is too low and no amount of "taking charge" attitude is going to help. Feeling like you're at the mercy of big powerful systems you can't understand because it's actually kinda true. Why even bother to read your bills? Maybe it feels like the best option is to do nothing and just hope it will be fine.
Or, what if you go to see a doctor about a medical problem, and the doctor doesn't believe you and doesn't take you seriously? I can see how that might feel like being tossed about by systems too big and complicated for you to do anything about.
Or if you're looking for an apartment to rent, and you found one but there are some things about it that seem like red flags, like they could be signs that there are going to be problems living there- but the rent in your city is so expensive, you can't afford any red-flag-less apartment. All you can do is sign the contract and just *hope* that there won't be problems. If you had more money, you could actually *do something* about it, rather than just *hope*.
So it is a privilege, to have a good understanding of what these systems and resources can do for you, and to have the money and resources to access the parts that you decide would be helpful for you.
But also, it can be the case that less privileged people are the ones who understand most clearly that the systems don't work for them, and the necessity of taking charge of one's own life. I've read about people who have had problems with doctors not taking them seriously- and they develop strategies for what to say, what to wear, etc. They know that the system won't just take care of them, and the only way to get any help from the system is to have a brutally honest understanding of what the system *is* and its inputs and outputs, and then strategize based on that. You are in charge of your own life.
So it's not necessarily "more privileged" or "less privileged" to believe there are all-knowing authority figures who will take care of you and totally understand you and should be in charge of your life. It's just naive.
---
The idea that someone else knows what I need, better than I know myself, just because they're an expert in some area (medicine, packing boxes, predicting the next token, etc). The idea that I'm not really supposed to understand what's going on, and I'm not supposed to choose for myself- I should just be good and follow the rules and trust that everything will be fine.
No, it's just not true. You have to be in charge of your own life.
---
Related:
From "Virtues Morality" To "Boundaries Morality"
When the Teacher Says, "Don't Look at Your Report Card"
In Which John Piper Doesn't Tell You To Find A New Job
I Figured Out What The 1-10 Pain Scale Is Actually About
I Don't Want My Baby To Be "Brave"
Honest Lent: "Seek First God's Kingdom" Doesn't Work If You Have Autism
Cut Out the Middleman (or, why I am the master and commander of my own life)
No comments:
Post a Comment