Raw LLM Responses

I'm noticing something similar, but it's likely because we're both operating in the data science/machine learning space. A lot of the subtleties of model development and measurement really can't be done by an LLM. You basically have to know where you're going to get there. Where these AI tools are valuable is that they really do begin to eliminate the need to remember the minor technical details. Let me give you an example. We're working on a project and we brought a new data scientist who wasn't familiar with coding in pyspark. He's learning, but he's been able to become extremely effective quickly simply by making declarative statements and having the AI write the pyspark code The same would work for someone who was a data scientist who only knew R or SAS and was transitioning to a python shop. Because they understand what the end result should look like they can make declarative statements to the AI to write the code. They could just write the code in R and tell the AI to convert it into python. The outcome of this would be that in a few years we shouldn't have those hard requirements that you must have X number of years experience with python or pyspark or R, etc, but you still need the fundamental understanding of building good models and designing experiments. You need to be able to do feature selection beyond simply by picking the most important features and eliminating those that are highly correlated. We're so backlogged with projects that we'd like to approach I'm hoping that this push to AI increases our productivity to a point where we can actually complete a lot of those projects and provide more value to the company.