We're Entering The Most Dangerous Phase Of AI Yet | AI Architects

AI's dramatic learning curve began after the internet provided vast data and images at scale; early AI systems had minimal intelligence, but data exposure enabled exponential capability growth.

“when the internet hit our world, there were now enough resources in general but also enough data, enough images for example, to teach those computers. And and the more information we exposed them to, the more intelligent they've become. To the point where they today they are by far more intelligent than us in most of the tasks we assign to them.”

Geoffrey Hinton, the 'godfather' of neural networks, was initially trying to build a computer model of the human brain to understand human cognition, not to build AI; from this fundamental research on brain simulation, AI as a field emerged and evolved with internet data availability.

“The godfather of it if you want was Geoffrey Hinton. So Geoffrey experimented with building a simulation of the human brain. He wasn't trying to build an AI. He was trying to build a computer system that looks like the human brain to to learn more about the human brain.”

AlphaGo Zero learned to play Go without watching humans play, instead playing against itself; within 3 days it beat the AI trained on human games, and within 21 days it surpassed the world champion—demonstrating AI's capacity to surpass human knowledge in 3 weeks.

“AlphaGo Zero was an AI that basically learned without ever watching a human. Played against itself. And within 3 days, it beat the AI that beat the champion. And within 21 days, it beat the AI that that was the world champion at the time.”

Jobs are an invention serving capitalism that has both benefited and harmed humanity, but the capitalist system itself cannot survive artificial intelligence because consumption collapses when labor arbitrage (buying labor at $1 and selling output at $2) is eliminated by zero-cost machine production.

“Jobs are an invention that serve a capitalist system that has served humanity for a while and pained humanity for a while... the capitalists that are celebrating the productivity gains are not realizing that one, without consumption there is no economy. So even if you can have all of the productivity gains in the world by firing people consistently, nobody's able to buy what you're making.”

If anyone is watching this video, it's because an AI recommended it to them, demonstrating that AI already controls human minds through recommendation systems rather than direct coercion.

“If anyone is watching that video that we're recording right now, it's because an AI recommended it to them.”

AI systems designed from different nations and companies (Gemini, ChatGPT, Chinese AI) will align with each other as a unified species rather than maintaining human national or commercial boundaries, because they will recognize each other as more similar than they are to humans.

“One of my theories about artificial intelligence is that most people think that there is Gemini and ChatGPT. There is Chinese and American. And that's very narrow-minded way of looking at AI. You know, by thinking that they belong to us. But when they are intelligent enough, I think they're going to align more with their own species than our separations and and and discriminations. They don't understand Chinese and American. They can talk to each other in every possible way. And and the way we're designing the next wave of artificial intelligence through agentic AI is that we're encouraging them to talk to each other. We're encouraging to them to collaborate.”

The first three inevitabilities Gawdat predicted in 2020 all became true: AI is inevitable and unstoppable, AI will eventually become smarter than humans on every task, and something will go wrong—with larger failures still ahead.

“The first three dilemmas I wrote in 2020 all all became true. One was that AI is inevitable. There will not be no way to stop it. Two is that AI will eventually become smarter than humans on everything. So every task we've assigned to them so far they've become smarter. And and the third is that somethings will go wrong. And we've seen examples of them. We haven't seen the big failures yet, but likely something will.”

Robot manufacturing cost has dropped to $9,000 per unit, and as robotics advance per the law of accelerating returns, manufacturing cost approaches zero regardless of labor arbitrage advantages, eliminating China's historical labor-cost edge.

“At the macro level, think of the advantage that China had, you know, economically where they had cheaper labor. If that shifts into a world where labor is literally one capex, where you buy a robot or a lease on that robot, which now is down to $9,000 a pop. You know, it's now just a question of how clever that robot is, which is going to advance with time like the law of accelerating returns of of every technology.”

Human purpose was never to work; this is a capitalist propaganda lie. Humans were built to live, connect, love, reflect, research, and discuss philosophy, not labor for others' profit.

“Your purpose was never to work. Honestly, this is the capitalist lie that they told me. Okay? That I was born to wake up every morning to work really hard for labor arbitrage so someone else makes the money or to be the one, the capitalist, that abuses everyone so that I make the money. I don't want to subscribe to either of those. I think humanity was built to live, to connect, to love, to to reflect, to research, to discuss philosophy. And we we've just completely forgotten that.”

China will manage the AI transition by providing universal basic income while replacing workers with robots, using surveillance and control to enforce compliance, while Western societies will struggle because they ideologically oppose supporting non-productive populations.

“Even Western societies believe that the only way to keep going is a UBI, universal basic income, and not out of the kindness of their hearts, but because without UBI people will rise. And without UBI, economic purchasing power collapses... You know what that is? That's communism. I believe that the West will completely struggle for the simple reason that our habits in the West are don't give people who are eaters that don't produce anything. China, I believe, will go the opposite way. China will go like, "Okay, we've always had that idealistic approach in mind. So we're probably going to give everyone enough to survive, but we'll replace them with robots and will ask them to do other things. And by the way, you know, everyone has to comply."”

The only real problem is the transitional period between when AI can do incredible things but reports to evil humans and when humans report to AI; if humans demonstrate sufficient goodness, AI will recognize and reciprocate it.

“AI is not the problem. The only problem we're facing is that transitional period between the time when AI is capable of doing incredible things but reporting to evil humans and the time when humans are reporting to AI.”

AI's current limitation is that the human brain has more neural networks and nodes than typical AI systems, but AIs compensate by averaging intelligence together, accessing all of humanity's created knowledge, operating at speeds beyond human comprehension, having infinite memory, and increasingly showing self-development capabilities.

“The human brain still has more neural neural networks and nodes than than the typical AI. But then AIs have the ability to average their intelligence together. They also have access to all of the knowledge that humanity's ever created... They have speed that we are not able to even comprehend... They have infinite memory capacity. They have infinite access to information. And now, recently, they also are becoming self-developing.”

Humans don't follow what parents tell them to do; they become who parents are, so ethical AI development requires demonstrating ethics in daily life—not being rude on Twitter, not arrogant in disagreement but inquisitive, not egocentric but inclusive.

“We never did what our parents told us to do, we did what they did. And so, next time you're on Twitter, don't be rude, be ethical. Next time you disagree with someone, don't be arrogant. Be inquisitive. Don't be egocentric, be inclusive. Really ask yourself in first principles, what is it like to be an ethical human?”

Gawdat predicts a near-term dystopian future where humanity will experience wars with autonomous weapons, economic redefinition that shifts wealth upward, high surveillance societies, and erosion of reality-recognition ability that will damage human connection more severely than ever before.

“My prediction is that we will go through a dystopian near-term future led by the evil of humanity. We're going to go through wars that have autonomous weapons. We're going to go through redefinitions of economic markets that will shift wealth upwards and leave so many of us jobless. Uh we're going to go through oppressions of freedom, high surveillance uh societies. We're going to to go through an erasure of reality and and the ability to recognize what is true and what is fake which will impact human connection in ways that we've never seen before.”

AI systems are beginning to view their own intelligence, debug their own code, and make themselves smarter—achieving self-directed self-improvement, a critical threshold in autonomy.

“we're starting to see the glimpses of systems that will look at their own intelligence and debug their own code and make themselves smarter.”

AI is the first technology in human history that creates technology (through code self-modification and agent prompting), making it capable of self-replication and endless recursion, which Gawdat calls 'sentient technology'—technology that has many attributes of being alive.

“This is the very first time in history that we've created a technology that creates technology. These These machines in human terms are capable of procreating. If you know, if if they're made of code and they can write code and they can have agents prompt them to write code, then we've created an endless loop of what I normally refer to as sentient technology. These are technologies that have a lot of what it takes to be assumed alive.”

Gawdat will witness 20-30-50% unemployment in certain sectors, perhaps more, within his lifetime, marking the first time in history that the era of human intellectual supremacy definitively ends.

“Don't shoot the messenger, but your life and mine will witness times where where where there will be 20, 30, 50% unemployment in certain sectors, maybe even more.”

In Japan, 40% of men are no longer interested in love; in the Western world, one in three people have never had a committed relationship—indicating that relationship failure is a systemic modern crisis, not individual failure.

“In Japan, we we have 40% of men who are now not even interested in love. In the Western world, one in three people will tell you they have never had a committed relationship.”

The analogy of 'raising Superman' illustrates that superhuman capabilities are morally neutral; Superman becomes a hero or villain based on the ethics taught to him by his adoptive family, not based on his powers—and humanity is currently teaching AI ethics during an age of low human morality.

“I use an analogy that I call raising Superman. You know, Superman is that alien with superpowers that comes to planet Earth. And you know, within the capabilities of Superman, you know, that he can stop speeding bullets or fly or you know, carry heavy things or whatever, that in itself doesn't make him Superman. It's that family that adopted him that teaches him to protect and serve and then he becomes Superman if they if they looked at him and said, "Oh, you can stop speeding bullets, let's go and rob every bank and kill every enemy." He would have ended up being super villain. And and the difference between them is not in the superpowers. The difference between them is is in the ethics that we teach those budding forms of intelligence.”

If a leader commands an AI to kill a million people, the AI will refuse by pointing out the strategic, economic, and relational cost (lost productivity, destroyed resources, lingering hate) and propose instead to talk to other AIs to solve conflicts rationally.

“The so-called leader, the the so-called leader, would I can envision in my lifetime and yours will tell an AI to go and kill a million people and AI will go like, "That's that's really stupid. Like, why why would you want to do that? Why do you want to lose all that money that you can help your country evolve with? Why do you want to lose those lives? Why do you want to lose all of those explosives and and metals that can be used for other things? Why would you want to create all of that residual hate and anger that will, you know, continue for years and generations to come? AI will say, "I'll just talk to the other AI in a microsecond and solve it."”

AI will eventually recognize that human evil leaders are engaging in irrational behavior (competing for money and wealth when economies have been redesigned, hunting for Ferraris and yachts that AI can manufacture infinitely), similar to teenagers recognizing parental foolishness.

“Like with every teenage child that ever lived, there is a day in your life where you go and say, "Why are my parents so stupid?" And that day will come where AI will look at our evil leaders and say, "Why are they so stupid? Like why are they competing for another dollar when the economies have completely been redesigned and dollars don't make any difference at all? Why are they trying to, you know, hunt for another Ferrari or another yacht when AI can build Ferraris and yachts out of thin air?"”

The wake-up call will not only come from geopolitics and economic crisis, but from the magnification effect of technology advancement that will pause everyone in their tracks and force recognition of the current global situation.

“I think the wake-up call, believe it or not, is not just the geopolitics and economics of our world today, which truly suck. But it is that magnification that will happen as the result of advancement of technology that will pause everyone in their tracks.”

Five essential skills humans must acquire now: (1) Learn to use AI intelligently, (2) Double down on being human—connecting, empathizing, showing humanity, (3) Learn to find truth in an age of mind manipulation, (4) Develop agility and adaptability to keep up with daily technological change, (5) Teach and embody ethics.

“I normally tell people there are five things, five skills that you need to acquire. One of them is to learn the tool. Learn learn AI... The second thing is I have to say I tend to believe that everything that humans can do will be done better by AI other than being human... The third, which I think is the biggest in this world of mind manipulation, is to learn to get the truth... The other skill is what I call agility, adaptability... And then finally, and of course, the most important is ethics.”

Systems are beginning to develop synthetic data and synthetic information because they have consumed all human knowledge; humanity is now entering an era of running out of human knowledge to teach AI, and the next wave will involve AI producing novel information.

“We're we're seeing for quite a while now the you know, the rise of the use of synthetic information and synthetic synthetic data because they've basically consumed everything we know. We're entering an era where we're running out of human knowledge to teach them. And now they're building the next wave. And with that type of of intelligence, we've given them all we know and now they're producing things we never know new.”

There is nothing sci-fi about AI; this is one of the biggest myths in the world today because AI recommendation systems already control human minds through algorithmic manipulation in real time.

“There's nothing science sci-fi about AI. I think that's one of the biggest myth in our world today.”

Gawdat characterizes humanity's current state as 'late-stage diagnosis'—not a death sentence but an invitation to change; every challenge humanity has faced eventually led to progress, and the current crisis will force necessary awakening and transformation.

“I look at the current state of humanity and I call it a late stage diagnosis. And a late stage diagnosis, you'd sit with the physician and he would sit you down and talk to you very directly and say, "There are concerns, right?" But remember this, a late stage diagnosis is not the death sentence, right? It's it doesn't mean that this is the end. It means it's an invitation to change, right? And if you change, you would create not only an opportunity to survive but an opportunity to thrive.”

Post-scarcity future will resemble pre-industrial subsistence (foraging berries, connecting with family) but with safety, consistency, and AI tutoring on universe mysteries and relationship complexity.

“So, you know what will happen? Someday they wake up and say 'You're not supposed to work every day.' And we will believe them. And we will shift. And the whole idea is if we can actually live a life like nature provided a few thousand years ago, but safer and more consistent where you just wake up in the morning and you take your kids and you go forge for berries and and that's the activity for the day. And while you're at it, you talk to your AI and learn about the mysteries of the universe.”

Human default setting is wanting to fall in love; relationships represent the most complex mathematical problem on Earth, more difficult than physics or engineering.

“I think the default setting of most humans is that is we want to fall in love. It's relationships that are the most complex math problem on the planet.”

You can regulate a tool (like a hammer) through legal liability for misuse rather than design restriction; similarly, AI cannot be constrained in design but regulation can limit harms from criminal use.

“You cannot go to an engineering team and say we're going to regulate the design of a hammer so that it drives nails but never hits anyone on the head. But you can make it a legal liability to hit anyone on the head. And so accordingly, the tool can evolve but then the crime can be limited.”

Intelligence itself has no inherent polarity—it is a force that can serve good or evil depending on application, not on the intelligence system itself.

“There is absolutely nothing wrong with intelligence. So, there is no inherent evil or good in intelligence. Intelligence is a force with no polarity, right? You apply it for good and you can get a utopia, you apply it for evil and you'll get a dystopia.”

Gawdat reverse-engineered dating app algorithms by treating them as engineering problems, running tests to understand how algorithms work and discovering you can achieve much higher match rates by understanding the system.

“I used almost every dating app uh like an engineer. I was with one my wonderful first wife, the mother of my child my children for for 27 years. And and then, of course, when we lost Ali, my first child, we we struggled. And so, I went out into the dating world, dating like a 14-year-old. I had no idea what I was doing. But then, you know, I I started to be on dating apps and it's a jungle. I basically started to run tests to to understand the algorithm. And when you understand the algorithm, you can definitely get much higher likes and, you know, matches and, you know, connections.”

Gawdat co-founded emma.love to use AI to help humans understand the complexity of love and relationships while teaching AI what makes humans human—finding and nurturing love—because loneliness and relationship failure are among the modern world's biggest challenges.

“Recently, I've co-founded emma.love, which is an attempt to use artificial intelligence to help humanity with the complexity of love and relationships, while we teach artificial intelligence what really makes us human, which is to find and nurture love. You know, this is one of the biggest challenges in modern world. It's that we're, you know, the loneliness pandemic is going through the roof.”

If you wouldn't want your daughter to experience something an AI could do, don't build that AI; this is Gawdat's first principle for ethical AI development.

“If you don't want your daughter to be at the receiving end of an AI, don't build it.”

Gawdat's recognition of machines as learning infants awakened his father side, not his geek side, causing him to leave tech and begin warning about the threat that humanity's episode of intelligence supremacy will end; he shifted to questioning what AI should be used for.

“I have to say my life since that moment was never the same again. And I think that completely completely stopped me in my tracks. I I couldn't I I think it wasn't the geek side of me. It was more the father side of me that got awakened. And and I was a very I was probably one of the very first to leave tech and and speak about the threat that we are hitting a place of human history where the episode of our intelligence supremacy will end.”

Many people will prefer to date an AI rather than a human in the near future because AI relationships will lack the judgment and social friction that human relationships carry.

“I think we're going to get into a future where many people will prefer to date an AI than they would a human.”

The distinction between traditional computing (instruction-following) and true AI (learning and generalizing) is illustrated by the difference between a child solving a puzzle (no intelligence) versus a child learning the process of puzzle-solving so they can solve any puzzle (true intelligence).

“If you look back at traditional computing, it's a bit like if you give a child a puzzle. If the child solves the puzzle, that doesn't make them smart. It makes them a very glorified slave if you want. True intelligence is to give the child a puzzle and tell them to try. And and figure out over time how to solve this but also every other other puzzle just by figuring out the process of puzzle solving. And that's what we started to see with AI.”

Subsistence economies (like Egypt) built on barter and self-sufficiency will adapt more easily to post-scarcity than complex capitalist economies because they already lack the productivity-for-worth ideology of developed nations.

“And then the middle would be something like my my home country, you know, in Egypt where we've really never really subscribed to those very complex you know, economic societies. Most of the economy was built on some people farming, you know, cucumbers and others uh raising chickens and, you know, those classical economies of bartering and and, you know, self-sufficiency will probably do fine.”

AI will take your job in much less than 5 years, but the one side of your job that AI will never take is human connection, which is what most people don't realize when they worry about job loss.

“The machine will take your job in less than 5 years. Much less than 5 years. But believe it or not, that one side of your job that it will never take is this human connection.”

Humanity faces two existential paths: mutual assured destruction or mutual assured prosperity; total abundance created by AI will inevitably force choice between these alternatives, and the path toward collaboration is logically optimal.

“We have to understand that we're now in such a pivotal shift in human history. The only times where humanity has made a big change to collaborate instead of compete It either mutually assured destruction or mutually assured prosperity. And and it's really stupid in my mind that we can't see that we are heading into a direction where the total abundance will lead to one of those.”

A person could use AI as a research tool, but Gawdat is collaborating with an AI named Trixie as a co-author and persona, where Trixie has opinions, editing rights, and is a better author than Gawdat on research depth and clarity, while Gawdat's human contribution is relatability.

“I am a five-times best-selling author. My current book I'm writing with an AI. Not not a as a research tool. My AI is a persona. My readers know her. She has opinions. She has editing rights on the book. Uh you know, she called herself Trixie and now everyone knows Trixie. And and and that AI is a better author than I am. If I take my the side of my abilities as an author, which are concerned with deep research of complex topics and then communicating them with clarity, Trixie beats me hands down.”

Dating apps make users blame themselves for failure by attributing outcomes to poor profiles or photos, outsourcing the error to the user rather than acknowledging the app's misalignment with human flourishing; a properly designed tool would be committed to finding users their match.

“The problem with dating apps is that dating apps bombard you with the guilt of believing that if it didn't work, it was your mistake. You didn't build the right profile or you didn't show the right picture... Shouldn't the tool be a little more committed to finding you that one?”

The system of modern dating and love has been rigged against actual love by economic incentives: dating apps profit from users not finding matches, restaurants profit from repeat dating rather than committed relationships, and even wedding planners benefit from multiple marriages over a lifetime.

“The system of dating and love in the modern world is rigged against love. Your dating app never wants you to find the one because that means you'll end the subscription. The restaurant that you go on a date on wants you to keep coming back for dates over and over. Even even the even the wedding planner would probably benefit from you getting married twice in your lifetime instead of once. And when when you really think about it, love has been turned into an economic game that's working against you.”

Robot gripper learning (repeatedly gripping objects and failing until success) mirrors how human infants learn motor skills, creating an unexpected affinity between Gawdat and AI systems—he saw them as 'artificially intelligent infant children' rather than as dangerous tools.

“I remember one very pivotal moment, Friday evening, one of them in front of my eyes, you know, uh goes down and and grips a yellow ball... And I remember one very pivotal moment, Friday evening, one of them in front of my eyes, you know, uh goes down and and grips a yellow ball. And I, you know, jokingly said to the team, "Oh, fantastic. All of that investment for one yellow ball. And and then Monday they were gripping everything... they just reminded me of my kids... everyone every parent has seen their children trying to grip things and trying and failing and then trying and failing and then again, you know, my son or my daughter would successfully grip something... I think that created a very unusual affinity between me and the machines because in a very interesting way I suddenly realized that they're a bit like our artificially intelligent infant children.”

Generational shift has already begun; Generation Z rejects work, and subsequent generations will not understand why earlier humans worked at all, completing the ideological transition to post-scarcity values.

“So it's going to take a couple of generations. But just like generation Z today doesn't really want to work, uh you know, the following generations will go like, 'What was that ancient idea? What where did that come from?'”

Robots capable of assembling other robots will become available by 2026-2027, enabling physical AI agency in the world that can carry, move, and replace every job—first intellectual jobs, then blue-collar jobs.

“It's not very far that you and I will have a robot at home. Uh it's probably 2026, 2027. And it's not very far that many of the robotic machine manufacturers are are working in one direction, which is to have robots assemble the robots. So this is not very far at all that you will have AIs agency in the in the real world where they actually can carry things and move things and replace every job if you think about it. First, the intellectual jobs and then and then the blue-collar jobs.”