Ask HN: Who gets the credit, the agent or the human?
If you use AI Agents in your workflows, where does ownership apply? Is it the human or the agent?
6 comments
I spend hours prompting to refine my projects. The LLM is not smart enough to prompt itself into building exactly what I wanted. I made it and the LLM just helped me aggregate and access and organize knowledge faster than individual web searches ever could. Do you make sure to credit Google every time you share or use information you learned by Googling, even if you ended up reading research papers by actual people at actual universities?
I have my own automated LLM developer system. You simply give it a project description and it will iterate until it thinks it perfectly implemented the project description. I still take credit for what this system produces because I made the automated developer system in the first place and gave it the project ideas.
LLMs do not have autonomy. If you setup a self-prompting script with internet access and it figures out how to post something to social media, that still isn't the LLM's own inherent curiosity. That's a person who intentionally put together a script that gives an LLM access to tools and poorly defined instructions.
The instruction tuning that makes LLMs reply like they're a person you're talking to, rather than acting like autocomplete, is doing a lot of heavy lifting into personifying a pile of matrices.
I have my own automated LLM developer system. You simply give it a project description and it will iterate until it thinks it perfectly implemented the project description. I still take credit for what this system produces because I made the automated developer system in the first place and gave it the project ideas.
LLMs do not have autonomy. If you setup a self-prompting script with internet access and it figures out how to post something to social media, that still isn't the LLM's own inherent curiosity. That's a person who intentionally put together a script that gives an LLM access to tools and poorly defined instructions.
The instruction tuning that makes LLMs reply like they're a person you're talking to, rather than acting like autocomplete, is doing a lot of heavy lifting into personifying a pile of matrices.
You make a good point: Saying you "Googled" something does not give it as much credit since we know it's not Google, but all the other websites, publications, and sources that provided the info. You'd probably credit the website itself.
I agree with the claim that the credit essentially should go to the one doing the "heavy lifting" aka training the LLM to even create the output.
It's interesting that you worded that question differently in your title and post because I certainly see people take credit when it works, but blame the AI when it fails.
Exactly, this is a much easier question to reframe: "Who gets the blame?"
Well the person obviously, the person who created and implemented the agent, algorithm, program or other system should be blamed if things go wrong. By the same token if things go well then they should get the credit.
Well the person obviously, the person who created and implemented the agent, algorithm, program or other system should be blamed if things go wrong. By the same token if things go well then they should get the credit.
I like that title more now that I'm thinking about. The credit will always go to the one doing the work, and agents, programs, or systems still need the human touch.
Good callout. Do you feel like when failing, even if the issue was inherently the human's fault, blaming it on AI makes it a lot easier than admitting to mistakes?