Augmentation Not Replacement

“I haven't completely offloaded any tasks for it. I pretty much just use it to augment what I do.”

“Use it as a learning tool and not a do it for me tool. If I can tell that you've clearly just thrown it in there and said, "Do this for me," then you kind of lose that personal credibility in my eyes.”

“The way I describe it is that for a research activity that would take a researcher alone five days to complete, if you look at it with AI alone, it might take a day, but in order to do a good job of it, the necessary human AI interaction, you might get closer to three days.”

“I've been looking out for, I guess, the effect that people discuss about how it kind of detracts from your own thinking or your own creativity to rely on AI to produce outcomes. And I don't think that's happened to me. And I think if it indeed has not happened, I think it may have something to do with how I see the AI interactions.”

“It's not sort of the replacement, but as an assistant, kind of a sounding board, if you will.”

“In a way, we've created an information environment where we need it. We need AI. And I really think that that's the biggest promise of it, is to stop using it as a potential replacement for humans and use it as a way for us to manage this infosphere that we've built ourselves.”

“I really don't trust a synthesis or analysis and synthesis done entirely by AI. But even with the overview, I think it misses a lot of important stuff that's between the lines. And by the way, it's not just that. I think it's part of the process for us as researchers to immerse ourselves in the data. If we skip that, we don't understand the data afterwards, and we are not retaining that important knowledge that is kind of layering in the back of your mind.”

“The problem is that that kind of analysis gives you tunnel vision. So you don't get the context in which that is said. You don't get if they said it before or after something else. Because the moment you code in a sequence, you follow the conversation, you follow the flow. There is some logic behind it.”

“So I don't think it impacts the way I'm thinking. It's just helping me work with my thinking, or articulate what I'm thinking, or challenging what I'm thinking. So in that sense, it's not the thinking, it's probably the how I work that changes.”

“it's like working with a partner because it gives me ideas that I couldn't really figure on my own, different insights, but of course I have to double check everything. So that's actually a good example of guardrails. Before, I used the same tool to create personas for users of a data cataloging tool that we're looking to buy. So I gave all the interviews to the AI and said, create these personas, and it created five great personas, but it wasn't based on the data, it was based on general knowledge.”

“I'm going to run it through AI first and see if it comes up with some starter ideas instead of me doing a whole exploration. And it came up with something that I thought was pretty neat, and I just rebuilt it in Illustrator and gave it more depth and just more human touch, if you will.”

“that's not something that I can go in and type to Copilot, "Find me a molecule to replace." You have to use it as a tool that augments what you do.”

“Everything we're doing is building the plane as it flies. And I have work in Figma which hasn't been fully translated into our product. And so that work is still there to actually do those refinements, and even to truly implement the designs from Figma into code while we're still building out new features and whatnot. And so sometimes I'm using it to do the work that a front-end engineer might do to clean up our implementation.”

“So we are using it a little bit for that too. But it hasn't gone back into Figma yet. I feel like that's still a work in progress. And then there are times like when I was taking that LLM kind of experience, the chat-based interface project, and I spent just a couple days just working on that and I was really designing as I was building it because I had my engineer's work to start from so I was refining their work, I was cleaning up what they had done. But there would be times when I'd give maybe a general prompt and the output, maybe 50% of the output worked and 50% didn't. So, I say, "Oh, that's a good idea. We'll keep that, but then change these five things." And it's just kind of like an iterative building process. There have been other times where I'm like, "Okay, I'm going to try and use Figma Make because I haven't used it very much" and I'll give it an idea that I'm working on and the output just took a while and it's not helpful at all.”

“So it would need access to GCP and Azure. It would need access to our Bamboo and the other tools that are in the infrastructure that are needed for this production so that it could do the full cycle. So right now I'm the monkey in the middle. I tell it what to do in the tangible code. It does it. Now I have to run the test, take the logs from it, and feed it back and say check the logs, did it work or did it not? So I'm really in the way of the velocity.”