Ep. 8 Has Generative AI Already “Jumped the Shark?” | AI Insights and Innovation

[David Linthicum]

Has Generative AI Jumped the Shark?

Has generative AI already jumped the shark? Let’s talk about it. Welcome to AI Insights and Innovation, your go-to podcast for the latest news, trends, and insights in artificial intelligence, including generative AI.

Introduction and Host

I’m your host, David Linthicum, author, speaker, B. Liz Geek, longtime AI systems architect and analyst at The Cube Research. Let’s get into the topic.

Defining “Jump the Shark”

Hey, Siri, what is meant by the term to jump the shark?

[Siri]

The idiom jump the shark is used to describe a moment when something that was once great has reached a point where it will now decline in quality. This is from IMDb.

[David Linthicum]

Generative AI’s Peak

So obviously that’s a bit clickbaity, but it’s interesting. Actually jump the shark came from happy days, way back when Fonzie, when the idea started to run out of steam, or the show started to run out of the steam, they had a episode where he jumped a shark with water skis, and I thought that was kind of appropriate.

[00:01:13]

But it’s often referred to as things that have kind of reached a pinnacle. In other words, it’s not a lot further we can go with it, we’re kind of trying to figure out what it is and what we can use it for, for additional insights. And I think that’s where we are in many respects within generative AI, even though it’s only been around a couple of years. So let’s figure that out.

Generative AI’s Challenges

So in this episode, we’re going to get into whether generative AI, despite its impressive innovation, has already reached its peak. We investigate the challenges organizations face in identifying practical use cases and justifying the substantial costs associated with their deployment. And what we’re getting into, and by the way, this is not asserting that generative AI is going to go away.

[00:02:00]

What I’m asserting is that we’ve kind of reached a point where there’s other options that are cheaper, there’s other options that are more lightweight, there’s other options that can be more tactically applied, that may be better options for many businesses. And so as a surge of adoption has occurred, and nothing I’ve seen in my career, and I’m 62 years old, I’ve been in this business since I was 18, has really matched what we’ve seen in the rise of AI, the re-rise of AI around the generative AI stuff, obviously with the rise of chat GPT, people could see the value in it, people used it in their daily lives to do very handy things like write notes, write emails, become a workflow automation system, operate cars, operate devices, the list goes on. And so obviously AI has always had a lot of value that it’s able to bring to applications to the business, and generative AI is just able to do that a bit better because it’s able to generate new uses of the information.

[00:03:06]

Now generative AI systems are normally defined by LLMs, large language models, and those are huge models that are created with massive amounts of information, and that’s why they become handy. The reason we use chat GPT and other LLMs is because they’ve been trained with a lot of the information on the internet, and so therefore they’re able to map patterns within that information to provide us with responses that we can access through natural language processing, we can talk to it, we can ask it about the weather, we can ask it about a particular topic, we can ask it as I just did with the Siri application about a particular term, and it’s able to find a path through the information and bring us the appropriate data to us. What comes into play here is that generative AI has a tendency to be very heavy weight systems, as I mentioned earlier, and so in other words it takes a huge amount of data to train these systems, a huge amount of processing, and a huge amount of money that needs to be leveraged to make this processing occur.

[00:04:16]

So as businesses are looking at the applications of the use cases for generative AI, they’re often a little confused about what they can use it for for their particular business, and also often confused about the expense of building these systems.

Hype vs. Practicality

And so that’s kind of where we’re at right now. We understand the power of it, we understand the value of it, it’s getting better as time goes on, all the new releases of the LLMs out there. But businesses, CIOs, CEOs, boards of directors are looking around trying to figure out if this is actually something that’s going to provide business value for them. In fact, many articles, and I’ll link them down in the comments, sorry, below, depict the fact that CIOs are running into trouble in finding use cases for generative AI.

[00:05:05]

And so here we are. We have access to this very innovative technology that came through the evolution of AI. Generative AI is a subset or a derivative of machine learning, and machine learning is a derivative of AI, and of course we have deep learning and other systems there. So we’re at a place where we’re asking the appropriate questions in terms of what can we do with it. And I think businesses are trying to figure that out right now.

Generative AI’s Pinnacle

And that’s the core of the discussion here. So why are we asking this question now about generative AI kind of hitting its pinnacle? Because I think people are getting a bit sick of the generative AI hype. Every cloud computing show out there now is a generative AI show, and virtually every enterprise software player is going to have and promote their technology through the lens of generative AI.

[00:06:00]

Even when I went to RSA a couple of months ago, even though it’s a security conference, everybody was, in essence, talking about generative AI, protecting generative AI, leveraging generative AI within their tools and technologies. And it was just so hype-driven that all of the vendors feel they need to align their technology strategies behind it, and that kind of continues to this day.

CIO’s Perspective

But I think as the rank and file out there, CIOs, people who are responsible for automating core business systems, they’re seeing a lot of hype out there. They’re seeing a lot of interest in it, certainly even from their staff members, their employees. But they’re not necessarily seeing very obvious use cases for it. And they also see, when they look at the cost of the thing, the fact of the matter is that they just don’t have enough budget to deploy some of these huge honking generative AI systems based on large language models for their particular business case.

[00:07:04]

Now, they can use LLMs, and certainly there’s tactical uses for LLMs, such as supply chain integration, logistics management, inventory management, things like that. But as far as them building a generative AI system, that’s a bit of a tougher putt for the rank and file CIOs out there, CEOs, the people who are actually running the companies.

Generative AI Saturation

So the overload continues, and people are telling me all the time that they’re just saturated with their generative AI hype. But some people, I get on briefings, and the first thing they say, we don’t want to talk about generative AI. They just get too much of it. And so social media platforms and content websites are becoming inundated with AI-generated articles, posts, videos, things like that. The amount of AI-driven stuff out there, people performing avatars and deep fakes and even leveraging AI podcasts that are generated from their voice, or even AI YouTube videos, and there’s a ton of them out there, is just enormous.

[00:08:07]

And so people feel that this technology is getting overused and probably misused in many instances, and they’re just getting a bit saturated with it.

Inaccurate Information

So there’s also inaccurate or misleading information, AI-generated news articles or summaries that contain factual errors or lack of context. As I mentioned with generative AI systems, normally they go a mile wide and an inch deep. And the reason is, is because they only have information that they’ve been trained to provide. And if the information is not there, they’re not going to give you any additional information or insights into it. So there’s no innovative differentiator. There’s no kind of common thinking in terms of what a particular topic means and how it’s defined, but what it means to you and some deeper insights into what we’re considering, some innovation, innovative thinking around using the information. And so if we’re leveraging AI to generate all these articles and generate all these videos, we’re not getting that perspective.

[00:09:03]

It’s also high cost with minimum ROI, and we already mentioned that.

High Cost, Low ROI

Companies investing heavily in AI-driven customer service bots define that customers prefer speaking with human representatives due to the bot’s inability to handle complex queries or offer empathetic responses. As a result, they return minimal amount of investment. So we’re seeing lots of organizations that are investing in chatbots as a means of providing customer service. And sometimes that’s not bad. I talk to a lot of generative AI chatbots that can help me solve issues and technical problems and things like that. But there’s lots of very bad ones out there that may provide you with erroneous information. They may lead you down the wrong path. And they’re certainly not going to be humans. You’re not going to be empathetic around what you do. And if they try to make them empathetic, at least where the technology is now, they sound kind of condescending. And so people are kind of pushing back on that.

Job Displacement Fears

Also, you’ve got the job displacement fears.

[00:10:00]

Everybody’s afraid that generative AI is going to take over their jobs. And in some instances, that’s going to be the case. So designers writing music compositions, increasingly concerned about AI encroaching on job security. And of course, we’re talking about their ethical point of view, what it’s able to do and what it’s not able to do. And even though we have seen some displacement with automating information worker kind of things, it hasn’t been to the degree, I think, that everybody thought it would be. And here we are at the end, you know, approaching the end of 2024.

Over-Promised Capabilities

Over-promised capabilities, AI-powered personal assistants, the claim to revolutionize the productivity. But, you know, they ultimately fail to understand the nuances of the commands or integrate effectively with other tools. And it becomes kind of an integration nightmare. People are going to use these things across systems, at least now. And we have some promising technology like Apple Intelligence that’s coming forward, which is going to provide us with an AI agent that’s going to be on our phone.

[00:11:00]

So the jury’s still out on that. And I think we’re going to perfect this as time goes on. But right now, the technology is not at a place where people feel that it’s not without its downsides.

Privacy and Ethical Concerns

And of course, privacy and ethical concerns, generative AI systems can do really cool stuff like backward engineer anonymized information and actually add identities to it. That’s not so good. That violates HIPAA regulations, privacy issues. Corporate bandwagoning example, you know, businesses, everybody’s jumped on the generative AI bandwagon. They did so from the beginning. If they don’t have some sort of a generative AI strategy or AI strategy, they don’t feel that they’re players in their particular industries that they’re in. So everybody, as I mentioned, all cloud conferences are now generative AI conferences. You know, every technology system out there is going to have a strategy around how they’re going to leverage, how they’re going to protect, how generative AI kind of fits into their ecosystem.

[00:12:04]

Novelty Wearing Off

So the novelty is wearing off. Early adopters of AI-generated art or music initially found the technology fascinating, cool. But over time, the novelty wears off. Creation lack of emotion, depth, and creativity in human artistry becomes missing from a lot of those things. And so the number of AI written drawings out there, and some of them are pretty cool, is just overwhelming. And so if you’re creating art, there has to be some sort of an innovative differentiator that’s going to set your art aside.

Complexity and Integration

Complexity and integration, examples enterprises attempting to integrate generative AI tools and other into their operations also face, often face significant challenges related to data management system compatibility, workforce training, and it leads to frustration in stalled projects. I’m a generative AI architect. I’m an AI architect who used to be a cloud architect, and he used to be an IT architect, and I’ve done those jobs as well.

[00:13:03]

It’s not a simple thing to make these things work and play well together, and certainly with generative AI, again, you’ve got to find a valid use case for it, and sometimes that may not be available. So people who are looking to build a generative AI system, in many instances, I’ll come and meet with them.

Cost vs. Traditional Technology

I’ll look at them. You know, traditional technology will do a better job at building the same system, and it’s about one-fifth the cost. So there’s no business justification for using a generative AI stuff to build, and the reason it’s so expensive is because of the complexity, the cost of the talent. AI architects make a lot of money, and AI engineers make a lot of money. Data scientists make a lot of money, and so if you’re going to invest in this area, get the talent you need, invest in the technology, invest in the processing power that you need, you know, massive amounts of GPUs, for example, it’s going to be very expensive, and there’s a bit of sticker shock going on out there right now.

Customer Experience Decline

And also customer experience declines. You know, e-commerce websites using AI for personalized recommendation that continuously miss the mark, leading to subpar shopping experience and customer dissatisfaction.

[00:14:09]

We’ve had lots of issues. I’m not going to name the fast food chain, but they were using AI at their kiosks to take orders from customers. Customers got frustrated with it, and then they pulled it back, and so we’re seeing those sorts of fallbacks to more traditional ways of doing something from early adopters of generative AI.

Misapplication of Generative AI

And by the way, that does not mean that it’s not going to improve the ability to build these kiosk systems, these intelligence systems. What I’m asserting is that we’re moving too fast in many instances in leveraging generative AI and abuse them for use cases where they have no real chance of solving the issues. And so we’re trying to get smarter with the technology, but right now I just see the misapplication of generative AI more so than the correct application of generative AI for the business use cases out there.

[00:15:03]

Future of Generative AI

So what’s the future of this? Generative AI is not going away, and I’m not making the assertion that it is. I’m not saying we’re all sick of it, and therefore everybody’s going to get off the technology and it’s going to die a quick death. I think we’re normalizing the use of it and becoming a bit more familiar with the advantages that it’s able to bring and also some of the downsides of it as well, cost being the downside over use of technology and the ability not to use it in the proper context for the business. Also the fact that everybody in deploying generative AI systems are deploying heavyweight AI systems and not necessarily looking at lighter weight technologies like agentic AI.

Agentic AI as Alternative

We talked about that in the last podcast, the last video, which is more of a lightweight deployment of AI-based systems that can leverage generative AI and the ability to have intelligent agents that interact one to another, maybe a technology that’s going to have more use cases for the lightweight tactical use of AI, which I think is where a business is going to focus.

[00:16:10]

They’re not going to build LLMs. They’re going to build very small language models, they’re going to build tactical use cases for AI, doing supply chain integration, logistics processing, inventory control, things that really make them money where we’re not building these huge systems that take 10 days to train out of massive amounts of information, but they’re doing very tactical, very productive things for the business and utilizing AI when it’s going to provide value within the application and agentic AI may have more of a promise.

Generative AI Passe

Dave Valenti just wrote an article about this, talking about his generative AI passe and I agree. We’re kind of talking about the same thing here. The reason we’re talking about that is because there seems to be lighter weight, more cost effective ways of leveraging AI, agentic AI being the latest key trend, which is based on an old trend of leveraging AI agents.

[00:17:02]

Those went around for a long time. We worked on those 20 years ago, but it’s going to be cheaper and therefore deliver more value to the business.

Agentic AI’s Advantages

They’re a little bit more flexible in being distributed and the ability to be embedded with certain business processes, not necessarily try to take over major systems. I think businesses are going to be more attracted to that kind of a model versus the generative AI model, which is very heavyweight and very procedural based.

Pragmatic Look at Generative AI

We need to have a pragmatic look for this stuff, has to transform the potential in various domains, software development, where it can boost productivity, automating repetitive tasks. I think it has an application there, however, you have to look at the potential values of that, which depends on overcoming challenges and finding sustainability, valuable applications for this stuff. It’s one of these things where we go through it all the time as an industry.

[00:18:04]

Industry’s Shiny Object Syndrome

The technology industry does this a lot. In other words, we’ll chase a particular shiny object. Fifteen years ago, it was cloud computing, before that it was service-oriented architecture and integration and distributed systems and the PCs and LANs and client server and all these trends that kind of came along. We move very quickly after that particular trend, invest a lot of money, everybody feels left out if they’re not talking about and leveraging this technology, but eventually it normalizes around the real value that the technology is able to provide.

Generative AI’s Normalization

I think generative AI is doing that now. When I say jump the shark, we’re reaching kind of a pinnacle in the hype and I think we’re looking around to trying to find the real value of this stuff for the business. Business leaders are asking that question and they should be asking that question because in many cases, IT is asking for five, six times the budget for building these applications and building these generative AI systems when there’s no real business advantage in doing so.

[00:19:08]

That’s just going to be a recipe for going out of business. If you’re starting building these systems for just the reason of getting the generative AI checkbox on your quarterly reports. So you don’t want to be that person, you don’t want to be that company.

Generative AI’s Role

So while generative AI is innovative and it’s not going away and it’s going to be a form of AI pretty much forever, very much like machine learning, it’s a derivative of machine learning. Other AI-based technologies, the genetic AI and the ability to look at deep learning and all these sorts of disciplines that have spun off of AI. Generative AI is going to have a role. It’s going to have something that it’s able to do well that businesses and industry are going to be able to use. The ability to create LLMs around a particular vertical industry such as healthcare and the ability to do centralized diagnostics versus one hospital owning and maintaining the LLM, but it’s an LLM that may be leveraged by a hundred different hospitals and therefore they’re able to use that technology in an economical way and also allows them to scale.

[00:20:15]

So all these sorts of things are being considered now.

Industry’s Awareness

I’m in a lot of these meetings, I’m in a lot of briefings talking to technology providers, they’re seeing the same thing. And while it’s kind of being whispered around now, I think people don’t want to be, I guess, the person like myself who provides the message that this may be something that we have to kind of take a harder look at. Those kinds of things are being back-channeled a lot right now.

Normalization of Generative AI Hype

And I think people are seeing the normalization of the generative AI hype and utilization of different mechanisms, the ability finally to find applications for it and the ability to figure out how we’re going to make this financially viable for the smaller companies out there and even the Global 2000 companies in leveraging this technology effectively.

[00:21:03]

We have to figure out how to deal with governance, strategic integration, ongoing innovation. We have to deal with security. All these sorts of things are problems that still need to be solved.

Generative AI’s Future

So this is not necessarily trying to, this is not me telling you about the swan song that generative AI is playing. Technology never goes away. This is exciting technology. It’s going to have some value. I just think it’s being oversold right now.

Businesses Aligning in Other Directions

And right now, I think businesses are aligning themselves in other directions in some instances. In other words, they’re not all in with generative AI. They’re looking at other options, agentic AI, non-AI systems, for example, workflow-based systems, RPA, robotics process automation, things like that, to provide more tactical uses of AI.

Smart Approach to Generative AI

And I think that’s a smart thing to do. So keep that in mind.

Conclusion and Call to Action

Well, that’s all I have for you this week. Don’t forget to like and subscribe.

[00:22:00]

Promoting theCube Research

Also, you know, check out theCube, check out theCube Research, check out some of the work we’ve done. Check out the great videos that are put out by theCube. They spend a lot of time on the analysis of those things. I use them as resources for these videos and for my work as an analyst, my work as an architect. So good resources there.

Call for Comments and Questions

You have any comments, questions for me, hit me up in the comments below.

Closing Remarks

Else, I’ll see you next week. Cheers, guys.