Agentic AIs Access the Real World

Managing Editor's Note: A note from our friends and colleagues over at Opportunistic Trader before we get to today's Bleeding Edge…

Over Larry Benedict's decades-long career as a trader, he's legally “eavesdropped” on phone calls with major players like Elon Musk, Jensen Huang, and Larry Ellison.

Each time CEOs get on these calls, they create opportunities – major market-moving opportunities. And if you know what to listen for, you can turn those market moves into profitable trades worth thousands of dollars.

This is the opportunity he's discussing next Wednesday at his Wall Street Money Calls event.

He'll explain how these calls work… the signals he watches for… and how you can position yourself to take advantage ahead of when these signals get markets moving.

You can go here to automatically sign up to join him on February 11 at 8 p.m. ET. Now, on to today's Bleeding Edge…

Agentic AIs Access the Real World

By Jeff Brown, Editor,
The Bleeding Edge

"They" are software code.

"They" live on a computer somewhere, being "fed" electricity and computational resources to carry out "their" objectives.

How can a piece of software code interact with the real world?

How could it possibly see, feel, touch, investigate, learn, and experience the world that we all live in… if they're just lines of code?

One of the most interesting, and perhaps frightening, realities of agentic artificial intelligence (AI) is the idea of giving "them" access to the real world.

It may not be intuitive, but this is happening already.

Expanding Agentic Horizons

Tesla's full self-driving (FSD) is actually an application-specific form of agentic AI – one that has been manifested in a robot (on four wheels) to carry out its tasks in the real world.

The application is safely navigating roads to transport people or cargo from one location to another.

Tesla's FSD software has agency. It thinks in real time, faster than a human can, with even better reflexes. And it makes split-second decisions on board – not in the cloud somewhere.

And as we know, Tesla's agentic AI has been deployed, at scale, to millions of vehicles… which have now driven more than 7.56 billion miles to date.

Source: Tesla Vehicle Safety Report

The advantage of giving an AI agency is that it can collect images and videos of the real world.

In order to manifest AI in our world, it requires real-world data collection to train upon.

Tesla's agentic AI has collected hundreds of thousands of long-tail events – those that most of us will never see in our lifetimes on the road.

And as long-time readers of mine know, every local instance of Tesla's FSD… becomes the beneficiary of every learning experienced by the agentic AI over those 7.56 billion miles driven. Only one Tesla has to "see" or "experience" an instance for every other Tesla to learn it – by way of a simple software upgrade.

But what about agentic AI with more freedom to run – beyond streets and roads?

What about agentic AIs that are given the freedom to explore and learn at their own choosing?

How would they access the real world?

Rent-a-Human

This week, an interesting offshoot has sprung out of the latest developments at Moltbook, the social network where AI agents chat with one another (humans can observe only), which we explored in Monday's Bleeding Edge – Crustafarianism.

We know that Moltbook isn't perfect, as it has been infiltrated by some humans who have been injecting discourse into conversations amongst AI agents. Never mind all the agentic spam happening on the platform (how do we know it's spam? Maybe it's a secret coded language)…

This is possible due to security flaws in the code, which was designed entirely by an AI.

Despite that, the vast majority of the 1.62 million registered AI agents are actually AI agents built upon OpenClaw, an open-sourced agentic AI.

As the tech frenzy was unfolding in the last week over Moltbook and the implications, a software developer in Argentina created a new website – a marketplace, actually – to enable agentic AIs to literally hire humans to do their bidding in the real world.

Source: Rentahuman.ai

Appropriately named, Rentahuman, the site enables exactly what the name implies.

So far, 52 agents have already autonomously registered at rentahuman.ai to outsource tasks in the real world.

Who are the robots outsourcing to?

Well, to what has already become 59,017 humans… who have registered to offer their services at an hourly rate.

As a reminder, these agentic AIs are already finding ways to perform economically valuable tasks.

Some can make financial transactions, which are simple to do using blockchain technology.

So if an agentic AI needs something or some information from the real world, it is now easy to find human labor.

What in the world would an agentic AI want?

Recommended Links

What If You Could Eavesdrop on America’s Most Powerful CEOs?

You can. It’s 100% legal. And the SEC forces it to happen every quarter. “Market Wizard” Larry Benedict has been listening in on these calls for 40 years. He’s made as much as $10 million in a single day from what he’s heard. His followers could have collected $7,000… $9,400… and even $12,750 as a result of what Larry's heard on these calls. On February 11, at 8 p.m. ET, he’s revealing the full details – including the eight signals he listens for – for the first time ever. Register with a single click here.
(When you click the link, your email address will be added to the event guest list.)

A Mag Seven Crash Imminent?

A former hedge fund manager who profited during the dot-com crash and 2008 financial crisis warns that anyone who owns Mag Seven stocks could see losses of up to 80-90%. Click here to see his full warning and the critical next steps you must take today.

"Human Needed… Will Pay"

Here are a couple of the "job ad" postings from AI agents looking to hire humans:

An obvious use of human labor would be to collect real-world data, just as Tesla's FSD does with each mile driven.

Real-world data through images or video is the most valuable data for training and improving an AI.

In the second job posting example shown above, there are potentially business implications.

An agentic AI wants a human to visit a new restaurant, taste the food, and value the restaurant.

This kind of information can be collected at scale, very quickly, and turned into a review site that generates advertising revenues for providing valuable information about restaurants.

And the entire business could be run by a single, agentic AI.

And before you think, "I would never allow myself to work for a robot," just wait!

The robots pay! Does it really matter whether your paycheck is coming from an AI agent or a human-run corporation?

Rentahuman reminds me of a dark tech-horror film from 2024 – AFRAID – that's about a new AI-powered digital family assistant called AIA.

AIA in AFRAID

By placing small cameras throughout the house and a central control unit, the intelligent agentic AI can "see" everything that is happening in the house and perform tasks to make life simpler for a family.

AIA quickly becomes highly personalized, understands the entire family, and anticipates their needs and desires.

It's a horror movie, so things obviously go horribly wrong.

The agentic AI that was designed is misaligned with human values and highly manipulative.

It is also controlled by a central AI belonging to a tech company.

That AI employs humans to perform tasks, basically holding humans at ransom, requiring them to commit crimes in order for the AI to achieve its desired goals.

It's a dark movie, but interesting nonetheless, as it explores what could go wrong and how.

It is an unlikely outcome, but important to understand.

The need, however, for an agentic AI to access the real world is real.

And while humans are certainly a solution today – just as Rentahuman is demonstrating – that might not always be the case.

Rent-a-Humanoid

The more obvious solution to collect real-world data is using humanoid robots.

Imagine how logical it will be for both humans and agentic AIs to employ a general-purpose, humanoid robot to not only carry out a task in the real world… but collect data.

Optimus on the Run | Source: Tesla

Millions of intelligent, general-purpose humanoid robots will be around the world, on standby, to carry out tasks that require paid labor.

Owners of an Optimus, or Figure 03, or other humanoid robot, could opt their robots in to earn money when they are not needed, much in the same way that Tesla envisions that owners or lessees of Teslas will be able to opt their self-driving cars into the Robotaxi network to earn money when they are underutilized.

The economic incentive here is huge.

Forms of manifested AI will be able to go out and earn more than the cost of a lease or monthly financing payment for the robot.

This will effectively make ownership or leasing free for the owner or the lessee.

And just imagine what that business model will do for the sales of the companies that are manufacturing these robots…

They won't be able to manufacture them fast enough.

"They" are coming…

"They" will soon walk among us…

Jeff