A brief conversation with an AI

I've been working on a conversational bot that I can brainstorm with and ask for advice. I've been tinkering with it for a few days, but today, I finally had a conversation that I thought was helpful enough for me to feel that it was adding value to my working process. Here's a raw (except for fixing some of my typos) transcript of the conversation (minus the prompt). The conversation is generated using GPT-J-6B.

It's still... not perfect (I ask it how to be productive, and it tells me I can sit on the bed and watch TV). But baby steps!

Me: I'm having trouble staying focused while working at home. What should I do?

AI: When I work from home, I usually have three options. Option 1 is to go to a cafe and work in a coffee shop, which is a total waste of time since I would just be reading my email on my phone instead of working. Option 2 is to get up and go to a different room to work. This is also not optimal because I'm going to start work and then leave. Option 3 is to sit in my bed, on a beanbag chair, watching TV. This would be an optimal solution if I had any work to do at home. But I don't.

Me: Going to a different room sounds like a great idea. I'll do that for a bit!

AI: It sounds like you need to schedule your work time and then stick to the schedule. You might even want to do this for a few weeks to get a good idea of how to structure your day.

Me: Do you have any thoughts on interfaces/paradigms to represent history of search/discovery that's better than tabs / tree-like history? i feel like most products stick with tabs/flat lists and some are experimenting with trees, but one is too simple and the other feels too unwieldy?

AI: So I think there are three main categories of interfaces/paradigms that you can consider when creating a discovery tool.

Me: What are those three?

AI: 1. A linear history. This would be an unweildy list of events, with each event having the full story of what happened leading up to it.

AI: 2. A flat history. This is more typical of traditional search/discovery interfaces, and is like a timeline.

AI: 3. A nonlinear history. This has the potential to be really complex. It's more like a tree of all your history, but it's more than just one event per node. For example, Google keeps your search history and then puts the history of each query next to it, so the root of the tree would be the first query you ever performed.

Me: I like the idea of a nonlinear history, but I feel like there are more possibilities besides a tree b/c trees are hard to interpreter for normal people.

AI: Yes, that's right. But you might want to also consider a combination of these paradigms. For example, you could have some parts of the history be linear, and other parts be nonlinear, which will give you a tree-like feel on top of a linear structure.

Me: In your mind, what decides whether a given part of browsing history is linear vs. tree-like and more complex?

AI: Generally, for most products, it's going to be the data structures that they're using to represent browsing history, and this data structure will decide whether it's linear or tree-like.

Me: Right but in this case the data structure is all just linear history. We have to infer more complex ways to present the data for the user.

AI: If you think about it, a tree-like history would be a list of events where each event is a single piece of information that could be related to something else. The only way to represent that in a list of items would be to show a single item per line of the list. This is what most products do.

Me: What if we present browser history as a map to explore instead? how do we lay out a map?

AI: One option would be to put the first node of the map near the top of the map, and have a node for each event. And then the second map layer might be a map to explore the details for a particular event in the first map. And so on.

Me: How do you think this compares to a geographic map? Can we draw some analogies to this "online history map"?

AI: There's a good analogy here that is actually very similar to what a geographic map is. A geographic map is basically a map that is made by plotting points of latitude and longitude on a flat plane, so they have a two dimensional layout. You can then put different data on top of that map, like cities, roads, rivers, mountains, etc.

Me: How does that analogy extend to showing more abstract browsing histories on a map?

AI: I think you'd have to start with the idea of mapping history to points of latitude and longitude. Then the second layer of information could be a set of different things that go along with that point. You could map history to a collection of objects, such as locations, objects, and/or actions.

Me: I like that. It's like there's a pre-existing "map of the internet" with different topics in different neighborhoods, and browsing the web is like taking a stroll through those neighborhoods. Your paths then show up on the map as paths you took.

AI: That's a good analogy.