Looking Ahead To AI In 2024

Does it feel like the pace of change is speeding up? That’s because it is, especially in the world of AI.

Google Trends is showing that even with all the hype that AI has gotten in the past decade or so, it has positively exploded in search interest in the past twelve months:

You know that if your parents are texting you on a weekend asking about ChatGPT, then we’ve definitely “crossed the chasm” and AI is not just interesting to technologists and researchers, but to pretty much everyone.

For those of you who thought you could predict where 2023 went, well, that would be amazing. But given that we know where 2023 has gone, we can use some of those trends to guess where we think 2024 might be going with regards to AI.

With that in mind, what does AI have in store for us in 2024?It’s crazy to think Generative AI really came into widespread use only in the past 12-15 months, and mostly in 2023. The pace of Generative AI change was intense in 2023 with new models being released on an almost weekly basis. So obviously, if Generative AI was big in 2023, it clearly will still be a star of the show in 2024. Generative AI is going to continue to be huge and impactful and make lots of news.

Generative AI Embedded in Everything

More importantly, Generative AI will get embedded into everything, from software to hardware devices, and applications of every kind. Simply because it’s so easy to embed generative AI into these applications. In the early part of 2023, if you wanted to use Generative AI, you had to actually use OpenAI’s Chat GPT online or in the API , or in other solutions such as Midjourney, Stable Diffusion, or Google Bard. But by the end of 2023, Generative AI was being embedded into the applications you were already using. This means you might be using any of a number of generative AI foundation models without having to leave your existing application experience.

As a result, given the ease of embedding generative AI in everything, we’re starting to see generative AI embedded in every product. Of course, It should not be surprising to find AI in places that you would expect, such as in your spreadsheet or in your word document because that’s where you can expect it to be.

But we’re even going to find generative AI in places where we don’t expect it to be and as a result, perhaps where it shouldn’t be. Maybe in the future there’ll be a toaster where you can have it draw a picture on toast with generative AI, or perhaps a generative AI alarm clock, where instead of just telling you what time, maybe it’ll tell you a story, or create some unique response for the day. Voice assistants will write a story for you about the news. There will be so much generative AI, we’ll come to expect it when it’s not there and not expect it when it is there.

The problem with generative AI is that the problems of generative AI haven’t gone away. The issues we have today around hallucinations, bias, copyright problems, and then challenges with the truth will still be there, but now just embedded in our word processor or toaster.

Raised Expectations for All AI Applications

One of the side effects of generative AI being everywhere is that people will come to expect more from other applications of AI that might be more difficult to implement. If generative AI is so easy, why are autonomous vehicles still so difficult? How about the other Seven Patterns of AI such as pattern and anomaly detection, predictive analytics, goal-driven systems, hyperpersonalization, recognition, and autonomous systems? Some of those patterns might be implemented using the very powerful foundation models that will get more powerful, but the other patterns will continue to be difficult to implement.

Regardless of how difficult they might be to implement, the casual AI user will come to expect them to be as easy as generative AI. In effect, generative AI has “raised the bar” for other AI solutions. They need to be as easily usable and accessible as generative AI tools. The casual user won’t want to be a data scientist or data engineer, as much as you want to label them as “citizen data scientists”. What we know as low-code today needs to be as no-code as generative AI really is.

So while AI is moving fast, it’s also still very sluggish in adoption in enterprise and government applications. Enterprise AI adoption, outside of generative AI will still remain sluggish because the enterprise-specific applications, outside of simply applying generative AI, will still require greater skills and all the challenges of implementing AI projects that exist. 80% of AI projects still are cited as failures.

So in 2024, people are going to become impatient with AI solutions. They’re going to say, why is it taking you months to build an AI solution when I can do something right now with my generative AI thing in a few minutes? This cognitive dissonance comes from people who are not AI experts, but who have played with generative AI and are going to come to expect more than might be possible from other AI solutions.

Vendors Add Just a Sprinkle of AI to Add More Hype

The AI marketplace continues to be frothy, even though venture capital funding continues to be challenging. Vendors who might have a not-so-sexy enterprise solution or consumer product will be highly motivated to sprinkle a little bit of generative AI or other simple-to-embed AI into their solution so that they can magically rebrand their products as hip AI products. This “old wine in new bottles” approach has been going on for many years, but it will be especially accelerated by generative AI, because as I said above, it’s going to be easily embeddable in everything.

Unfortunately enterprises and government agencies are still behind the ball when it comes to AI implementation. Since they’ll feel compelled to get with the picture in 2024, they will no doubt implement many of these new AI-based solutions that are really the old thing in new clothing. Enterprises will need to focus on the problem they’re trying to solve and making sure that AI is the right solution. Not just AI for AI’s sake. The problem is that enterprises don’t know how to separate the good solutions from the bad ones.

That means 2024 might be a year of great dissatisfaction with AI tools from folks who might have thought they were buying something new, when really they were buying the same old thing with a dash of AI. The combination of a still-poor VC environment combined with foundation models and open source solutions having ever-more capabilities and enterprises who have been burned by poor purchasing decisions means that there’s going to be a lot of crash and burn vendor failures in 2024. We should expect to see a lot of vendor consolidation, layoffs, and a reduced field of startups as incumbent vendors embed AI into their solutions, old stuff gets rebranded as new stuff, and those who don’t have AI acquire those who do.

AI Laws and Regulations Get Tougher in 2024

2023 was the year of the AI kind of laws coming to fruition. The European Union introduced the EU AI Act at the tail end of the year, the US Senate had a large round of discussions, and governments around the world look to tighten how AI systems will be used in the public and private domain. No doubt, 2024 is going to be the year that governments worldwide are going to need to act on AI.

Since generative AI will become embedded in everything, it’s going to be impossible to ban the use of generative AI. Even though governments are angling on how to restrict or narrow the use of generative AI, the reality is that generative AI will be difficult to regulate. However, the real-world impacts of generative AI are significant, so we can expect to see requirements for moderation, watermarking, and restrictions on the use and embedding of generative AI in specific situations. 2024 is an election year, so we can just imagine what sort of generative AI chaos there could potentially be. It should be very well expected that there’s going to be lots of use of generative AI in so different forms to drive fake news, generate text and images that aren’t real, or generate fake social media content. It’s so easy that I think we should expect it. And when that inevitably happens, it is going to just further the calls for increasing regulation.

The EU AI act and things coming out in the US Congress, will be trying to put some restrictions on the use of generative AI and requirements when they are used. As a result, what’s going to end up happening is that we most likely will start to see more heavily moderated generative AI systems with results that could potentially become increasingly worse and muted over the iterative generations.

We can already see that the generative AI solutions in some cases are getting worse as they get watered down to try to comply with existing and emerging regulations. People are already saying that ChatGPT is getting “dumber”. You might see generative AI systems start to refuse to respond to your prompts that they were happy to comply with in previous iterations, or the results are nowhere near the quality that they were before. Given that generative AI will become embedded in everything, this means that those embedded solutions will also get worse. Generative AI solutions, especially the big hosted ones, are going to face all these pressures that may force users to have to revise the use of those systems.

Trustworthy AI will be a huge theme in 2024.

AI Euphoria Gives Way to AI Realism

After the big AI sugar rush of 2023, it’s inevitable that we’ll face the sugar crash down to reality in 2024. The combination of AI vendor overhype, increasing regulation, and the watering down of generative AI solutions will lead to a more realistic perspective on AI solutions in 2024. The overwhelming positive, enthusiastic attitude to AI in the past few years will give way to more of a neutral and cautious attitude. That’s because people are seeing not only how the technology is actually being used, but also how it’s being misused and misapplied.

With AI becoming commonplace, and with data scientists still hard to find and hire, the notion that data scientist is the “sexiest job title of the 21st century” will start to wane. Already we’re seeing more positions opening up for prompt engineering and for more middle-level folks who can apply aspects of AI to their business, rather than needing specialized training or degrees. This means that the job market for AI will loosen as a wider range of job titles start to embed AI capabilities in their jobs.

In much the same way that specialist secretaries and word processors-as-a-job gave way to everyone having desktop productivity suites, so too will AI start to migrate from highly trained, highly salaried staff to more common job roles and lower salary levels. This means we can expect to see a decline in AI specialist hiring in 2024.

The role of project management will become more important now that it’s just becoming so much easier to add AI capabilities, especially generative AI capabilities, into your projects. The bottleneck is not the hiring of data scientists or the long process of tool procurement and implementation. So that means that the emphasis will be on knowing what you’re doing and doing it the right way. This means that finally, AI project management will get some respect as organizations realize that the technology of AI is the easy part – the people and process part of managing AI projects is much harder.

So are you ready for where AI is heading in 2024? AI and 2024 are both here, so it’s here whether you’re ready or not!