An exploration of how Artificial Capable Intelligence will transform employment, human purpose, government, and the very fabric of society.
This will be quite a long blog post because it pervades all of society. There will not be one single life action that will not be uprooted by AI. After all, it is a general intelligence. Unlike the past 2 essays which are largely about how the architectures work & the tangible bottlenecks, this is less based on research & more an intuition about what will happen. After all, OpenAI is not another B2B SAAS tool, it requires innovation on the frontier. Therefore, all future insights are by definition predictive.
This essay probably does not begin with a rigid place. However, this essay does come with the perception that ACI will happen. We will get Artificial Capable Intelligence. In the digital world, this will come in the form of agentic workflows. Essentially, you have digital workers of all kinds now that can more or less perform their job equally as well from home as they can from the office. Thus, in this world, what do they do? What jobs/skillsets are still important? How do we acquire meaning? Do we all become data harvesters for the bots? Does it change how we mate? How long we live? How a government functions in this civilisation?
This are all questions we will explore.
Firstly, my own predictions as of June, 2025.
This is the one on humans minds. All over LinkedIn, all you see is "You won't be replaced by AI, you'll be replaced by someone who uses AI". I generally disagree. On a long enough time horizon, all human labour will be replaced. While people reason from analogy saying all technological cycles create new jobs, ones that were unheard of before. This time it's a different outcome. There may be different jobs. Why wouldn't AGI just do those jobs?
Fundamentally, if we create an intelligence that is more intelligent than us, why wouldn't we allow it to do all the work? To think that humans will still have jobs in the digital or physical world holds the belief that we can somehow do tasks/actions that robots/agents cannot do. This is betting against the progress of technology.
While many people say that we already have psuedo-UBI in corporate America. The "jobs" we hold today would seem trivial to farmers from 20,000 years ago. They are still necessary. We still are yet to have worldwide abundance. There is still shortages of food, clean water, healthcare, shelter, electricity etc. Therefore, there is an onus on mankind to continue improving the efficiency of the economy such that we slowly move toward that. On the presumption that we do reach ACI, all jobs will be eliminated.
Some people posit that new jobs will be created that are uniquely human. Humans crave human content so there will be human influencers. AIs need human data & therefore humans will log their lives for data for the AIs. Humans will test products, recipes made by human business owners before they are complete. These things could all be true. However, I'd argue the demand curves for these jobs won't meet 8 billion people and thus it will be useless anyway. Each human only has 8 hours in their day, and therefore, they cannot watch a dozen content creators even if they wanted to. Plus, most of these "creative fields" have power law dynamics anyhow. 0.1% of people get 99% of the fruits. That isn't going to change. Data & Recipe testing could be one for a little bit.
I could plausibly see the government implementing mandatory jobs though. Humans have to show up and do something to the economy. While it will not actually drive forward economy, it could prevent Brave New World type tendencies coming to fruition. Meaning & wellbeing are so important to the function of a nation that it might become essentially to artificially create meaning through jobs like this. How this would conflict with objectives of a high-performing efficient economy? I am not sure.
This is quite a long-term view though. Obviously, agents & robotics would be the 2 necessary pre-requisites to this. My guess is agents will be here in a very meaningful capacity within 5 years. If you do not have a senior role within 5 years, it'll be tough to contribute meaningfully to the economy. It seems a while before humans stop steering the company's direction as well which means that taste will remain important, and therefore, there will be jobs in a meaningful capacity.
In the short-term, I think junior roles are most at-risk. They are the most simplistic, execution-based work that agents will quite easily be able to handle with better scaffolding, memory & task-length. Think job function matters less than the skill level that the individual has gotten. 10x marketers and 10x devs will both have jobs for quite a long time. Perhaps 5+ years. But, average marketers and average devs are both in trouble. It doesn't really matter what function. Boilerplate CRUD coding seems to be automated. SDRs that just hit send on cold email functions will go in the next three years. Marketers that spent their time churning out blog post after blog post & doing media buying on Meta/Google will be gone. Book-keepers are in trouble. So on, so forth. So, it's less about job function than it is about skill level.
Though, physical jobs seem to have an advantage or digital ones. Mainly because of moravecs paradox. What humans find simple is difficult for robots. This is because (1) there is insufficient training data for real-world interactions that NVIDIA is trying to build at the moment (2) human evolution has favoured physical tasks; digital tasks like software engineering & even language are relatively new & thus it has not been scrutinised by evolution nearly as much. With companies like FigureAI, Optimus, Boston Robotics, I suspect this will be solved as well.
Over a long enough time horizon, this won't matter. All jobs will be automated. There won't be jobs a human has uniquely do, that artificial intelligence won't be able to do. In the short-term, this is a tricky question.
Physical > Digital businesses. Blue-collar labour is harder to automate. Plumbing businesses, construction businesses, healthcare businesses, industrial businesses. These are all better businesses than marketing agencies, recruiting agencies, IT services companies, accounting businesses. Maybe, even software companies. The western world could become muchlike China where Software isn't all that big & doesn't command 10-20x ARR multiples because the amount of STEM graduates means everyone builds the tools internally.
Industrials & Real Estate seems quite useful still. It'll be a while before the value of raw materials drops considerably as AI invents ways of doing processes more efficiently for manufacturing or industrials company, while it also can build cities so quickly lands become irrelevant. If I owned property in Bronte, LA, NYC, I'd most certainly be holding.
In terms of skills, my sense is that most skills take quite a while to develop. So, hedging by developing skills that require deep problem-solving will probably hold the most utility. These are all inherently creative problems around (1) storytelling: brand & media (2) technical product know-how in the form of software, hardware or biology (3) sales. That's my overall intuition. I'd pick some combination of those 5 things. I'll probably pick software, sales & storytelling. Though, I'm unconfident because software might become less useful in 2-3 years because AI can be trained on all of Github which makes it difficult to determine if it'll still be functionally necessary. Skills I wouldn't learn are accounting, financial analysis, performance marketing, law. These seem prime for digital disruption. They seem quite low leverage & replaceable. The fact Deloitte employs 500,000 people seems proof of this. Generally, people like Sholto Douglas say to pick a creative/technical domain and get to the frontier.
Creators/Digital World. I'm not too invested in this world nowadays. Though, I spent lots of time thinking about it when I made 50 youtube videos when I was 15/16. Hormozi thinks there will be tonnes of non-human AI creators. Tonne of dogs & cats that people watch all the time. I am not quite sure how this is true? Perhaps for entertainment? Surely not for education. My sense is people will still want to learn from other humans.
Most humans get their meaning from sensations. Physical sensations, emotional sensations (visual & auditory) & intellectual sensations. Your job, your relationships & physical activity are your biggest predictors of wellbeing. This will not change. The worrying thing is actually that most humans rely on the form of a 40 hour/week job to satisfy one of these sensations, typically intellectually. If jobs are to be automated, this paradigm will shift. People will be forced to find new ways of satisfying that craving, or they will have to lean on other sensations.
Both could work. Humans could further find more enjoyment from developing relationships with other humans. They could satisfy such cravings by seeing more things, hearing more sounds. They could also explore intellectual pursuits for hobby, tinkering reasons alone. Most people are not the best at what they do & still find it satisfying to some extent. Perhaps the allurement of career progression & nature of competition/status is the actual reason.
In many ways, the pursuit of meaning/purpose becomes more authentic in such a world. In creative world, you are not constrained by having to mine coal everynight & instead are free to build interesting castles & ships.
Human Status/Mating. A huge subset of meaning/purpose is the ability to mate. Historically, this has been based on status. Status for females is generally in the form of fertility & attractiveness. For men, it has been contingent upon their ability to gain resources & wealth. If men can no longer compete for resources, what status game will be play. That book says the three status games are virtue, dominance and competence (resources, intelligence). Perhaps, the other status games become the important ones. Dominance was the first big status game. Then competence. So, maybe virtue becomes the next biggest status game? Women's status seems to be fairly predictable. It has not changed much, only how they tried to show it off.
ACI will make the government take this threat more seriously. Will they intervene & regulate AI? Will they nationalise the leading labs? Leopold has very strong opinions they should do this. AI labs need to make the clusters & the labs themselves more secure. This is a national security project that has vast implications for the military. We cannot have some random SF startup having access to the weights & everything else alongside it, rather than the government.
There are fears that this would signal to China this is an arms race, sparking the next cold war. Arguably this is already happening. DeepSeek has met with 2nd in command of Chinese government. Therefore, it may be already be underway and thus is worth putting in the hands of the government already.
I am not too well read on this. Tyler Cowen thinks economic growth never really gets above 1%. This seems a little silly. EG seems to be the right index to measure the extent of ACI? If you have ACI, then the amount of goods & services in the economy should proliferate and be much much higher than it otherwise would be.
For one case study, the cost of digital services should dramatically decrease and thus there should be far less of them. Consumers should be able to build their own software. This will overtime extend to everything. The price of physical products like clothes, beds, housing, food, equipment, hardware should decrease substantially as robots are able to do it & we can pay them far far less than humans and they become more efficient so they can produce more per hour than we possibly could. Hence, I take this side & believe there will be an abundance of growth akin to China's in the late 1900s after they implement Belt&Road trade, SEZs and whatever other policies they did.
Mckinley — American was really protectionist at first.
All non-tech sectors have had inflationary outcomes. Housing, Education & Healthcare all gone up. Tech affected sectors all decreased. Eye Surgery the lone exception in healthcare.
Middle class been squashed. Top tier knowledge workers in cities. Middle class have too commute to work — have house 3 hrs away. Lower class stay in social housing.
Issues very obvious. Energy supply becoming more abundant & cheaper. AI Data Centres being built out. Models getting better. Billions of robots being deployed.
Immigration. Need for low-skilled immigrants anymore? High-skilled ones?
AI automated researchers. Closed loop on software engineering = first to ASI.
Coding research to improve Llama tool chain.
Most of code going toward effort, it can run tests, find issues & writes higher quality code than most people on the team.
Not one lab that dominates. Not one general app that does everything. Specialisation between the groups.
Software Engineering 100x. Acquarian → Small % of humanity. This lead to creative/cultural pursuits & people spend less time working, more time on culture.
How does Meta provide customer support?
Automate jobs. Not how history of technology has worked. Jobs are pursued that otherwise wouldn't have been. For example, customer support waste of time. Now, Meta does it because 90% can be handled meaning hiring the other 10% might be important as well.
Davidson estimates 10,000 more. 10x for algorithmic efficiencies & 1000x more compute. No amount of compute would be able to take you to fully automated.
GPT4 = 1,000,000 more compute than GPT 2. Not many differences in algorithmic efficiencies.
How fast will it go?
Years
Trillions of dollars of compute clusters will be built
Potential for it to be built in the middle east — infrastructure there
Cannot have it in the power of foreign enemies. Even if middle east promises it. US regretted reliance on middle east energy supply in the 70s.
America has power to build it — natural gas & deregulatory agendas
2023 was AI wake up — TWSC & NVDA were very obvious buys early on once AI infra took off — march 23 was when it really started happening.