Ep. 15: The Answer to AI Problems Is...

This solo episode of AI Hope starts with a question I’ve been chewing on for months. AI brings a long list of real problems, from job losses and data center sprawl to cybersecurity, rogue agents, and what I call the dumbification of society. Hollywood and the press amplify the existential fears because doom sells tickets and gets clicks. Meanwhile, industry hasn’t offered a real answer, and few of us expect the federal government to act on our behalf.

My answer is humans. I can’t control what Washington does, but I can control how AI gets adopted in my own organization, and so can you. That starts with use cases that serve your mission instead of reckless, use-it-for-everything deployment.

It also means knowing the difference between efficacy, which is achieving your mission, and efficiency. When I hear executives tout AI efficiencies, I increasingly hear code for layoffs, and unless retraining and redeploying people is part of the plan, you should see yellow and red flags.

I also get into the AI lobby, why elected officials only move when they’re in hot water, and why every agent needs built-in human triggers instead of rogue internet access. Even though I work in this business, I think the frontier model bubble needs deflating. These models are a department store: You get a little of everything, and none of it is great. I expect niche models and SaaS platforms to gain ground once the big IPOs pass.

AI’s biggest problem is that we’ve let it go unchecked, deploying it with a shiny-object attitude that it would cure the world. The fix is us, whether that’s a human check inside a workflow or a bigger strategic call about how your organization uses the technology.

Listen today on Spotify, iTunes, YouTube, or wherever you get your podcasts!

0:42 Is there one answer to all of AI’s problems?
 2:23 The answer is…
 6:37 The department store problem and the rise of niche models

What is AI Hope?

What's your AI success story? What made you realize you had to change? And what does a hopeful future with AI look like to you? AI Hope with Geoff Livingston asks three simple questions of people doing the hard work of adapting and thriving in the AI era.

The Answer to AI Problems Is...
===

[00:00:00] Happy Sunday. Welcome to AI Hope. It was a big race day today, so I'm wearing my running gear. Please excuse me, but, ~uh,~ it was kind of fun here in Washington. We had a pouring rainstorm, and a bunch of us ran the Army 10-Miler. It was crazy. But, ~uh,~ a lot of fun things happened. We're here to talk about AI, though, and whether or not we can have hope with things and how to look at those particular issues.

And I wanted to ask you what you thought the real answer was to all these AI challenges, 'cause I think there's always one singular answer to problems with AI, whether it's how it's being deployed and, ~uh,~ whether or not that's costing people jobs, the efficiency issues, if you would. The existential issues of whether or not AI is going to kill us all, which is, again, to me, fear-mongering, but I know it's a big issue for a [00:01:00] lot of folks.

And, ~uh,~ of course, Hollywood and reporters love that story. Hollywood loves it because it sells tickets, and reporters like it because it helps them get clicks, but also because I think that they have their own fears with AI, which are, ~uh,~ kind of self-evident given that AI is obviously something that can replace reporters with writing.

I wouldn't say well, but in theory, it could, and, ~uh,~ that's obviously alarming. And then there are the challenges with the data centers and then all the other things that are occurring. So AI brings many challenges, cybersecurity, rogue agents, all the things, the dumbification of society. All these things make it very difficult to adopt AI 'cause we know that there are real problems with it.

What are the answers? And there doesn't seem to be a real answer coming from industry, and I think we all have, ~uh,~ very little faith that government's gonna act on behalf of the citizenship, although we are [00:02:00] fighting on many levels, and most notably on the local level with the data centers. So the answer in my mind is humans.

We are the answer. We have an opportunity to interject ourselves into the debate and resolve a lot of these issues. And how do we do that? ~Well, ~first of all, where we are matters. So I can't necessarily control how the federal government's gonna respond to AI or adopt it.

Can address how AI is being adopted within my individual organization, how I'm using it. I can use it responsibly and ethically. So, for example, if you're concerned about data centers and their environmental impact, ~um,~ first of all, you could try to find solutions that are environmental. There are some AI solutions that are out there.

They're not necessarily competitive with the frontier models or some of the other more niche-oriented technologies. But another way to look at it is, if I know that [00:03:00] this is creating a big drain, how am I using this tool? Am I using it recklessly however I want to, or am I creating opportunities that are use case-based that are gonna serve my mission and achieve efficacy for my organization or my company and deliver great results?

And by the way, efficacy does not mean efficiency, okay? Efficacy means achieving mission. ~You know,~ we are effective, as opposed to efficiency is cutting costs and the like in operations. Now, that, that doesn't mean that AI shouldn't be used to save time, et cetera, et cetera. But I, I am increasingly sensitive to when I hear executives talk about using AI to create efficiencies, because for me, that's code often for layoffs and for cutting people's jobs.

And so I think we really need to be mindful about this and talk about, okay, so we're gonna create efficiencies. How are we gonna use this [00:04:00] to, ~uh,~ make the folks that are involved more employable or retrain them for other work, or how are we gonna use their time? 'Cause if that's not part of the conversation and you're just hearing efficiency, efficiency, efficiency, I, I, I just think you should be seeing yellow and red flags.

~Um, ~all right. Back to the topic there with efficacy. Use cases that are on point to achieve efficacy are gonna help further your cause, your team's cause, your enterprise's cause, and that's really what we wanna see, right? And so I think that's mindful use. That's responsible use of AI. ~Uh, ~if we're seeing organizations embrace AI in a manner that's not, shall we say, mindful, and I think we see that with some of the frontier companies, then we should be registering our outrage.

I don't think elected officials act unless they're under hot water. I, I just don't see them acting in a manner that's best in the interest of the consumer or the [00:05:00] business person unless they're, they're absolutely boiling. And we know there's an AI lobby, and they're throwing a lot of money out there, and we can see that even with certain White House, ~uh,~ moments.

But, ~you know,~ don't kid yourself, your elected representatives are getting the same money too and the same treatment. ~Uh, ~so with that in mind, how do we really-- before we get humans involved, we should be the ones leading the conversation. We should be checking AI. We shouldn't be offering agents just rogue access to the internet.

We need human triggers all the way through, and we need to insist that technology companies build those triggers into the AI. So to me, that's the real big hope. I really think that the answer isn't necessarily that AI is bad, it's that we let AI go unchecked, and that's the real responsible, and I would say pragmatic view of what's happening.

We've let AI companies go unchecked. We've deployed this [00:06:00] recklessly and with a shiny object syndrome attitude that it's gonna cure the world, and now we're kind of reaping some of the results of that. The other thing too is, ~uh,~ I, I really I know I'm in the business and I know I'm here kind of promoting AI, but I, I still think that we need to do something here to kind of deflate this bubble that's happening with these frontier models.

I don't see them as the most efficient way to execute AI. I think it's really kind of a department store model where I go to Macy's and I get a little bit of everything and none of it's great. And I would really like to see more and more, ~uh,~ niche and smaller models, and I do anticipate that happening.

It may not be until we get through these IPOs with these big companies, but sooner or later, we're gonna have to figure out what it means to be an AI for, ~um,~ operations or an AI for the technology department and managing data and data governance. And increasingly, I think those software as a service [00:07:00] platforms are going to start asserting their own power and control of the matter.

So that's it. ~Uh, ~those are my thoughts on, ~uh,~ AI hope and AI existentialism. What is the answer for AI as a problem? It's us. It's always us. We need to assert ourselves and be the humans that are present in AI, whether it's through a check or through a larger strategic decision. I look forward to your thoughts on this.

Thank you.