Mark Zuckerberg of Meta fame posted a manifesto. He lays out an idea of a "future for everyone", although I don't think his group of everyone comprises those outside his tax bracket. The screed itself is too long and glazes over ideas that are inconvenient, as one might expect of a silicon valley bro archetype. It's almost certainly AI-generated, and thus should be discarded as half-thought-through slop, but let's give it the time of day. I'll break down particular pain points and how his ideology fails to accommodate the real world inside his "positive AI future."
Mark Calls His Shots (And Misses Every Time)
Let's start with the obvious: Silicon Valley at large, and Mark Zuckerberg in particular, has not had a good track record on what the next big thing is. Not even 5 years ago he renamed his entire company after an already dead project, the Metaverse. His decision to focus on Metaverse and Web3 projects has cost the company tens of billions conservatively. This man is not to be trusted with which way the literal wind is blowing, let alone what the future of society looks like.
What growth there has been in the Meta family of products has come from extremely expensive purchases of already popular services (and a bunch of much less successful things in between), and those services are nearly universally worse off for it. Meta, like Google and Microsoft, simply absorbs things and squeezes the dollars out until the next thing comes along that they can acquire or clone. So in between the acquisitions Mark is left with a lot of time to think and too much money to fail, a truly nasty combination.
This Manifesto Is Not About The Future
The article can largely be broken down into four parts: the intro (~700 words), plans for the technology (~790 words), risks of the technology (~4400 words), Meta's "philosophy" and wrap-up (~610 words). By word count, this is in no way a piece about how AI should be for everyone; instead it's a large list of all the ways Meta is "responsibly" developing this technology, or things it can't control that he feels they should get in front of. We'll still walk through his talking points, but we are not required to accept his premise. His focus betrays what the actual message is.
Intro: AI Is Cool, Trust Me! - Cool And Trustworthy Guy Mark Zuckerberg
Mark is as divorced from reality as one might expect from his ilk. His manifesto centers upon an idea of "abundant intelligence", but I don't think we are lacking intelligence in today's world. There are so many smart people doing smart things, well outside of AI development. Many of these developments come from the steady progress of scientific understanding, and our continued ability to operate on larger and larger scales. AI is not the one doing these things, people are. AI will not be what fuels an intelligence boom, it will be what kills it.
He also insists that this fantasy of "artificial super-intelligence" is just around the corner with no actual backing to that claim, as models are still failing to consistently pass many benchmarks, which they themselves are also flawed as measures. Claims of AI super-intelligence/superpowers/whatever else are not new, but are particularly galling since they've been regularly made every few weeks for the last several years. It's like the claims of the end-times being imminent; so many times we've heard the same story with no evidence that I see no reason to even pretend to believe them.
Mark then harps on how humanity has endured technological changes before, and we as a society will do it yet again! Those other revolutions didn't operate in the same way that the "AI revolution" seems to be shaking out however. In the industrial revolution, there were jobs lost in manufacturing as it became mechanized, but those jobs shifted to operating the machinery, logistics, and white collar work. There were other sectors that came about to support the shifting needs of society as it became something new.
The computer revolution then came and still required people to operate the computers, and very skilled individuals to make and maintain these early machines. The white collar and knowledge sector shifted focus and new ways of working came about with networking, mass storage devices, and other developments. Again, society shifted but there were still ways and places for individuals to make a living.
The "AI revolution" is something else entirely. It aims to subvert and replace the human at every level. AI agents are already being used to coordinate and spin up other agents for tasks, threatening to replace entire departments and fields (provided you don't care about being wrong or it working long term, which the C-suite often doesn't). These things can expand and replicate in a way never before experienced. There is some work for humans in cleaning up and fixing the mistakes of the agents, but those jobs will shrink too if even a fraction of the improvements the AI bros squawk about come true.
See, these sorts of screeds only seem to come out after 99 other instances out of CEOs, CTOs, and whoever else gloating that they're cutting jobs in favor of the AI future, reducing headcount and losing valuable actual intelligence. Every so often they feel a need to placate the masses and reassure them that there's going to be a place for them in the AI future, it's for everyone, can't you see it?
REWORD: Here's the rub though: people don't eat intelligence. People can't pay their rent or insurance premiums with intelligence. People aren't too stupid to succeed by default. We're not limited by our intelligence, we're limited by our resources; said resources that are actively being hoovered up by every fucking company doubling down on the AI boom. They'll supposedly be diffused through unclear means, and our lives will be better for it. That's not how capital flows though, resources don't "trickle down". Given the last several hundred years, there's no reason to believe that we wouldn't be asked to do more with less, and to be grateful for the opportunity to do so.
His half-hearted attempt to relate to real problems falls flat: "Soon [artificial intelligence] will increasingly help discover new knowledge -- ranging from discovering new drugs to cure a family member's disease to finding new ways to improve your business." With what money am I buying these assuredly prohibitively expensive drugs, Mark? There are many diseases that could largely be eradicated if not for the extreme costs set by the healthcare companies controlling them. There are radical new treatments for cancer stuck in Cuba because we as a nation hold a grudge and seek to punish the Cuban people. Our problem is not that we lack ideas, it's that most lack the means to access those ideas.
In Mark's world, you can just girlboss your way to success from zero to comfort via small business. Reality says that you need money and stability to have the option to pursue your own business, hobbies, and the like. You cannot start a business from nothing barring very special circumstances. AI does not give you starting capital, or a safe place to stay while things get going. AI can generate ideas but if the real money was in what the AI was generating, why would Meta/Anthropic/OpenAI not just do the things themselves? It's the same reason people sell courses about how to sell courses instead of, you know, making the businesses they bloviate about; the money is in selling shovels.
Sidenote: The False Dichotomy
His promise of imminent super-intelligence is overshadowed by his fixation that this technology will either be absolutely controlled by a select few (of which he would certainly be one), or be totally free and in the hands of every Tom, Dick, and Jane. He harps on the idea of this technology needing to be in the hands of everyone a LOT. I'm not going to say no one would benefit, but these benefits should be questioned. In my view, the utility of technological advancements has significantly diminished as time has gone on. The Industrial Revolution change a lot of lives, computers a lot too; I can't say I feel like our recent developments like 5G, satellite internet, or iOS 17 did the same thing though. Other more useful developments like genetic testing and at-scale renewable electricity are still outside of the reach of many individuals. While AI's "use" case can be broad, there are very real harms that are being papered over that will be litigated in the future, once the damage has been done.
Part 2: "Real" "Specific" Examples
Mark struggles to paint a convincing picture of what this proposed AI could do (the majority of which is stated here). Big picture, I don't understand what Mark thinks AI can do that the internet hasn't already enabled. People already have access to all these tools, and free knowledge, and benefit from science! My life isn't better because I can have a computer program spoon feed me information that was already accessible. My life cannot be transformed by technology that just sucks money and spews toxins. There's a claim of utility that I just cannot and do not see playing out in reality, and Mark has a very significant and real conflict of interest when it comes to convincing people it can.
First up is the personal agent, the one that is your 24/7 friend, support, and administrative assistant. He talks about it having access to every bit of your personal life, including finances, household information, and personal relationships. The huge risk involved in that is acknowledged, as he spends the other half of the paragraph talking about how it'll be fine because they'll have the best privacy and security, trust them. Please don't look at the pile of security incidents that grows every day because of the acceleration of AI-enabled hacking.
He tries to draw from his own life, but he should understand that he lives in an entirely separate reality from almost anyone else reading this piece. His agent "flags interesting information" and helps prototype ideas, which, cool I guess? Sounds like you're just replacing a trusted coworker or peer with a robot, but alright man. Engineers have fleshed out prototype ideas with rubber duck debugging for decades now, and I'm not convinced this is at all better. He also talks about baking with his daughter, which for some reason requires personalized recipes (an impulse I just can't understand, why do I need custom recipes?) and the agent to order the groceries (which could be a cute outing for any normal person), and then he has the AI watch them bake because apparently it's got insights they just couldn't have come to otherwise.
Next is the "creative multiplier" ideal of AI, where it supposedly enables people to do things far beyond what anyone could do before. Again, he brings up this poor child, who can code and produce videos, a thing children have already done for decades. He got started coding at an age not older than her, and there are languages specifically developed to be more approachable to children as well, so why involve AI? The problem is that he's not excited for his child being creative, but is instead excited for the veneer of high production value. Does that make her videos any better, her code any more functional? No, but it makes Mark feel good, so it's worth it.
Mark loves small businesses, and seems to believe that every single person should be a small business owner. That totally won't flood the market with similar looking slop and starve out everyone in the process. He fails to recognize that businesses have costs, and thus require capital to start. "Oh" he'd cry, "you just don't get it! AI reduces the costs of starting a business!" But really, seriously think about it for a second: what kind of business does it reduce the costs of? Likely ones that the internet is already flooded with: shitty guides, drop-shipped garbage, and AI-generated slop-media empires. There's no reason to believe that AI is going to magically enable everyone to run a super-efficient, valuable company that actually provides goods and services.
The super-tutor is another benefit he touts of AI, as though the internet isn't already flush with learning resources for every level of learner. I strongly suspect this will actually lead to worse learning outcomes because a.) the LLMs are prone to errors and misunderstandings, and those learning with them won't have the faculties to know that what they're being taught is wrong, and b.) we've developed a lot of different methods for learning because there's been a lot of research into how people learn, being spoon-fed exactly what you want to learn is likely to lead to gaps and poor understanding of what is being taught. There is something to be said for schools being underfunded and unable to meet the needs of children, but schools that have attempted to supplant individualized help with AI have had disastrous results.
His claim that everyone will finally benefit from science really frustrates me. It seems to be attempt at repackaging work that's long been done in the open science community as an achievement of the AI era. Things like Folding@Home have long allowed people to get involved in science if they'd like, or programs that have users help identify large batches of images through distributed work. The idea that AI enables people's "direct involvement" would require them to first trust science, which Meta has helped undermine along with current administration, as well as again have the means to afford the treatment if something was developed. Without the means, the treatment being there is useless. We have treatments for HIV/AIDS, but it's not widely available to affected communities because of cost. The other end of that is people trying to make their own replications of new drug treatments, which I don't think I need to explain the problem with. Science itself is already accessible, and access to the developments from it are aided by money, not AI.
Finally, they'll make sure everyone has access to their product, not because daily active users are valuable and can be advertised to, but because they care. Trust them™.
Part 3: The Real World Risks Which Just Happen To Mention How Meta Is Being Like, So Responsible
Note: I want to work through each risk that is deemed worth mentioning, which I am reminded is a solid 2/3 of the length of the original manifesto. This will be a long section.
Mark hems and haws about how there are he takes the risks with AI very seriously, but also have you considered that actually the problems would be fixed if everyone had AI? That seems like a stupid way to justify his choices, and it is. His "thought experiments" are nonsense, assuming AI is both superintelligent and unbiased, both of which it shows no inclination of being. The most confounding is his example of cybersecurity, as I'm pretty confident that everyone in cybersecurity is aware of and has some level of access to AI, but yet here we are with new incidents every day. Isn't this what is supposed to fix all of our cybersecurity problems? How many more do the engineers need to use AI? At what point are we going to reach our mythical security peak? His talk of AI lawyers and AI businesses are similar crap, again requiring that you suspend your disbelief for this future product that is both perfect and tailored to every individual, without replicating their anti-social behaviors or harmful biases. It is easy to say we'll figure it out, its another entirely to claim that it would fix everything.
I think he shows his cards when talking about people, something I'm not convinced Mark understands whatsoever (emphasis my own):
When people are empowered, they naturally compete and check each other economically, socially, politically, and in all other planes of human interaction. People also check and balance the power of institutions including businesses and governments.But if the power of superintelligence is held by a small number of individuals, businesses, governments, or AI itself, then that will naturally lead to outcomes that are less favorable for everyone else. This is not a technological principle. It is about the balance of power. There is no such thing as a singular benevolent superintelligence.
Therefore, the key to a positive future for everyone is achieving a balance of power that favors individuals. The solution is to ensure that superintelligence is broadly distributed to empower people.
Even if he won't call himself one, Mark writes like a social darwinist, which is both a pseudo-scientific ideology and not how things really work. People do not "naturally compete" as far as I'm concerned, which is something I'm equally able to claim as mark, given he has no background in any relevant field. I believe people tend towards cooperation, because we understand that we can accomplish more when working together, working towards a social good. An expectation of competition is what drives competitive behavior.
It's also very funny that he is so fixated on the individual, when also calling for AI to be a universal good. In his mind it is reasonable to say everyone needs access to AI, but doesn't require that access that same access to healthcare, housing, or food. This is the thinking of a broken individual, one who is wholly disconnected from the actualities of living; by itself this should be plenty to know that we are not dealing with a serious person.
I find it personally offensive that a company complicit in so much harm is also trying to brand itself as the guy working for the people, calling its mission (since it's "founding", although I'm pretty sure he's not talking about Facemash, or Facebook for that matter) "focused on putting power in people's hands." That's like Cargill saying it's been about sustainable agriculture since day one, or Monsanto (now owned by Bayer) that its for empowering farmers. It's just a lie; a convenient rewrite of history for the sake of current day myth-making.
Job Growth and The Economy: I find Mark's argument for why jobs won't be hurt short-sighted and ill-informed. He says there's an infinite demand for "new experiences and... new problems to tackle" but I see an economy that has already expanded well beyond what consumer dollars can support, which is why so much business is B2B or surrounding the business of investing in businesses. He fails to account for the idea that while company headcounts may shrink, support divisions would face an outsized impact, and beyond that B2B would shrivel as well, being replaced by bespoke, AI-made systems. There are multiple layers of employment that take a hit as these various workforces shrink. His only claim is that a bunch of jobs that we can't even imagine will take their place, like slop-toy maker, writers for some reason, and "personal biologists" making personalized treatments, which totally sounds like a cool and regular job people will have (and again, if many people don't have jobs, how are they paying for these things?).
A Long Tangent On How This Should Effect Richland Parish Specifically
Building AI Infrastructure with Communities: Yeah Mark, if you give towns something in return they tend to be more forgiving of developments, really groundbreaking stuff. What's interesting is what isn't said though; he's very careful about what he claims and what he doesn't acknowledge. First off, his claims for the local area in Richland Parish seem... questionable at best, which he frames as Meta being so generous and just wanting the town to flourish, vs. their contributions being legally required taxes and splashing some cash into vanity projects.
Richland Parish is a town of roughly 20,000 people, where the deal for Meta to build the data center was passed in secret with NDAs and now locals finding themselves priced out of the area. Additionally, through legislation and corporate loopholes Meta has flexed most of the risk of this development to the state and its people (please see the NYT piece linked just above for much more on that topic) With this in mind, reality starts to look quite different from the benevolent angel persona Mark tries to put forward in his manifesto.
I can find no detail on what this "Future Is For Everyone Fund" he mentions actually is. Would love to learn more about it, I see mentioned it being a $1B fund, but they don't so much as have a press release about it, so who knows? He touts the America's Workforce Academy as investing in the community, not acknowledging that a.) these construction buildouts stop one day, and b.) most of the individuals accepted into the academy will not be from Louisiana, let alone Richland Parish. This is not long-term investment in employment, and the surrounding area is not able to support the numbers of people working on the datacenter staying in the area. They'll have more than 7,500 construction workers for the build-out according to Meta, but only 1,000 jobs once complete, which will also require an entirely different skillset.
He talks a big game about keeping electricity cheap, without talking about what that new power will be facilitated by, which is of course not one, but seven (maybe ten? conflicting numbers) new natural gas plants, which will almost solely be for the power needed by the data center, according to their numbers (5GW datacenter max, 5200MW produced by the completed plants at peak). Also included is an unclear expansion of nuclear power plants (which is better in my mind), and the only talk of renewables being "commitment from Meta to help fund up to 2,500 megawatts of new renewable resources", which I'll believe when it's built. Beyond that, there are escape clauses written into the deals for Meta, which means if they exit the development, locals will be the ones to pay the price with higher energy costs among other things.
They're also supposed to be the most water efficient data centers in the world, committing to being "water-positive... by 2030." However, their projected usage is still 23M gallons of water per day in the meantime, just under the amount used by the town itself. They do insist it will be a closed loop, cycling the same water instead of constantly drawing, and I don't see immediate concerns of droughts/overdrawing from local resources, but according to an article in The Advocate, Christopher Dalbom, director of the Tulane Institute on Water Resources Law and Policy said “There’s no incentive for (Meta) to use less water because they don’t have to pay for it, and it’s not regulated or reported."
Curiously, Mark also brings up that "Historically, towns and cities that brought railroads, highways, electrification, and broadband became centers of research, business, and industry, with population and economic growth that follow." This is true, but not particularly relevant when our technology has developed so your can have single (maybe double) digit millisecond latency to the datacenter from across the country. Location is less important than ever when it comes to these developments, which is partially why I feel they tend to keep picking less populated areas (besides the tax benefits and other incentives).
Finally, we end with the gem of "American communities will have greater prosperity and security if America and its allies lead in AI compared to geopolitical rivals," which sounds insane to claim. Prosperity is doubtful, but they will continue to tout jobs, developments, etc. Security though? America just got its shit kicked in by a much smaller, less militarily powerful rival, how is AI going to fix that? Neither America nor Meta gives a fuck about its allies being ahead in AI, it's about them being king of shit mountain. It's rude to lie about what your priorities are, particularly hidden so deep in a piece about a supposed utopian future.
Now Back To The Regularly Scheduled Programming
Securing Against AI Misuse in Cybersecurity, Bioterrorism, and More: A lot of shoulds here, but literally nothing concrete. Just hand-waving and talking about how there's a real big risk. Remarkably meek compared to his response in the section above. It speaks to how little he has to offer in the way of stopping these things from happening; this is one of the real risks, and by acknowledging it he can at least say "but we're trying!"
Protecting Freedom and Preventing Government Tyranny: It's crazy comparing how Mark talks about how people should have a right to access personal AI versus their very wishy-washy handling of speech on their platforms. The self-described capitalist thinks that his product in particular should actually be a right afforded to all Americans. He also tacks onto here the idea that the government should be more responsible for making sure models are safe instead of them, a thing that I'm sure the government is going to get right on.
Ensuring American Leadership: This section is little more than a loud fart sound laced with xenophobia and fearmongering about China. A lot of unsubtle hinting that actually this could be a national security threat so the government really needs to partner with someone (Meta) to lead (Meta) in (Meta) Open (Meta) Source (Meta) models (Please pick Meta). You can practically hear the begging and groveling through the paragraphs.
Alignment With People and Addressing Existential Risk: There's some really unhealthy grandiose talk that tends to pop up in these conversations, hinting at how if unchecked AI could kill us all. If this technology is so dangerous then it simply should not be integrated into things that would pose such a risk (i.e. power grids, governments, etc.) Instead, we get to see Mark attempt to talk his way out of a room with no exit. There's no effective way to control for what he's talking about, even with what he describes, which is essentially each agent being tailored to an individual's sense of reality. The only winning move is not to play, which we have learned before and will learn again.
Maintaining Control of Superintelligence: Mark is straight up making shit up. We have no hint of any system being anywhere close to able to perform like what is being described, instead his ideas are based in science fiction. He keeps mentioning super-intelligence and recursive self-improvement, so I want to make it crystal clear: AI super-intelligence and recursive self improvement do not exist, and are not reasonably on the horizon with any known means of LLM training. If they were, we would have more than these freaks wandering about jamming it into every third sentence, there would instead be, you know, evidence.
Part 4: Ideology, Meta, Mark, And Lies We Tell Ourselves
This is the natural conclusion of all he's set up. The manifesto is not to explain why AI is important, or why it needs to be open source, but instead why Meta should be the state-sanctioned winner of the AI revolution. He tries to make the argument that Meta is the best choice for everyone, a good tool of the state, but also the state shouldn't get too involved, because we'd hate for pesky things like comprehensive legislation to get in the way of AI "innovation".
Unfortunately for Mark, he has forgotten that Trump values blondes with cosmetic surgery above all else, and most damningly I don't think Mark is willing to put out. So, this manifesto is not going to change the minds of anyone important. Instead it serves as a monument to the moment; a meandering walk through Mark's concerns and interests with no cohesive ideology or philosophy to be found other than Meta (and thus Zuckerberg) being the right choice for the government to throw their weight behind. It's a shallow view of the future with little interest in the details he deems beneath his purview, which happens to be most of life.
I think at the bottom of all this, Mark is scared. He's looking out at a tech landscape that might be slipping away from his grasp. His acquisitions have kept him involved with each wave of tech thus far, but AI could be something else; a money printer in the way that each past fad was supposed to be. Normally, a CEO might rebrand his entire company around a new focus, but he's done that a bit too recently to run that play again. So instead he must plea and grovel at the feet of the government to become the USA's preferred special AI boy. I hope for our sake he does not succeed, nor any of the other AI bigwigs. May this era be remembered as a weird blip of social contagion where Silicon Valley successfully gaslit the entire world.