>I'd say smartly used LLMs make for better choices because you can do research much more easily. Benchmarks, prototypes, asking it to search for you whether a given library has a good reputation or not and is well maintained, etc.
Honestly I agree with this take. The problem is that, under the context of the people banned for using LLMs, no one makes decisions like this.
For instance, say I want to do some data analytics on a pretty large csv file. Of course before AI I would need to think of 1.How to open the CSV file 2. What is the specific data I want to extract 3. How to extract the data 4. How do I change the data into a specific format (say a list, a set etc ..) 5. How do I sort it. 6. How do I visualize it .... and so on. Those kinds of steps on how to write the code forces you to research the libraries and methods and decide what is required. (for instance using pandas in python)
Using AI for that task however, I wouldn't have done any decision for how to extract the data in question. Maybe the AI's decision was marginally better than mine (Writing a csv parser from scratch is faster for that specific data for example). Maybe in the case of data analytics, there is no point in the code cause in the end what is desired more is the visualization. But dumping the code into communities where people were interested in how you came to that decision falls short cause you didn't do it in actuality. I think that is the current issue at hand.
Honestly I agree with this take. The problem is that, under the context of the people banned for using LLMs, no one makes decisions like this.
For instance, say I want to do some data analytics on a pretty large csv file. Of course before AI I would need to think of 1.How to open the CSV file 2. What is the specific data I want to extract 3. How to extract the data 4. How do I change the data into a specific format (say a list, a set etc ..) 5. How do I sort it. 6. How do I visualize it .... and so on. Those kinds of steps on how to write the code forces you to research the libraries and methods and decide what is required. (for instance using pandas in python)
Using AI for that task however, I wouldn't have done any decision for how to extract the data in question. Maybe the AI's decision was marginally better than mine (Writing a csv parser from scratch is faster for that specific data for example). Maybe in the case of data analytics, there is no point in the code cause in the end what is desired more is the visualization. But dumping the code into communities where people were interested in how you came to that decision falls short cause you didn't do it in actuality. I think that is the current issue at hand.