And now we talk to oh, when we get into phone trees, often we're talking with these bots, they're getting much better and it's not frankly so much of an improvement of the software as an improvement of the hardware. Like it's just processing more and they've got bigger data sets. But what I keep being preoccupied with is the limit of those data sets
Romie herself
@Romiesays · 3:55
So I was defending myself like whoa, whoa, it wasn't that bad. And they thought that they were arguing with me because they thought I still didn't think I'd done something that was mistaken. And I feel like I run into that all of the time and maybe that's even an attraction of the idea of, like, oh, we could get robots and there's going to be no emotion, so there's going to be no misunderstandings
Romie herself
@Romiesays · 2:03
But simultaneously, I guess it is a little bit like some of the automation of other forms of medicine or therapy or things like that, where it's a tricky kind of ethical bind because you want to be able to provide this service to as many people as possible. Which means that getting a lot of people something that is not perfect but that is just okay is still better than what they had before
And yes, the point that you make in the end about trade off between making things accessible to a lot more people who don't have access to certain It's skills, certain expertise, certain resources, but then keeping the option of that one on one consultation with the expert, if you will, or one on one support by the expert, is also equally important. And it is a tough trade off. I was watching the listening to a radio program about a doctor who is taking bringing specialized medicines
Romie herself
@Romiesays · 4:59
It's interesting because this kind of came up in a couple of different ways for me this week, kind of on both sides of it, in terms of both how the technology has really expanded access and also of, again, running up against kind of a limit of like, I don't actually know how to automate this part
Swell Team
@Swell · 0:15
Romie herself
@Romiesays · 3:29
So right now I'm going to do a much better job than a computer. But if we could get that audio processing into a computer that would eventually, in theory, be able to do a better job than me, where it could put up captions for both people that are talking simultaneously, which I can't do. So it's interesting to think about in the context of which problems we choose to solve in the same way that 50 years ago
And that's where I think the need for the human then captioning is always going to be there to be able to provide the right training, which then the AI can scale up and have a lot more content much faster. That actually brings me to a question. This is slightly tangential, but I was curious about it. Like, when I watch live TV shows, sports events, for example, the live captioning that I see often is inaccurate
Romie herself
@Romiesays · 4:45
But in terms of rebroadcasts of live episodes, I think that pretty much companies at this point are in the clear and can just rerun their existing captions rather than having the piece recaptioned. But there are some companies that definitely do it. I mean, I can think of a few different sports networks that do it, and those tend to be ones that are, like 24 hours sports type networks where they don't know where they basically just schedule captioners for ali. Of the day
Romie herself
@Romiesays · 1:25
Although Caption companies will also frequently have access to the teleprompter in order to pull names that will need to spell correctly and stuff like that. And so we know what stories are coming up and that can help us get much closer to verbatim. But that kind of error that I'm talking about would normally be a clue that there is not a live Captioner and they're not using software, they're feeding a teleprompter