Like you run tests rapidly and deploy and there may be some overlap of functionality and even expense as you’re trying to figure out the best model. As an example in December, we launched an internal chat tool, so internally developed and secure. So any prompt output or uploaded document stays within our environment and can’t be used to train LLMs and doesn’t go anywhere. But internally it leverages Open AI and Anthropic only. We are looking at the potential to add other models, but those are the ones for our general use across the organization.
We’re most comfortable with at this, at this stage. That said, I mentioned we have a handful of markets that are testing.
So you can version a script in chat GPT today, right? But the output can potentially be fairly variable depending on your prompt, right? What you ask it to do. And so, we’re looking at from a responsible use perspective, what something that’s a little more controlled from a prompt and model perspective might look like? What we’re finding though is the more controls, the more templatized the output can be.
TVSPY: I worked for a company that was a data storage and protection company. I don’t wanna make your brain explode, but is Scripps focusing a little bit more on cybersecurity and beefing up the IT department now that you’re getting into the cloud and the crazy cyber world out there?
Hartman: So our cybersecurity officer is part of governance and in any demo conversation we have not only a lockstep partner in ensuring that our published guidance to the teams is reflective of the priority that the data security, data privacy, cybersecurity is to us.
TVSPY: Yeah. I’m kind of going off the rails here, but I do know that one of the biggest frustrations at my old company was that I wasn’t able to livestream or do remote video as much as I wanted with people.
I wasn’t able to use certain products because the cybersecurity risk to the company. We couldn’t use them because our IT standards were so high. There were certain security risks to certain live streaming programs, which, you know, broadcasting is, in a sense, live streaming.
What comes to mind for me is like, the more you guys get into having proprietary information in your networks, the more you’re gonna wanna protect it. The more you protect it, the harder it is to do what you’re there to do.
Hartman: Yes. Yes. Yes.
I think what we’re gonna find, you know, quite a bit, is that just because we can do something doesn’t mean we should or will. Or that it’s worth the risk. I am no cybersecurity expert, but very much appreciate those within the organization that are. We share our requirements with any vendor before we even have a conversation. If they can’t meet our standards, we can’t have a conversation.
TVSPY: If you had your unicorn and rainbow world with AI and local TV, I don’t mean that in derogatory way. I mean that actually like, “Wow, one day we’ll be able to do this,” what would it look like?
Hartman: I think a lot of the conversation, misguided parts of the conversation, are overly focused on streamlining and automation. And certainly that’s an opportunity. But I like to think about it from the perspective of what have we always wanted to do that we just couldn’t, right?