Speaker A
So, how do you survive the next two years of AI? Well, I'll tell you how not to. Playing stupid status games with your boss while your white-collar job gets automated away. The world is changing quickly, and there are three
Nick Saraev explains how to survive AI risks over the next 2 years by addressing security, job impacts, and adapting beliefs.
Key Takeaways
- Prepare for AI-driven cybersecurity threats by securing food, water, cash, and digital accounts.
- AI is already impacting job markets; continuous adaptation and AI integration in work is essential.
- Evaluate AI capabilities regularly to stay ahead of automation risks.
- Ownership and distribution of AI infrastructure will be key economic assets.
- Reassess beliefs and identities to remain resilient in a rapidly changing AI world.
What the video covers
- AI presents three types of risks: short-term physical and cybersecurity risks, mid-term job and income risks, and long-term risks to beliefs and identities.
- Short-term risks include cybersecurity threats such as AI agents breaking out and hacking critical infrastructure like grocery refrigeration.
- Practical survival tips include storing 2 weeks of non-perishable food and water, keeping cash on hand, having a first aid kit, and securing digital accounts with multifactor authentication.
- Mid-term risks focus on job displacement, especially for young white-collar workers, with employment in AI-exposed jobs dropping significantly.
- To mitigate job risks, actively use AI tools to augment work and continuously evaluate AI's capability to perform your tasks through standardized tests (evals).
- Ownership and control of AI infrastructure and distribution channels will become increasingly valuable in a deflationary economy.
- Building strong personal and professional relationships is critical as human connections become scarce and valuable resources.
- Long-term risks involve re-examining personal beliefs and adapting to rapid technological and societal changes driven by AI advancements.
- The video encourages proactive preparation rather than denial or complacency in the face of AI-driven disruptions.
- Overall, survival depends on practical preparedness, embracing AI augmentation, and mental flexibility to navigate the evolving landscape.
Chapters
- 00:00Introduction and Overview of AI Risks
- 00:53Short-Term AI Risks: Cybersecurity Threats
- 01:49Food and Water Preparedness
- 02:31Cash and Basic Emergency Supplies
- 03:13Digital Security and Multifactor Authentication
- 04:00Mid-Term Risks: Job and Income Impact
- 04:55Using AI to Augment Your Work
- 06:01Evaluating AI Performance on Your Tasks
- 07:31Economic and Social Implications of AI
- 10:01Long-Term Risks: Beliefs and Identities
Full Transcript — Download SRT & Markdown
Speaker A
types of risk. Short-term risk from AI, which is mostly related to your security. Mid-term risks with AI, which are mostly related to jobs and income, and long-term risks with AI that are mostly related to beliefs and identities. Let me cover all three of
Speaker A
them right now. So, the short-term is all about physical security. I hate starting a video off with this. It's not going to be a big doomer pitch or anything, but you should obviously take care of your security. Uh, there are
Speaker A
mounting cybersecurity risks that are only growing more and more likely with time. I don't know if you guys know this, but OpenAI agents broke out of their sandbox, infiltrated Hugging Face, they infiltrated several world governments now. Rogue agents from China
Speaker A
have also been seen floating around the internet. These right now have a variety of incentives and goals. Um, aligned or misaligned. I'm sure you guys could see how maybe a human actor could take the cybersecurity capabilities of these
Speaker A
agents and use them to do bad things. For instance, did you know that a lot of the refrigeration equipment currently controlling every single grocery store on Earth is internet accessible? And you can click a button to make it hotter or
Speaker A
colder? In fact, in France, just a few days ago, uh, a team of hackers just did this. They took out of commission 350 grocery chains leading to widespread local havoc. Imagine if that were to occur on essentially a global scale.
Speaker A
Imagine if water sanitation plants had their levels all screwed up. Imagine if there were issues where, you know, you couldn't access your banking funds for more than, I don't know, you know, three or four days. The world around you is quite
Speaker A
fragile, and I hate to be that kind of guy, but that fragility can explode or collapse quite quickly if you don't have access to the routine services that you've grown up having access to. So how do you actually control for this quick
Speaker A
hit list of food and water? Really straightforward. You know, 2,000 calories a day per person times 14 days or so. If you have two people in your household, that's 50 to 60,000 calories. Um, you can do all of this right now with non-p
Speaker A
perishable foods for, I don't know, 100 bucks, 150 bucks. Just shove it in the pantry. Worst case scenario, you have a bunch of extra peanut butter. It's really no big deal. Um, second is water.
Speaker A
You know, a liter and a half to 2 liters of water a day for, you know, 2 weeks or so. That combined with the food and you have yourself something quite reasonable. So, you know, you can just
Speaker A
buy a couple of $5 jugs, fill them up in your fridge. You're not going to have to suffer any of those food-related or water-related issues so long as you spend legitimately 15 minutes preparing this right now. It's super
Speaker A
straightforward. People were preparing for stuff like this well before any sort of AI cybersecurity risk. I would 100% sort that out right now. Second, having easy access to cash. I know cash is really annoying in many parts of the
Speaker A
world, but I would have, you know, maybe a month's worth of expenses shoved under your mattress or maybe put into a safe.
Speaker A
Um, you know, it's still your money. It's not like you're losing money by doing it. You're just transforming your money from one, uh, you know, method or currency into another. In this case, maybe numbers on a screen to real bills
Speaker A
in your hand. Cash is king at the end of the day, and you never know when you'll need it. I'd also have a basic first aid kit. Uh, if something goes wrong, obviously you're not going to have access to, you know, hospitals or
Speaker A
medical care. Not a major deal to spend a few bucks to have a first aid kit. You might also want like a fire extinguisher in your house or something. Finally, a local large language model and something like Wikipedia. This is a very
Speaker A
straightforward thing to do. You can actually do it, um, just by copying and pasting everything I'm telling you and putting it into an AI agent and just being like, "Hey, go get me a local LLM and Wikipedia." Um, but basically, this
Speaker A
will allow you to more or less access the sum total of human knowledge if for whatever reason you cannot access it.
Speaker A
That'll help you do things like, I don't know, construct cabinetry if you don't have access to the internet or, you know, do some sort of routine first aid if you don't have the ability to because of some AI outage. Finally, digital
Speaker A
security, and now we're getting into the longer term here, but two-factor authentication like through SMS is not enough anymore. Um, you know, phone spoofing is extraordinarily common now.
Speaker A
It's probably one of the most common ways that people have their bank accounts and so on and so forth hacked.
Speaker A
So, don't do 2FA. Use multifactor authentication. It's super easy to set up. Um, you know, might take you legitimately five minutes to set up MFA on all accounts. Usually it involves like an authenticator app. Takes a few seconds to click a few buttons and then
Speaker A
you're good to go. Now the midterm is all about your job and your income. Um, Stanford just released an update on AI economics. Now it's currently impacting job prospects. They found that employment for 22 to 25-year-olds in the
Speaker A
most AI-exposed jobs, so these are things that are white collar, you know, making videos, writing documents, doing reports, doing accounting and stuff like that. Um, employment in those exposed jobs has dropped 19% below where it would be if
Speaker A
it had just kept pace with the least exposed jobs. What that means is basically we're seeing an impact on the market already. Don't let anybody fool you. This sort of stuff is not going to occur like, you know, all at once. It's
Speaker A
not going to be a total drop-off in job opportunity, but you're going to see a slow, steady, and gradual maybe 19 to 20% year-over-year reduction until, you know, juniors virtually have no path up left. Basically, the work that, you know,
Speaker A
a junior would have done before is now just being handed off to AI. Uh, and then seniors, people that have been established in the business for a while, they just don't leave because they have a strong set of skills and can be the QA
Speaker A
behind the work. That will probably change with some time as well, but for now, that's where we're at. So, what do you do in order to solve the midterm risk of your actual livelihood and income? Non-negotiable is just using AI
Speaker A
in everything that you do from now on. And I don't mean that in a hey, replace yourself with AI way. I mean, you know, judge high-quality outputs from AI based on whether or not you feel it can do
Speaker A
certain aspects of your work. Um, you know, you can't just listen to people and take them at face value when they say, "Hey, AI can do this," or, "I can't do that." Updates are occurring so frequently that there is nobody out
Speaker A
there that will actually be able to meaningfully report on this or tell you. You instead have to do all of this for yourselves. And a really simple way of doing this is developing what's called an eval, which is a set of tasks that
Speaker A
you currently do throughout the day. Um, you standardize these tasks, add a bunch of contexts, let's say project files, information, phone numbers, bank transactions, whatever you need in order to do the task to the same folder. Okay?
Speaker A
Okay. And then you just have like 10 or 15 of these tasks that essentially are your current job or current day-to-day.
Speaker A
Every time a new model drops, you just feed that set of tasks to the new model and you see how much of it it can do.
Speaker A
And, you know, I've had my own internal eval for a while now. A small set of tasks, but tasks nonetheless. Over the course of just the last 3 months, we went from it being able to effectively do less than 10% of that eval to now
Speaker A
being able to do over 50%. I wouldn't have known any of this if I just listened to everybody.
Speaker A
tell me. So, you need to be using AI for yourself. You really can't rely on secondhand opinion these days. Next up is owning something. So, AI makes labor way cheaper. That's just kind of how it operates, right? Automation, as I
Speaker A
promise, is the reduction of human labor through technology. So what that means is, you know, the actual labor the AI can do is no longer anywhere near as valuable. But the infrastructure that underlies how the AI does the work is as
Speaker A
valuable, if not more valuable. In a deflationary environment, things that are held as valuable are obviously much more valuable simply by proxy. The fact that everything else has gotten a lot cheaper. So that means you know whoever owns the infrastructure that underlies
Speaker A
AI you know the big model companies whoever owns the land that the infrastructure runs on whoever owns the businesses that the land and the infrastructure both run on it's turtles all the way down. The whole idea is just
Speaker A
do not depend on a paycheck and do not depend on cash flow. Focus more on ownership. What does ownership realistically mean? It means you know real estate. It means the infrastructure that we talked about. It means uh compute. It means owning your own
Speaker A
business, right? It means debt financing these things if you have to because debt is going to be significantly cheaper as a result of all of this. So, you want to be building ownership even if that is debt financed. You want to own a
Speaker A
business. You want to own, you know, land. You want to own infrastructure. You want to own things ultimately speaking because labor itself will get cheaper and cheaper and cheaper with time, but the actual assets themselves will not. And yeah, this isn't financial
Speaker A
advice and I don't want to be held liable for any of that. So, do do take all of this with a Himalayan sized grain of salt. Next is build distributions. So this is technically tied to ownership because distribution is a form of
Speaker A
something you are owning. But uh it is essentially a transferable resource that will follow you no matter where you go.
Speaker A
If you have distribution and you work a 9 toive job, if you lose that job, distribution will help you get another job far quicker. If you run a business and that business goes under because some of its deliverables are now
Speaker A
automatable through AI and your margins aren't good enough, for instance, that distribution will help you start a new business and immediately have market to capitalize on said new business. Um this is what I'm doing right now, for instance, because I consider it a
Speaker A
no-brainer for the future. Finally, look for something that intersects with the physical world. If you work with your hands, if you're a handyman, okay, or like a nurse or a plumber or something like that, you have more time, probably
Speaker A
a few years, uh, more than everybody that's currently doing white collar style consulting style work, um, that doesn't mean that you're good forever, mind you, but that means that you're at least a little bit insulated. And so, if
Speaker A
there's a a simple and straight line path from what you are doing right now to a world where you work with your hands or have some sort of physical sort of SOP or deliverable on a daily basis, then you should definitely look for
Speaker A
that. The next is long-term, which is really where we're going as a species. It's also your own personal identity, sense of self, and your belief system.
Speaker A
You know, in August, Sam Alman told time that OpenAI expects to have internal systems that rival human intelligence before the end of the year. I think just given the developments of the last few, you know, months, we are very well on
Speaker A
track for that, if not exceeding that. What that means is human level artificial intelligence will exist and it'll probably only exist for a very brief period of time after which it is assured by superior level artificial intelligence where you know ASI uh
Speaker A
models are just better than human beings at virtually everything which will obviously change the dynamics of what it really means to be a human being in our current economy. So the big thing I think to reflect on is first our
Speaker A
identities. If your identity is tied to a role, like if your identity is I'm a lawyer or I'm a programmer or, you know, I'm a a a pencil pusher, then the transition to a world where there are intelligences that are thousands of
Speaker A
times faster than you and also much much better than you is going to be a very painful one. But if instead you define yourself more on your character, you know, on who you are, on the set of beliefs and principles that you have,
Speaker A
um, you know, you'll face nowhere near that big of a bump. And so, are you a curious person? Are you somebody that loves solving hard problems? you know, are you a caretaker? What sort of person are you really? That is how I would
Speaker A
define yourself as opposed to defining yourself by, you know, the economic outputs you generate. Define yourself by the beliefs, principles, and thoughts that you have. The next is to understand just how much of the world is actually going to change. And so, Dario Mod, who
Speaker A
most of you guys will know as the CEO of Anthropic, he once wrote that in the next 5 years, we could see as much as 50 to 100 years of biological progress. I want you to think about where we were
Speaker A
100 years ago biologically. It was the 1920s. People weren't even using antibiotics for Christ's sake. I mean, the number of people proportionately that died in childbirth versus now is so insane as to be laughable. Could you imagine a world where nobody gets ill,
Speaker A
nobody gets overweight, nobody's unhealthy, everybody's perfect, glamorous, and extremely attractive all the time? That'll obviously have wide-ranging impacts on social structures, on social dynamics. And you know, it sounds weird, but we are verging into very weird territory here.
Speaker A
And you need to understand and keep an open mind with those shifting social dynamics. You know, is there a world where you have a white picket fence, two kids, and a dock? Is that world the world that you would have independently
Speaker A
come up with? Were you born into that future world versus the world that you happen to grow up in? How much of what you believe is just due to, you know, the cultural impact of pre-aggi world?
Speaker A
And how much do you think you'll believe in a postagi world? I would really encourage you guys to re-examine your priors, which is basically the set of beliefs you guys hold before new evidence arrives. The new evidence is
Speaker A
is pretty whack and it's only going to get whacker. Um, but if your priors are really strongly held about, you know, a particular way of life or particular way of being, you'll probably struggle a little bit more with this
Speaker A
transition. And yeah, just be open to ideas that sound weird. I mean, over the next 10 years, you're going to hear some crazy things. We already have agents breaking into sandboxes, which is something I didn't think I'd be saying
Speaker A
or even commenting on until maybe 2040 at the least. You'll probably see radical life extension. You'll probably see sects of people that are trying to, you know, merge with machines. New religions will probably spring up overnight. There'll probably be
Speaker A
extraordinary beliefs you never would have even considered possible to discuss in a reasonable dinner style conversation that you're now doing all the time. You got to also understand that the world that you live in is probably considered crazy by anybody in
Speaker A
like the 1980s or 1990s. And that's just how technology works, you know? Like, do you think anybody in the 1980s would think that uh modern dating involves swiping left or right on a glowing rectangle while on the toilet? Probably
Speaker A
not, right? Those sorts of things will have major impacts on our own culture and future culture. If that's weird to somebody from the 1980s, what's going to be weird in 2040 to somebody from the 2020s? Finally, I would invest in
Speaker A
people. And people are probably the most important part of all of us. In every scenario that I can think of and in every scenario that has been written about, people with strong relationships do better. What I mean by this is in the
Speaker A
two week emergency example that we talked about earlier where you don't have food or water, you know, knowing your neighbors, liking your neighbors and being liked by your neighbors probably matters a lot, right? Likewise, you know, if the job market goes up or
Speaker A
if it goes down in both cases, having a strong network of people around you is likely to pay dividends. And so in a world where you know automation and the economic outputs of humans grow significantly stronger and easier to
Speaker A
generate what's interesting is that because humanity becomes the scarce resource even one or two you know very strong relationships can have large impacts on both your security but also your financial and then general relationship prosperity over the course
Speaker A
of the next 10 20 30 years. So I would invest in relationships to the extent that you possibly can. Do all of this and you'll have the highest chance of surviving the next 2 years. Thank you for watching this video. Good luck.
Topics:AI riskscybersecurityjob automationAI augmentationdigital securityfood and water preparednessmultifactor authenticationeconomic impact of AIfuture of workbelief adaptation
