I have some further thoughts about these GPT-LLMs everyone is talking about. Mostly I’m talking about Bing, Bard, ChatGPT, and I suppose some other things (there are so many other things). See my earlier ramble about it here.
I do think it likely, greater than 50% chance my gut tells me, that we’re getting into another technological shift. That already feels quite bold, but I suspect that if anyone reads this they might think I am lowballing the odds. Maybe it’s even 51%, heh. Hey fella, it’s my gut that’s telling those odds, I’m just reporting what it says. I don’t even really know how reliable it is.
By technological shift, I mean something similar to, I don’t know, Wikipedia, or cell phones, maybe even early Google search. FourSquare, Twitter in the before times, or Facebook, but not social media. New vistas await, possibly big vistas. Or maybe it’ll just be a better Siri, or predictive typing suggestions for your phone. There will be some minimal amount of technological progress capture from this.
I think though, first things first, we (me, you, us all) don’t need to do anything hasty here. None of those career doomsayers have a crystal ball any better than yours or mine, even if they have a little more information than either of us. They just have some new dart that they’ve thrown at the wall of humanity. Maybe it will stick. Maybe it won’t. Maybe it’ll just dangle.
Here are some of the challenges I see. In the current iteration, these things are very casual with the truth. They are creative with facts, and you don’t have to try hard to get them there. This could be a fool me once situation. By that I mean that we, the users, feel tricked by this early iteration and take a step back for a while.
But maybe it’s good at helping you edit your own set of facts, editing something that you’ve written. I wouldn’t know, I haven’t tried that. I want to write my own words. Maybe I’ll try a human editor someday, if they haven’t been replaced.
There are also confidentiality concerns, liability for these errors, and just flooding the space with drivel which might poison the well. That last is only about as overblown as those people yelling about the immediate reshaping of the global economy. It’s just sleazy hype.
I’m also not sure how well it will be made to work in practical situations. In real, messy settings, and in more regular situations. Like forms at the DMV, or at the county clerk’s office. I’m sure they’ll get the tax forms, that’s too big of an apple to not take bite. Anything smaller than that, some Wasco county building permit form, I bet it’ll trip up. They’ll leave that for those county folks to deal with.
Somehow making this more complicated is not the way, and “prompt engineering” just sounds more complicated to me. Specialists will be able to make some hay with a complicated tool, but a true revolution is not what it is yet. For some folks it’ll even be a kind of fun. If it doesn’t exist already, I see prompt golfing games around the corner.
And yet, some change seems likely to land. ChatGPT has been intriguing, maybe even fun to play with. It surprises you. Don’t know what that change will look like. So keep your eyes open
Feel free to reply to this toot if you want to.
To Reply: Email me more thoughts.
Posted: in Notes.
Other categories: garo.lol.
Back references: none.
Tags that connect: [[AI]] No Government in the Cloud, Initial thoughts about LLMs.
Tags only on this post: none.