The Leading in a Crisis Podcast

EP 31 Incorporating Artificial Intelligence into a crisis response, with Dan Smiley

Tom Episode 31

Send us a text

Prepare to navigate the stormy seas of crisis management as we welcome the unparalleled expertise of Dan Smiley, whose storied career has taken him from the decks of the Coast Guard to the forefront of oil spill response. Now, as a consultant and podcast voice on crisis leadership, Dan steps aboard our show to chart the integration of AI into the realm of incident command. With anecdotes from the field and an insider's perspective, he brings to life the transformative impact of AI tools like ChatGPT in sculpting dynamic response strategies and propelling software innovation in crisis scenarios.

The conversation then shifts to a more turbulent current – the murky waters of AI-generated graphics and the scourge of misinformation. Battling through the challenges of crafting our podcast logo with AI, we expose the ironic complexities of technology that's simultaneously advancing and leading astray. Witness how deceptively realistic visuals, such as those of a fabricated ship fire, can undermine trust during real-life emergencies. Dan and I explore the indispensable role of Public Information Officers in this new age, armed with the latest rhetorical armor to combat an onslaught of AI-spun fallacies that threaten the integrity of truth in our digitally dominated world.

Reach Tom Mueller at tom@leadinginacrisis.com
Reach Marc Mullen at marcmullenccc@gmail.com
Reach Dan Smiley via LinkedIn.

We'd love to hear from you. Email the show at Tom@leadinginacrisis.com.

Tom Mueller:

Hi everyone and welcome to the Leading in a Crisis podcast. On this podcast, we talk all things crisis management and we deliver that through storytelling, lessons learned and experiences as shared from experienced crisis leaders. I'm Tom Mueller. With me again is my co-host, mark Mullen. Mark, welcome back. Hello'm Tom Mueller. With me again is my co-host.

Marc Mullen:

Mark Mullen. Mark, welcome back, Hello Tom.

Tom Mueller:

All right. Our guest today is joining us from Washington State. He is Dan Smiley and I'm going to hand it over to Mark in just a moment to introduce Dan. But he does bring deep crisis response experience, particularly in the maritime environment, and is the host of his own podcast called the Tactics Meeting, which also focuses on crisis management and crisis leadership issues.

Marc Mullen:

Sure Tom, and again it's my privilege to introduce Dan Smiley to join us on the podcast today. By my profession, the one thing I look for the most in a response is an incident commander who understands the dance and understands how to work with the chicken, with the PIO, and in all the exercises I've been at with Dan I've been impressed that he gets it. He totally understands how we have to work together for the good of the response. So I'm looking forward to conversation with Dan about incident command and the perspective from that structure into the JIC, and also his comments or understanding of the PIO role itself. So, dan, happy to have you with us

Dan Smiley:

Well thank you Marc. Thank you, Tom. I'm excited to be here. Thank you for inviting me.

Tom Mueller:

Dan, take just a minute and give our audience a quick summary of your bio. I know you've been working for a few years with a real focus in kind of maritime trades and that, so give us just a quick thumbnail of your experience and what you're doing today.

Dan Smiley:

Yeah, thanks, tom. I got here in a roundabout way. My experience was maritime. I joined the way. My experience was maritime. I joined the Coast Guard back in 1983. I was a bosun's mate. I sailed on the Coast Guard cutter midget and when that tour of duty was over I sat for my deck officer's license and I sailed passenger vessels and then freight vessels in Alaska until 1997, when, in an attempt to not be gone 260 days a year and to stay married, I took a job as the captain of a little coastal oil spill response vessel owned by Clean Sound Cooperative called the Western Gull. It was tied up in Bellingham and it was about a 15 minute walk from my house. So that worked out really well. So that kept me home. The staying married part didn't work out quite so well, but I stayed home.

Dan Smiley:

And now I'm in the oil spill response business and I don't know anything about it at all and I learned to skim oil and put out boom and do all those kinds of things. And then later I became the supervisor in Anacortes for Clean Sound and then a supervisor for Marine Spill Response Corporation, started going to tabletop exercises. I remember my first tabletop. All I did is sit with a yellow legal pad and keep adding up boom totals. I had no idea what I was doing. All I could think of was there's got to be a better way to account for these resources. This is crazy, and from there I ended up as a response manager for MSRC.

Dan Smiley:

Like you, I was at Deepwater Horizon. I was in the on-water recovery group in Houma for 105 days. After that I went to work for NJ Resources doing consulting, and today I am the general manager of response and preparedness for the Washington State Maritime Cooperative. I'm also an associate on the response roster for Gallagher Marine Systems. I'm a podcast host of the Tactics Meeting, as you kindly mentioned, and have my own consulting business, terminus Inc. So that's how I got here. I had no idea that I would spend all my time designing drills, going to command posts. I'm a ship driver by trade, but we all get here in a weird path.

Tom Mueller:

Right, and all that history is good experience for putting you in a knowledgeable position within a unified command somewhere right, because that expertise that you develop over time it can come into play in many different ways in a response. So the terrific background that you have for that. I wanted to talk a little tech with you if we could, and you know thinking about AI. You know it's the hot topic today. You know it won't be too long and it's going to decimate the planet and wipe us all out, but until then, I wonder yeah, I saw that movie.

Tom Mueller:

Until then, I wonder, you know, if you you know, as you peek into your crystal ball looking into the future, what role you see for AI in the incident command process, and is there something there that responders today should be starting to think about in terms of how we make use of that emerging technology?

Dan Smiley:

Yeah, that's a great question. I've been using ChatGPT myself for about eight months now. I pay for the $20 a month pro subscription so that I have access to ChatGPT 4 instead of 3. And I use it in one way or another nearly every day. As some of you may know, I do some software development and it's a really great tool for writing code that you can then just tweak a little bit and stick into your application. So I use it for that.

Dan Smiley:

It is a great tool for, I think, writing plans, and I haven't used it. I'm not PIO, but I haven't used it for writing press releases, but if I was a PIO I probably would.

Dan Smiley:

There was a response that I was on last year where I needed to write a pipe cut plan that had four components on it a cut team, a lift team operating the crane pollution mitigation team and, because of the risk of fire, a fire watch. And we'd had a planning, meeting that with all the parties involved, contractors and whatnot and I had a lot of notes and I decided you know what I'm going to see, what AI can do with this. I uploaded all the notes to AI and then I gave it some instructions Say, write me a plan. It's going to be in four components here's a copy of the 201 so you can get some of the response information from it and gave it a few more instructions and said go and darned if it didn't get me 80% of the way there in its first try.

Dan Smiley:

I think, if I had tried to write the entire thing from scratch, it probably would have taken me five or six hours. With the help of AI, I did the whole thing in an hour and 40 minutes.

Tom Mueller:

That's amazing. And then I assume you kind of went in and did some final editing and tweaking to make it fully fit for purpose.

Dan Smiley:

Yeah, what I did is I took what it wrote and I dropped it into a Word document and then I edited the first section of it, copied that, dropped it back into AI and said review this, edit it for clarity, remove any redundancies and make any suggestions based on industry best practices for this kind of work. Interestingly, it came up with some ideas to include that I hadn't considered, and so I edited the document, put those things back in and it went back and forth in this way, you know, over the course of a half hour or so, until I got it to a position that I thought met the objectives of the unified command. Then I slapped a cover page on it and we were off to the races.

Tom Mueller:

And so it sounds like there might be applications for writing an environmental plan or a beach cleanup plan, some more kind of across the incident command toolkit.

Dan Smiley:

You know, it's got access to a lot of knowledge and if you give it good instructions, I think one of the keys to being able to use these large language models is to give it really good prompts. You know, one sentence probably isn't going to get there right. You might tell it. Hey, you're the environmental unit leader in a unified command for an oil spill that's being managed under the Oil Pollution Act of 1990. I mean, give it a lot of information about. You know what is it it's trying to do. The more you can give it, the better result it's going to be, and I think that's going to be one of the things that people are really going to be trying to learn to do. There's probably going to be college level classes in this is how to write a prompt, a question for AI that returns a useful result.

Marc Mullen:

So you're saying that in the long run - Tom back to your question - in the long run AI would be a great tool to use for your planning documents of any type, which becomes signed-up documents, which become the base of response activities, which becomes what you see or what you do out on the water. So you're saying that Unified Command would, trust, depend on sign off on an AI-generated document?

Dan Smiley:

I don't know that I would present it to Unified Command as an AI-generated document. I would use it myself as a tool. I like to think of it as a force multiplier. I'll use it for a tool to produce the document that I am then going to turn into the unified command, but it's my document.

Dan Smiley:

You know, it's like you know it's like using spell check in Microsoft Word, right? If so, is it not my document If I use spell check in Microsoft Word, right? So is it not my document. If I use spell check, you know, is it not my document if I start with a report template that has been developed in Word?

Marc Mullen:

Well, it is my document, right, so I see it as the same thing, right? Well, my only point was you use the tools you have, and you use the tools to increase your capacity as a force multiplier. All I'm asking is it would be acceptable?

Dan Smiley:

there's there's no reason for you to think that it would not be an accepted plan because you used AI. N o, I wouldn't think so, even if all I did is drop the complete, a completed plan in and say edit this, review this for spelling and edit it for clarity, rather than. You know, I could do the same thing by dropping it into Grammarly or something and say review this for spelling and edit it for clarity, rather than I could do the same thing by dropping it into Grammarly or something like that. But the AI is going to make some very specific recommendations and I find it, even if you don't use its recommendations.

Dan Smiley:

It's nice to be prompted to consider them, and you might find that it prompts you to include a section in your plan that you wouldn't have otherwise. Right, I think it's useful.

Tom Mueller:

Dan, I used AI to develop the graphic for our podcast here and it was. You know it was way more frustrating a process than I expected it to be because it took me about 50 different attempts to sort of get you know what we see here now with our podcast logo, and I was just surprised. I thought it would be smarter, easier, faster, but it just, you know, it built a little bit more every time I added or clarified prompts. But that's been almost a year ago now. I think if I went back to that same model, it would probably react very differently to the same prompts that I put in, just because the knowledge base and training of those language models is growing so significantly. You think that'd be true?

Dan Smiley:

I know it's true. I've used it to generate a lot of graphics that I use in PowerPoint presentations, and it's improved dramatically just over the last three months. One of the things that I find quite interesting is that it almost never spells things correctly in the graphics.

Dan Smiley:

I don't know if it's trying to avoid copyright issues or not, but I'll tell it. Create an image of a middle-aged white female wearing a white incident commander vest that is labeled, quote, unquote, incident commander, and then it all ended up with incident commander. That's got like three eyes in it, like I and I, I, I, whatever. I mean it. Just it becomes really weird and if you're persistent with it it's like no, no, I need you to spend, I need you to make the graphic look like this. You might, you might get it right. In some cases I've literally had to take the graphic, drop it into the slide and then take a text box, put the label I want in it and stick it over the graphic so that it's spelled correctly.

Dan Smiley:

But it creates some great graphics. I made one for an inject for the wildlife group so that it was supposed to be a picture coming in from the public showing a duck swimming in a rainbow sheen water and I mean it was like great graphic, I use it. It freaked him out, but you know the news media are using these graphics. One of the articles about the Genius Star 11 ship fire showed this fire sitting at a dock at a container terminal, which was not the case with this. Ai kind of generated fire on board, and I guess they thought that was good enough. We have a ship. It's on fire. It was a fake picture and nothing about the situation that they were showing was remotely true to the incident. Right, there had never been flames outside of the hull of the ship, there'd just been some smoke.

Tom Mueller:

And that was a news organization?

Dan Smiley:

Yeah, it was in a news feed, so you start to see these things pop up all over the place.

Tom Mueller:

Yeah, you also got the potential for mischief around an actual response right, when you might have fake images of ducks mired in oil or something like that that could be generated and shared online as real photographs. Are we headed toward that kind of mischief now?

Dan Smiley:

I'm sure we are. It's probably happened already. If we've been paying a lot of attention to the social media feeds around the Baltimore Bridge collapse, you may have seen that kind of thing. I'm not saying that I haven't seen it, but I'd be shocked if that kind of thing isn't out there already. I haven't seen it, but I'd be shocked if that kind of thing isn't out there already. So that's going to be a really challenge for you as PIOs, as JIC managers, as liaison officers, and not only now do you have to tell the story that's real. You have to go out and refute the story that's not, and that that's going to be a challenge.

Tom Mueller:

Yeah, I mean there's a whole other level of response here. You know, I guess it falls back to rumor control or you know that piece of the PIO team to try and keep track and refute those things it's. You talk about a force multiplier for misinformation. The potential there for mischief is just huge. Not looking forward to that challenge.

Dan Smiley:

No, or you know they, they take a, they take a video, you know. Create a deep fake video that is a version of the official press conference and have the unified command saying something completely different. The unified command saying something completely different. And although I don't have any AI tools that'll do that kind of thing, they're certainly out there. We've seen some of that pop up on Instagram and and other places. It's in just a fun way. But you know, I'm not looking forward to the day we you know we see presidential candidates saying crazy things because we're watching deep fake videos I'm not sure we'll be able to tell the difference I if they're.

Dan Smiley:

If somebody's got the right hardware and software, we may not be able to tell the difference.

Marc Mullen:

That's going to be the problem now, this is what happens when mischief intersects with technology. Because, tom, you remember, you remember the Thunder Horse platform and in that response they were having to deal with images of some platform out Brazil that had actually sunk. So even back then they were dealing with the reality that you can find imagery that counters the story and try to change public perception. So you're right, now we can do it much more seamlessly and much more rapidly.

Tom Mueller:

We're going to have to get smart in the technology that allows us to quickly deconstruct those images and figure out which ones are AI generated and which ones may be real. There's going to be an opportunity for personal growth for your PIO teams going forward.

Dan Smiley:

Yeah, there's a whole new section of the JIC right.

Marc Mullen:

Well, we'll take the section that used to be - oh the RP altered this photograph! and turn it into oh the citizenry altered this photograph. That's right, you're hitting kind of close to home there, mark, I know. Oh, the citizen re-altered this photograph.

Dan Smiley:

That's right.

Tom Mueller:

You're hitting kind of close to home there, Mark.

Marc Mullen:

I know. Sorry Tom. Probably need to back away and get the counseling going again.

Tom Mueller:

Yeah, All right, hey Dan, thanks so much for joining us on the podcast today. Really appreciate you taking out some time and it's been just a fascinating conversation, so I hope you'll find an opportunity to come back and join us again soon.

Dan Smiley:

I'd be thrilled to do so Thank you both.

Marc Mullen:

Thank you.

Tom Mueller:

And that's going to do it for this episode of the Leading in a Crisis podcast. Thanks for joining us. If you like what you're hearing, then please like and subscribe to the podcast and please tell your friends and colleagues about us as well, and we'll see you again soon on the next episode of the Leading in a Crisis podcast.

People on this episode