Our T-SQL Tuesday topic for this month comes from Damien Jones who asks us about our wishlist for Microsoft OpenAI. AI has been a hot topic in the last few months. Some see it as the next best thing while others are a bit more fearful. I think the answer is somewhere in the middle. To read the full invitation post, click the T-SQL Tuesday link to the right.
Microsoft OpenAI and PowerShell
PowerShell with some built-in OpenAI functionality could be a time saver for all kinds of roles. Imagine needing a script for automating a repetitive, mundane task. I’d be happy to see an option built-in where you could make a request for the script, have PowerShell give you a solution, and include -WhatIf so you can test and have better idea about what the solution would do. Just like ChatGPT can be too confident in stating “facts” that are not true, you can’t blindly run a script and expect it to do what it says it’s going to do. How many workers are going to blame an outage they caused on code that an AI generated for them?
Good Enough For Basics
Not being perfect doesn’t mean it’s something to be scrapped altogether. It just means we need to be cautious and use it for the right circumstances. I like how some have compared ChatGPT to self driving cars. There have been significant improvements so far with self driving modes. If you’re commute to work was 95% interstate like mine was before being remote, you would barely have to bother manually driving at all. This type of self driving has been pretty successful. On the other hand, if my drive was in a big city with traffic lights, city construction, and other abnormal driving conditions, self driving could be downright dangerous.
The point is, just because something isn’t “all the way there” yet doesn’t mean it’s not helpful right now. Maybe with OpenAI will get there where you can trust anything it returns to you. I’m guessing it won’t. But that doesn’t mean we can’t take advance of using it for the more basic tasks right now.
Don’t Be Like This
The security risks make me more nervous than someone copy and pasting code that they got from an AI. More articles like this are popping up, mentioning workers putting confidential company information into ChatGPT. Assume a gun is always loaded, assume a conference call mic is always live, and assume whatever you’re entering online is public.
Thanks for reading!