How to Navigate the Future of AI

🏝 TheTechOasis 🏝

part of the:

In AI, learning is winning. While Thursday’s newsletter discusses the present, the Leaders segment on Sundays/Mondays will inform you of future trends in AI and provide actionable insights to set you apart from the rest.

10-minute weekly reads.

😟 Navigating A Future AI-Driven World 😟

It writes. It talks. It’s even flirtatious on command. Ever-supporting and ever-helpful. Worst of all, it’s not human, so it doesn’t sleep, and it doesn’t need to stop for a break.

Always willing, ever-obedient, always present.

All things considered, it might come after your job or company. There’s no way to sugarcoat this; with every new technological revolution, some survive, some perish.

That said, the AI revolution could have the most dramatic effect on jobs ever, potentially eliminating the need for them. And before you laugh and dismiss this, Elon Musk and Sam Altman defend this precise idea.

Decades from now, we might even compare it to the discovery of fire or agriculture, step-function improvements that elevated humanity to new heights, as King Charles III recently suggested.

But here’s the thing: You must not be willing to just wait and see if that’s true; you must take action.

For those reasons, last week we covered the ‘Am I Getting Steamrolled by AI?’ Framework, an easy-to-understand, practical guide to evaluating your current strength against AI’s winds of change.

But that framework was never a static exercise; it must evolve alongside AI. Thus, this week, we are covering what research predicts is the natural evolution of frontier AI, and what considerations and actions you can take today to future-proof yourself… and your kids’.

💎 Big Announcement 💎

This month, I am launching TheWhiteBox, a community for high-quality, highly-curated AI content without the typical bullshit, hype, or ads, across research, models, markets, future trends, and AI products, in digestible and straight-to-the-point language.

But why would you want that?

  • Cut the Clutter: Value-oriented, focusing on high-impact news and clear intuitions to extract from them (why should you care).

  • Connect with Peers: Engage in valuable discussions with like-minded AI enthusiasts from top global companies.

  • Exclusive Insights: As the community grows, gain access to special content like expert interviews and guest posts with unique perspectives.

With TheWhiteBox, we guarantee you won’t need anything else.

No credit card information required to join the waitlist. Premium members have immediate access, but are anyways very welcomed to fill in the waitlist form.

The Inevitability of Change

What used to take years will now take months; what takes months will turn into weeks, and weeks into days.

On Thursday, we saw how ChatGPT-4o, OpenAI’s new flagship model, coded an entire game in less than 60 seconds with no bugs. A programmer’s week-long project is now a mere 60 seconds away from people who can’t even code.

AI isn’t only changing things; it’s doing so fast. For those reasons, we are covering what cutting-edge research says about AI's short- and long-term future to give you an overarching idea of ‘what to expect in order to act.’

Then, I will try to inspire you to take action and suggest different ways you can do so.

Exploring Multimodality

Importantly, as we discussed on Thursday, these new models like ChatGPT-4o are the first “multimodal in / multimodal out” models, meaning that they can read, see, listen, and write, draw, and speak, all natively from one single model thereby allowing it to reason across modalities.

Sending an audio of a bark and the model generating an image of a dog as a response is a prime example of what reasoning across modalities is, as the model understands they both are ‘dogs’.

But here’s the thing: we do not yet understand this model’s power. As we explore its capabilities, we might encounter use cases that were impossible beforehand but not anymore.

But we all saw how ‘human’ ChatGPT-4o has become, to the point that you unknowingly might find yourself forgetting you are speaking with a robot at times.

The question then is: can AIs take the supportive role for humans?

I don’t have clear evidence on the subject, but the impressive engagement and retention results that companies like Character.AI are having, and Google research proving that AI is already potentially better than real clinicians at diagnosis dialogue (interrogating patients to understand their illness), might signal that doctors are in for a nasty surprise.

But let me be clear: I still stand by last week’s opinion on the future of doctors, but this shows that you can never be fully certain about AI.

Still, what about the next frontier in GPT-5?

The Conquer of Reason

While truly multimodal models have AI eyes, ears, and mouth all at once, an essential component of human intelligence, it doesn’t make AI models reason as well as humans.

We must not confuse the ability of our current frontier models to imitate reasoning with actual reasoning, just as we must not confuse reasoning with rote memorization. Knowing all laws of physics by heart doesn’t mean you actually understand them, and the same applies to AI.

And while we can’t claim to know how to make AI truly reason, AI labs are looking in one direction: long-inference MLLMs. I won’t go into too much detail as I’ve talked about it many times, but the idea is this:

Instead of simply responding to a request with the most probable concatenation of words as fast as possible, we give the model time to think. Thus, just as you will explore different solution paths when solving a complex math equation, models do the same.

You can do this in two ways:

In both cases, the results are dramatically superior, to the point that the only reason all our models aren’t working that way already is that they are extremely expensive.

A third option is neurosymbolic systems, like AlphaGeometry by Google Deepmind or rabbit’s R1 product, which combine human-crafted code and neural nets to create human-level reasoners.

The impact of such models is hard to predict, but we could be looking at AIs that could one day obliterate most white-collar jobs.

Of course, this is pure speculation, but you must remember you are competing against machines with Internet-level knowledge that can reason through that knowledge, and generate a million samples in seconds before deciding on one, making them fierce competitors for humans across the board.

Naturally, the impact would be asymmetric (some jobs would fall first). But which ones?

Toward a Value-Only Job Market

In particular, I would like to point out some corporate jobs, especially at the middle management level. Usually summarised as ‘meeting Joe’ or ‘meeting Jane,’ these jobs solely aim to attend meetings and allegedly coordinate projects or budgets.

Although they are sometimes useful, even vital in some cases, in more cluttered corporations, their job is basically to meet with other people.

Now, take a proposal like Google’s ‘AI Teammate,’ which will be released in 2025, and pack that with ‘GPT-5’ level intelligence with Internet-level knowledge and connected to all corporate systems, and you have the ultimate coworker.

However, it can get worse, as another possible consequence of ‘enhanced smartness’ is multi-agent systems.

An Army of AIs

One of the most promising avenues to increase AI performance is combining multiple AIs to solve a given problem.

As proven by research such as Society of Minds by MIT researchers, where the simple action of making different AIs discuss a possible solution even led to finding correct answers despite both having wrong intuitions at first:

Also, in a famous research paper, they created a software development agency based solely on AIs. Each agent (software developer, UI/UX designer, and even an AI CEO) collaborated to create entire video games with little human interaction.

Stanford researchers also implemented an impressive multi-agent system by putting 25 AIs into a simulated world. Eventually, the AIs developed unexpected behaviors like throwing parties or celebrating Valentine’s Day.

Simply put, if you let models interact, magical and unexpected things occur.

And with tools like Microsoft’s Autogen, building these AI armies becomes easy. In this multi-agent scenario, the need for entry-level workers would diminish considerably, as a seasoned employee could generate the output of many. However, for multi-agents to become useful, we need better AI models.

Moving on, at this point, we’ve mainly focused on white-collar activities. Does that mean that blue-collar jobs are untouchable? No, as AI robotics can also considerably impact the future of these workers.

But how?

**Sponsored Content**

A Newsletter You’ll Love: Stay ahead on the business of AI

Need ONE newsletter to stay on top of everything?

Prompts daily will make you more productive at work and help you look smarter in front of your friends (seriously!)

It’s also FREE to sign up!

A Physical Awakening

Until now, research by OpenAI suggested that white-collar workers, especially the higher your salary, are the most disrupted by AI.

According to this other study by the US National Bureau of Economic Research, historically, technological advancements have affected the most the top and lower ends of the wage spectrum. The former are replaced by automation, the latter see massive decreases in wage growth.

Although AI seems to be poised to affect higher wages more, mundane and physical jobs are also primed to be disrupted by AI through robotics, with companies like:

  • The recently announced The Bot Company, which intends to automatize home chores,

  • Cruise, Wayve, or Waymo to enable autonomous taxis,

  • or 1X’s Eve and to bring to life general-purpose robot androids,

Unsurprisingly, companies worldwide, specifically those belonging to The Magnificent Seven, are pouring billions into these start-ups or directly leading the charge, like NVIDIA.

But considering that, today, training a robot to tie a shoelace is considered ‘publishable’ in research by, among others, Stanford, it’s fair to say these robots are far from being a real threat for now.

That said, while fully autonomous robots are not expected soon, human-controlled robots are much closer than you might think.

For instance, BRIs (Brain-Robot Interfaces) are an increasingly popular application of AI too, with examples like Stanford’s NOIR, which allows you to mind-control robots, or BCIs (Brain-Computer Interfaces) like Neuralink, which allow you to mind-control computers.

And, of course, there are even potentially more disruptive changes that will eventually take humanity to AGI or even ASI (superintelligence). But I would be lying if I told you I know how to handle AGI if scientists can’t even get to agree on what the hell that really is.

But enough of my rant.

I have probably convinced you by now that things will evolve, and you will need to fit yourself in this AI world (because it won’t be the other way) and continue to evolve, too. Luckily, plenty of ways to handle this uncertain future and take action now exist.

The Circle of Life

I’ve divided the set of actions you can put into play now into three stages of AI disruption, aka ‘surviving AI’: Outrun the Prey, Lead the Pack, and Feed the Young. But before venturing into the three categories, consider this:

The goal of this post is that, after reading it, whether it’s about your job and future career prospects or your SMB or start-up, you can confidently evaluate your real exposure to future iterations of the technology.

The ‘Am I Getting Steamrolled Framework’ evaluates your current exposure, but the actions discussed today will also provide ways to evaluate your future exposure and take immediate action.

You don’t have to be better than AI.

In the first stage of AI disruption, AI will mainly be a tool for increased productivity. And as AI is predicted to have a huge deflationary effect on all industries (less time to market, overall better products, making survival a matter of minimizing costs of goods and services sold, etc.), the temptation for companies to ‘lose some weight’ will be enormous.

Based on CEOs openly acknowledging their layoff intentions in a PwC survey, its effects are becoming apparent already. Therefore, I don’t think you have the luxury to wait and see.

Thus, you must ‘ProAct’ instead of ‘ReAct’. But being proactive and starting to use AI in your daily workflows isn’t about beating the machine. Because here’s the thing:

You don’t have to. You simply need to beat other humans.

Like saving yourself from being eaten by a lion boils down to outrunning the other human prey, surviving the first state of AI disruption will also be about outrunning your colleagues.

As the industry saying goes, AI won’t substitute you; a human who uses AI will.

But how do I ensure I objectively outrun my others? Simple, metrics. You must first confront a potentially harsh reality: Is your performance measurable?

If the performance in your job is hard to measure, people are simply going to assume you don’t create value, and that’s the first step into dismissal.

For example, we have salespeople and ‘meeting Joes’. For the former, you have an objective, unequivocal way of measuring their value to the company: how many sales they’ve made. The path to salvation is clear for salespeople: sell more than the rest. 

On the other side of the spectrum, in most companies, especially corporate America, we can easily identify the typical employee whose job is meeting with others. And if you can’t think of that someone, that’s you. Therefore, as AI forces cost-cutting, having a metric to back up your value could save your job.

One simple test is to have someone above you, your boss or your bosses’ boss,(or a customer if you’re an entrepreneur), articulate how they measure your success. If the response is a bunch of meaningless jargon, let me tell you:

They don’t have a single clue. And when the ‘AI wrecking ball’ arrives at your company and their boss asks them to justify why you shouldn’t get fired, well, you will get fired.

For example, circling back to the ‘meeting Joe’ example, with Google’s AI Teammate or David Sachs’ new start-up Glue, in a world where AIs will have total access to all project and product history and will also have ears, voice, eyes, and hands to communicate and support anyone who requires the typical information these people would gather in never-ending meeting sprees, those types of ‘hard-to-measure-its-value jobs’ will be begging to be destroyed.

Bottom line:

  • If your job or company has a metric that clearly defines their value to your boss or customers, laser focus on it, so while you won’t be better than future AI models, you are better than other humans who will take the fall for you.

  • If you are one of those employees or entrepreneurs whose value is hard to define or based on colleagues’ or customers’ goodwill towards you, there’s still hope; being an early AI adopter while exploring career or company pivots will certainly buy you time. Each company and job is unique and must be addressed carefully, but self-awareness is the first step.

But at a later stage of AI, when AI truly embodies this idea of ‘hyper reasoning machines’ or ‘embodied intelligence,’ things get much trickier. In fact, in the ‘Leading the Pack’ stage of AI disruption, outrunning other humans won’t be enough.

Subscribe to Leaders to read the rest.

Become a paying subscriber of Leaders to get access to this post and other subscriber-only content.

Already a paying subscriber? Sign In

A subscription gets you:
High-signal deep-dives into the most advanced AI in the world in a easy-to-understand language
Additional insights to other cutting-edge research you should be paying attention to
Curiosity-inducing facts and reflections to make you the most interesting person in the room