I like to use AI as a sort of brainstorming tool. For example, if I have an idea for a software system, I will describe the general idea and ask for three different possible designs. I will tell it not to generate code but to provide three ideas with pros, cons, and tradeoffs. What it provides will often trigger some new ideas that take me off in a different direction. I often mix and match aspects of the ideas it provides with my own. I keep iterating on this till I think I have a design I like.
I use it to explain and explore things I don't understand and I will have it give me code examples in multiple languages, explain the benefits and differences which each language offers for the task, and ask questions about each. I stick to the ChatGPT free plan and when I hit the usage limit for the top model, I close the window and switch to writing code. by the time my limit resets I will have much better and more focused questions to ask but generally I wait until I have a real reason to go back, bang my head against the problem for awhile.
I have found that it is much better at answering questions if you start with code it wrote instead your own code or someone else's code, so I boil my question down to a simple programming task and start by having it write that code. For example, there were some things I was unsure about with VMs/bytecode interpreters/compilers, so I started my session by asking ChatGPT to write me a simple Forth VM in C and then used that as the jumping off point.
Spending time facing a problem and having no idea how to solve it is an incompressible time in the process of learning/understanding something. So I think it could be: 1) forcing yourself to think without any help, 2) asking AI after some time if no solution comes up to you (and there is a certain time constraint to deliver), 3) understanding perfectly the AI produced solution, no matter the time it takes.
Introduce friction. Use it only through an API, and replies are emailed after a day. Have it refuse to answer questions missing a minimal repro case. Let it chasten you if the same question was asked before (within a year or month).
Request quizzes or hints before showing the answer. Have it go through a Socratic dialog style, where you can reply, "Figured it out" to abort.
Kind of funny sounding, but you could ask AI some tips on how to do this...
I think naturally some output from AI will be incorrect and this forces people to question the output constantly, so I'm not sure if there is much actual current risk of such a loss of skills
I believe people will lose their critical thinking skills, when they use AI blindly. May be we should write the code for a requirement, and then ask the AI to review and suggest any optimizations, so that we could probably learn the mistakes we make.
- use it in "learning" mode
- invest time in doing things without AI - to improve the speed of doing things with AI. Training mode (without AI) vs Performance mode (with AI)
this a logical doom loop.
are you are a fucking punch press operator, or something else, worse?, such as a temporary hire untill the punch press no longer requires you.
my advice would be to pay attention to punch press design and optimisation, and keep your fingers (mind) clear of the works and keep snagging the custom work that cant be fully automated.
I like to use AI as a sort of brainstorming tool. For example, if I have an idea for a software system, I will describe the general idea and ask for three different possible designs. I will tell it not to generate code but to provide three ideas with pros, cons, and tradeoffs. What it provides will often trigger some new ideas that take me off in a different direction. I often mix and match aspects of the ideas it provides with my own. I keep iterating on this till I think I have a design I like.
I use it to explain and explore things I don't understand and I will have it give me code examples in multiple languages, explain the benefits and differences which each language offers for the task, and ask questions about each. I stick to the ChatGPT free plan and when I hit the usage limit for the top model, I close the window and switch to writing code. by the time my limit resets I will have much better and more focused questions to ask but generally I wait until I have a real reason to go back, bang my head against the problem for awhile.
I have found that it is much better at answering questions if you start with code it wrote instead your own code or someone else's code, so I boil my question down to a simple programming task and start by having it write that code. For example, there were some things I was unsure about with VMs/bytecode interpreters/compilers, so I started my session by asking ChatGPT to write me a simple Forth VM in C and then used that as the jumping off point.
Spending time facing a problem and having no idea how to solve it is an incompressible time in the process of learning/understanding something. So I think it could be: 1) forcing yourself to think without any help, 2) asking AI after some time if no solution comes up to you (and there is a certain time constraint to deliver), 3) understanding perfectly the AI produced solution, no matter the time it takes.
Introduce friction. Use it only through an API, and replies are emailed after a day. Have it refuse to answer questions missing a minimal repro case. Let it chasten you if the same question was asked before (within a year or month).
Request quizzes or hints before showing the answer. Have it go through a Socratic dialog style, where you can reply, "Figured it out" to abort.
Kind of funny sounding, but you could ask AI some tips on how to do this...
I think naturally some output from AI will be incorrect and this forces people to question the output constantly, so I'm not sure if there is much actual current risk of such a loss of skills
This if you do have some critical thinking to start with, for the less gifted ones LLMs only make things worse.
I believe people will lose their critical thinking skills, when they use AI blindly. May be we should write the code for a requirement, and then ask the AI to review and suggest any optimizations, so that we could probably learn the mistakes we make.
I paste URLs to docs into Gemini and ask it to generate quizzes.
- use it in "learning" mode - invest time in doing things without AI - to improve the speed of doing things with AI. Training mode (without AI) vs Performance mode (with AI)
Technology gives and technology takes.
How can I sit in a chair all day and not ruin my cardiovascular health?
I think the answer is, you cannot.
this a logical doom loop. are you are a fucking punch press operator, or something else, worse?, such as a temporary hire untill the punch press no longer requires you. my advice would be to pay attention to punch press design and optimisation, and keep your fingers (mind) clear of the works and keep snagging the custom work that cant be fully automated.
John Henry
https://m.youtube.com/watch?v=uYh9MwKQnb8