Are AI Agents Just Hype?

Iron Man was one of my favorite superheroes growing up, and I’ll admit that I’ve seen all three Iron Man movies.

Inside the metal suit is Tony Stark, a fictional industrialist and brilliant inventor played by Robert Downey Jr.

Stark invents what is essentially a large language model (LLM) that he calls J.A.R.V.I.S., an acronym for Just A Rather Very Intelligent System.

Later J.A.R.V.I.S. becomes a fully functional AI system that serves as Stark’s assistant.

In this form, J.A.R.V.I.S. is like a human butler on steroids, able to execute any task that Stark requests, no matter how complex.

The movie is a predictive of AI agents and where things are now headed.

Like I said before, I believe that over the next couple of years — and certainly over the next decade — AI agents will radically transform our world.

And I’m not alone.

Bill Gates has said: “In the next few years, [AI agents] will utterly change how we live our lives, online and off.”

OpenAI’s Sam Altman recently wrote: “We believe that, in 2025, we may see the first AI agents ‘join the workforce’ and materially change the output of companies.”

And Nvidia’s CEO Jensen Huang recently announced: “the age of AI Agentics is here,” calling it “a multi-trillion-dollar opportunity.”

But all three of these AI industry titans have a vested interest in AI agents becoming the next big thing in tech.

While plenty of regular folks still believe that AI agents are all hype:

Go onto any social media site and you’ll see plenty of posts like this one.

So how much of what you’re hearing about AI agents is real, and how much of it is hype?

Maybe we can get some answers from someone who is as close to the real-life version of Tony Stark as it gets, aka Elon Musk…

Grok This

Grok was a term invented in the 1961 science fiction novel Stranger in a Strange Land.

To grok something means you understand it intuitively. In other words, you just get it.

And that seems to be the ultimate goal of Musk’s AI system called Grok: to just get us.

Just like J.A.R.V.I.S.

Musk originally launched Grok in 2023 through his company xAI, which he started after his contentious breakup with OpenAI. An improved version, Grok-2, was released in August of last year.

According to Musk, Grok-2 is meant to understand and respond to people’s questions and instructions in smarter ways than other AI programs.

He says it can think more deeply about problems and access up-to-date information from the internet, including from X (formerly Twitter) which he also owns.

Musk’s team says they’ve added special features that help Grok 2 break down complex problems into smaller, more manageable pieces.

They’ve also given it the ability to look up additional information when it needs to answer questions instead of relying only on what it learned during training.

I’ve tested Grok-2, and it reasons through problems fairly well.

It also has access to current information, which many other AI systems lack.

Last October, xAI added image-understanding capabilities to Grok-2, so paid users on X can upload an image and ask questions about it.

That’s a pretty neat feature.

But while xAI and Musk refer to Grok-2 as an “agent,” it’s not quite J.A.R.V.I.S. yet.

Fundamentally, it’s another LLM like Chat-GPT4 that learns by studying huge amounts of text.

That makes it really good at matching patterns in language and generating its own text, similar to other AI programs.

Additional features like retrieval-augmented generation and real-time data access can make it seem more functional than other AIs, and it’s designed to have a more engaging personality when interacting with users.

But I don’t believe Grok 2 truly understands what people are asking, like xAI suggests. It makes mistakes or provides incorrect information – problems that plagues all AI systems.

But the main reason I wouldn’t call Grok a true AI agent yet is because it still doesn’t have the autonomy that AI agents of the future will have.

Which brings me full circle back to my initial question. Is the hype around AI agents real?

Here’s My Take

I’ve shown you how Jensen Huang’s concept of “Hyper Moore’s Law” suggests that AI computing performance has the potential to blow past Moore’s Law and double or even triple every year.

As AI continues to become more powerful, its reasoning functions are also drastically improving.

What’s more, one of Trump’s first acts in office was to repeal Biden’s executive order around AI safety, saying it: “hinders AI innovation.”

So it’s full steam ahead for AI for the foreseeable future.

That means AI agents should be a LOT closer to J.A.R.V.I.S. by the end of the year.

I also believe that AI agents will have autonomy in 2025, although I’m not suggesting they will function with full autonomy from the start.

Instead, I believe AI agents will perform parts of individual jobs or portions of a job process.

These agents will work in conjunction with traditional automations and other agents with some level of human oversight.

So you might have an AI agent that handles a portion of a customer inquiry, seamlessly handing it over to a human if a task gets too complex.

But they won’t take over entire jobs just yet.

However, that’s a real concern… and one I’ll cover in our next issue.

Until then, I’m curious what you think. Are AI agents real or all hype?

To share your thoughts, just send an email to dailydisruptor@banyanhill.com.

I look forward to hearing from you!

Regards,

Ian King
Chief Strategist, Banyan Hill Publishing