I still dislike working with AI in coding, but Ive learned to control it and when it cant help.
As it can take more and more data and have longer session lengths and dynamically ditch old chats, even this dislike could become just acknowledgment of limitations. Whereas now its like sticking your hand into an alligator mouth to get treasure. You just lose time and patience when you get bit.
The biggest problem is now at certain complexity it will only destroy elements, but this should be reduced, though will never go away with LLM tech.
But so many things I wouldn't have tried and completed without it because they would have taken too long.
Its need for direction is where the real divide is looming. Busyworkers and hacks being ultimately replaceable since AI slop is a lot cheaper than their slop. But a force multiplier for actual creatives.
The mirror metaphor has a lot of facets. Shining truth back on abilities is an unexpected one.
The very competent will use the tool wisely. The malevolent will destroy entry level/apprenticeship work and eat the corporate seed corn. Big corporations are generally malevolent.
I wrote a piano piece. I used to dream of doing a pyro musical to it. I'd talk about how awesome it would be if booms went off in the sky right here, or i wish i could get flutes here and brass here. Until a month ago, this was nothing more than wishful thinking. Now. I will have it exactly the way I heard it in my head when i came up with the melody and it's gonna be awesome. I cannot wait to do this next year on the 4th. I could never do this. I dont have the skillset, but making music with ai has unlocked a deep level of creative thinking that I never knew was there. I was blown away when I heard my tune, to a symphony orchestra, and it was great.
I was pleasantly surprised, Vox, when you began to experiment with AI for writing. Then, and now, the broad consensus has been that AI is fun, but not serious. You can use it to play, but it can't actually offer improvement over what humans produce. When I tried using Claude to write, using an idea from my youth where I had started and stopped multiple times, but never quite got into the flow, suddenly my idea was coming to life.
Like you, I've had to refine my approach, learn how to use the tool properly, but yes, I was able to collaborate with Claude and output at least 10x what I could do alone in the same span of time, with the end result more polished than I could have done on my own. Now if I could just get Claude to nag me about wanting to write more. I've probably got at least a three novel series in me for that story.
I too just started using Claude, but I had been using a different AI tool for some time now.
Each has their strengths and weaknesses; it's like a carpenter having a number of different saws each with their own function.
I used to be a skeptic until I started trying to do actual work. I don't have to be a software designer, coder, PCB developer, aerospace engineer, materials specialist. I supply the idea, AI does the heavy lifting, and I check the work by having multiple AIs "peer review" each other's output.
For those who are able to see it as a series of tools it will prove to be revolutionary.
Yes; it's easy to recognize if something is off in one's area of expertise but impossible for areas outside one's knowledge.
I have noticed too that there's an element of professional rivalry if when presenting work from one AI to another you mention it came from their rival. Oftentimes I get responses on how it could be improved further or if a product is unnecessarily complex.
As to why I was having trouble moving forward, the story was set in 1937, on an alternate Earth, where the discovery of precognition by a German medical researcher leads to a psionic arms race between Germany, the US, Russia, and Japan, with World War II unfolding very differently, now that the Reich can see glimpses of the future.
I needed help fleshing out the concept. What I had in my head was a few scenes, and a concept that was even less developed than the summary I just wrote. I pitched the idea to Claude and we worked out the arc of the first novel iteratively. We created characters and fleshed them out, and set up plot elements that would show up again in future novels in the series.
What I'm probably most excited about is seeing video production start to take off. I would love to produce a web series out of this story.
I was telling my friend the other day, lucky I part own the company, because I would have fired myself by now otherwise. I can do almost all my leg work for technical development in hours now what used to take weeks just by having my 19 dollar subscription to GPT. It is INSANE, how effective it is if you know when it is BSing or not.
"We are really going to need figure out how to integrate people people new to a field and work place."
Having them study a senior engineer's prompts and the AI responses would teach them a lot. A version of shadowing that is far more accessible since all those interactions can be easily logged.
If you say your neurodivergent and you subscribe Iām guaranteed going to follow back š
If you say your neurodivergent and you subscribe Iām guaranteed going to follow back š
I still dislike working with AI in coding, but Ive learned to control it and when it cant help.
As it can take more and more data and have longer session lengths and dynamically ditch old chats, even this dislike could become just acknowledgment of limitations. Whereas now its like sticking your hand into an alligator mouth to get treasure. You just lose time and patience when you get bit.
The biggest problem is now at certain complexity it will only destroy elements, but this should be reduced, though will never go away with LLM tech.
But so many things I wouldn't have tried and completed without it because they would have taken too long.
Its need for direction is where the real divide is looming. Busyworkers and hacks being ultimately replaceable since AI slop is a lot cheaper than their slop. But a force multiplier for actual creatives.
The mirror metaphor has a lot of facets. Shining truth back on abilities is an unexpected one.
The very competent will use the tool wisely. The malevolent will destroy entry level/apprenticeship work and eat the corporate seed corn. Big corporations are generally malevolent.
I wrote a piano piece. I used to dream of doing a pyro musical to it. I'd talk about how awesome it would be if booms went off in the sky right here, or i wish i could get flutes here and brass here. Until a month ago, this was nothing more than wishful thinking. Now. I will have it exactly the way I heard it in my head when i came up with the melody and it's gonna be awesome. I cannot wait to do this next year on the 4th. I could never do this. I dont have the skillset, but making music with ai has unlocked a deep level of creative thinking that I never knew was there. I was blown away when I heard my tune, to a symphony orchestra, and it was great.
I was pleasantly surprised, Vox, when you began to experiment with AI for writing. Then, and now, the broad consensus has been that AI is fun, but not serious. You can use it to play, but it can't actually offer improvement over what humans produce. When I tried using Claude to write, using an idea from my youth where I had started and stopped multiple times, but never quite got into the flow, suddenly my idea was coming to life.
Like you, I've had to refine my approach, learn how to use the tool properly, but yes, I was able to collaborate with Claude and output at least 10x what I could do alone in the same span of time, with the end result more polished than I could have done on my own. Now if I could just get Claude to nag me about wanting to write more. I've probably got at least a three novel series in me for that story.
I too just started using Claude, but I had been using a different AI tool for some time now.
Each has their strengths and weaknesses; it's like a carpenter having a number of different saws each with their own function.
I used to be a skeptic until I started trying to do actual work. I don't have to be a software designer, coder, PCB developer, aerospace engineer, materials specialist. I supply the idea, AI does the heavy lifting, and I check the work by having multiple AIs "peer review" each other's output.
For those who are able to see it as a series of tools it will prove to be revolutionary.
Multiple AIs "peer review" each other's output. Useful suggestion.
Yes, definitely. Claude was getting some minor historical events wrong, and tried to create a female FBI agent in 1937.
Might have just been a cross dresser, given the org.
Haha. Supposedly that one is a myth, but the guy had enough issues to fill a newsstand.
Yes; it's easy to recognize if something is off in one's area of expertise but impossible for areas outside one's knowledge.
I have noticed too that there's an element of professional rivalry if when presenting work from one AI to another you mention it came from their rival. Oftentimes I get responses on how it could be improved further or if a product is unnecessarily complex.
As to why I was having trouble moving forward, the story was set in 1937, on an alternate Earth, where the discovery of precognition by a German medical researcher leads to a psionic arms race between Germany, the US, Russia, and Japan, with World War II unfolding very differently, now that the Reich can see glimpses of the future.
I needed help fleshing out the concept. What I had in my head was a few scenes, and a concept that was even less developed than the summary I just wrote. I pitched the idea to Claude and we worked out the arc of the first novel iteratively. We created characters and fleshed them out, and set up plot elements that would show up again in future novels in the series.
What I'm probably most excited about is seeing video production start to take off. I would love to produce a web series out of this story.
Well heās smart but he is also an atheist
What does that have to do with anything?
Because everyone interprets the world through their beliefs, this does not invalidate his argument just worth take into consideration
The best argument against it would be if it failed to work or was immoral.
It is clearly demonstrated it works. Right use will do more, faster in many fields.
The morality comes down to use. Used for crime, bad. Used to create, good.
"Tools cant have morality" applies to Guns and AI.
I was telling my friend the other day, lucky I part own the company, because I would have fired myself by now otherwise. I can do almost all my leg work for technical development in hours now what used to take weeks just by having my 19 dollar subscription to GPT. It is INSANE, how effective it is if you know when it is BSing or not.
Unpaid interns btfo
"We are really going to need figure out how to integrate people people new to a field and work place."
Having them study a senior engineer's prompts and the AI responses would teach them a lot. A version of shadowing that is far more accessible since all those interactions can be easily logged.
They need to make their own projects.
They have every tool at their disposal.
I love it. Yes, we're taking back the culture, and using the tools that they are disdaining to do it.