I remember the initial Google Workspace Gemini ads showing someone writing an email from a few bullet points and blowing it up to pages of text. Then another scene in the same ad has a user summarize pages of text into bullet points.
This is by far the worst and most expensive inverse-compression system ever.
I have never corresponded to another human with an AI response, I think just going through the process of doing that is the most fucking embarrassing thing you could do.
At work people have started just pasting AI responses into slack threads. PR review comments are also entirely AI written. I can understand having your AI review your code before you share with others for review (I would rather a bot point out my stupid bugs/typos instead of wasting someone else’s time on them), but if you are reviewing something YOU need to review it.
I’m also getting annoyed because one guy keeps submitting entirely AI written PRs. They’re awful. Too many comments that say nothing, they don’t match our code styles, there are obvious refactors that would make the whole thing cleaner and more readable. It’s just fucking lazy and a waste of time to keep telling him to fix it, because by the time I point out all the issues I could have just rewritten it myself.
Initial studies have found that AI generally helps either the top performers or the bottom performers of a job/profession. If it helps the top the most, it usually means that the person has to know enough on how to use the AI tools and correct known issues. If it helps the bottom the most, switch fields since it is likely that the AI is going to push your role into whomever is cheapest.
Here’s a use case: you load an LLM instance with a large amount of highly specific factual data over the course of several weeks (ingestion of a large number of documents, daily KT sessions, call with screen sharing transcription, docs artifact generation), and use that LLM to generate answers to people’s requests under high time pressure (urgent project deadline) which is prohibitive to even type the message, nevermind to comb though the data which you’d have to cram in your head, then never use again. Both sides are aware it involves AI. I have used this process against a control group (similar tasks, no AI use) and the result was clearly superior.
I remember the initial Google Workspace Gemini ads showing someone writing an email from a few bullet points and blowing it up to pages of text. Then another scene in the same ad has a user summarize pages of text into bullet points.
This is by far the worst and most expensive inverse-compression system ever.
I have never corresponded to another human with an AI response, I think just going through the process of doing that is the most fucking embarrassing thing you could do.
At work people have started just pasting AI responses into slack threads. PR review comments are also entirely AI written. I can understand having your AI review your code before you share with others for review (I would rather a bot point out my stupid bugs/typos instead of wasting someone else’s time on them), but if you are reviewing something YOU need to review it.
I’m also getting annoyed because one guy keeps submitting entirely AI written PRs. They’re awful. Too many comments that say nothing, they don’t match our code styles, there are obvious refactors that would make the whole thing cleaner and more readable. It’s just fucking lazy and a waste of time to keep telling him to fix it, because by the time I point out all the issues I could have just rewritten it myself.
Yep, that’s it. What a waste.
My hypothesis is that the least competent people use AI the most, and it’s a big multiplier on their incompetence.
Initial studies have found that AI generally helps either the top performers or the bottom performers of a job/profession. If it helps the top the most, it usually means that the person has to know enough on how to use the AI tools and correct known issues. If it helps the bottom the most, switch fields since it is likely that the AI is going to push your role into whomever is cheapest.
This isn’t about who it helps though, this is strictly about usage.
Yes, this article isn’t. I’m talking about another study.
Also the ones whose bosses give them incentives (or direct ordered).
I do not think you’re wrong.
This is why government LOVES AI.
You should just automate closing his pull requests with similar AI gibberish, and gaslight him if he complains.
Here’s a use case: you load an LLM instance with a large amount of highly specific factual data over the course of several weeks (ingestion of a large number of documents, daily KT sessions, call with screen sharing transcription, docs artifact generation), and use that LLM to generate answers to people’s requests under high time pressure (urgent project deadline) which is prohibitive to even type the message, nevermind to comb though the data which you’d have to cram in your head, then never use again. Both sides are aware it involves AI. I have used this process against a control group (similar tasks, no AI use) and the result was clearly superior.