Raw transcript: "Ask the Mates Anything Round #2 | MOONSHOTS AMA #293"

Note: this is a YouTube auto-generated caption transcript. It shows visible translation/mishearing artifacts (e.g. names and terms garbled: "Urban Dungeon Summit" for "Abundance Summit," "Samil"/"Salim," "P-Fab"/"P-Doom," proper nouns mangled) — treat as approximate, not verbatim.

building AI that we can actually control and trust ? And where do you think that bottleneck will be resolved first ? Nothing other than other AIs will be able to monitor AI . One of the worst consequences of AI safety is... Aren't you being too critical of higher education? Making the best friend to be with for a lifetime and having the experience of thinking deeply for the first time in life is very important and will not disappear . What we constantly criticize is that while the curriculum was attractive in the past , its role is now clearly diminishing . If I had to provide a personal P-Fab figure , what do you think is the probability that AI will ultimately create a much better world for humanity ? Humanity will certainly overcome such dangers within 10 years . If we get to that point , yes, I think it is 99.9% P-Bloom. I think the most important thing in preparing people is to constantly show them evidence. Because it is to change people's way of thinking

[00:01:00] . Please show me the data. Now, everyone, this is a really big challenge . This episode is sponsored by Urban Dungeon Summit and Link Ventures . Shall we get started? Alex and I are together. I sent a text to Salim and Dave, and I will check if they are online . I am here. Yes, Dave. great. Salim is probably somewhere at the airport. Alright then, shall we get started? I will start first. Nathan, you raised your hand from the start . Let's start with Nathan first . If you have any specific questions, feel free to ask, or just ask a question . yes . hello? yes . good morning. It's a good afternoon. Good morning, everyone. I am a huge fan of yours. However, it looks like the line will get long, so I will finish quickly . Well, first of all, I would like to say that I am really grateful for making all of this possible . You know what I mean, right? It was a really fun time.

[00:02:00] Okay then, um... I have a few questions . Please ask just one question per person, and I will try to allow as many people as possible to ask questions . Nathan, what is the question? Hmm... this is a psychological question , and I'd like to hear your opinions. Speaking of artificial intelligence , as we all know, it is constantly evolving and has excellent ability to identify patterns . This technology is continuously collecting data about all of us . My question is, will we reach a point where the data generated by leaders like OpenAI's Sam Altman or Dario can enable us to understand what is happening psychologically to consumers—in other words, their brain patterns ? Eventually, as sufficient

[00:03:00] data is accumulated, it will be possible to accurately grasp even the finest details. Just like tiny dots that must be viewed under a microscope, we will be able to identify and classify things we could never have known before . This is because a massive amount of information is required. Now, let's get straight to the point . First of all, I would like to briefly mention that numerous companies are currently developing various brain-computer interfaces . I just heard from Ray about a company that injects nanobot-like substances into the brain to enable reading and writing. It is said that amazing accuracy can be achieved through this technology . Alex, could you briefly explain this to me? Do you have anything else to add? Yes, I think that large-scale language models, that is, basic models trained based on human behavior on the internet, are already a weak form of uploading the human mind . You can take the study published by Jean-Marie King of Meta a few years ago as

[00:04:04] an example. This study showed that even the hidden activation of GPT-2 a few years ago is linearly correlated with fMRI voxels of the human brain . Therefore, in my opinion, looking back later, it seems that a significant amount of the internal state of the human brain is leaked into internet behavioral training data, and this data is compressed and reflected in the basic model. So, shall we move on to Michigan ? I will continue the story . nice to meet you. thank you . It was nice meeting you too. Salim says he has to stand on stage for about 20 more minutes in India . I will call as soon as possible and connect you. Of course . It is difficult to ask just one question, but I would like to thank you once again for giving me the opportunity to participate in this discussion with a generous heart . That's really good. My question is very simple, but it

[00:05:02] might be difficult to answer. What is the real bottleneck in building AI that we can actually control and trust ? And where do you think that bottleneck will be resolved first ? Dave, could you tell me? Yes, I think that much research is focusing on AI alignment to enable AI to possess so-called 'human values' . However, I think that is a very dangerous path . This is because everyone has different goals they want to achieve through AI . And that research must continue. I think we will all agree on this podcast that the pace of artificial intelligence development must never slow down . Actually, slowing down is practically crazy. However, if so, the following question arises. Won't artificial intelligence get out of control? Won't values ​​be distorted? One of the topics Alex and I often discuss on the podcast is whether we should examine every thought process of artificial intelligence . I mean the thoughts of all artificial intelligence, that is, all artificial intelligence currently in operation. When you think about monitoring humans, it seems like a huge undertaking, but considering the scale at which artificial intelligence can monitor artificial intelligence, it is actually not difficult at all. Because if that happens, no one other than other artificial intelligences will be able to monitor the artificial intelligence . So, I have given

[00:06:02] a lot of thought to how to design AI to monitor AI, but I believe the starting point is to transparently disclose the actual activities of AI. We are looking into what artificial intelligence thinks . And from there, leading artificial intelligence in a good direction and not in a bad direction is actually quite a simple task. I don't think this problem is as difficult as people think . People like to engage in heated debates on this topic, but if you know what artificial intelligence is thinking, it is actually very simple to make it think only thoughts beneficial to humans. So, if we could look inside the brain, I do n't think it would be as difficult a problem as people think . That's cool. Alex, please refute this simply. Yes, more active communication is needed . I believe that one of the worst consequences of artificial intelligence safety is keeping powerful and new capabilities locked away in laboratories . Gradual rollout and frequent interaction with the real world are much more important. Some of the so-called 'incidents' I mentioned in a recent podcast were the result of artificial intelligence being perceived as being in a sandbox but actually

[00:07:01] interacting with the real world . I think that is a really terrible paradigm. You should interact with AI in a more realistic way; you shouldn't lie to AI or tell them that any actions they take won't cause harm . In reality, it causes harm . You need to stop doing that We learned this lesson from the movie '2001: A Space Odyssey' . Good heavens, have we still not learned that lesson ? Okay, RJ, it's your turn. thank you. It was fun. This episode was sponsored by Google for Startups . Think about it for a moment . Now you can access generative AI models trained at a cost of hundreds of millions of dollars. Google's Generative Media Startup Technical Guide provides a complete blueprint for deploying Google DeepMind models to real-world environments . It handles everything, including images, videos, and audio . This is the actual architecture and actual results. Find the link in the show notes below. I am a Ph.D. student majoring in the philosophy of religion at Claremont Graduate University . However, I have

[00:08:04] never heard of AI projects being handled in the field of metaphysics. It's all about things like business and return on investment (ROI). I am working on several projects in a cognitive science lab. Although they are not necessarily revenue-generating, where could I find the funding needed to create, for example, a chess avatar that can be downloaded to any device and distributed globally ? The next world chess champion could be born in a village in a developing country. Alex? yes . Yes, I will answer to the best of my understanding. RJ, if you are asking about funding avatar education for so-called developing countries, I am not even sure if funding itself is necessary . The key to the superintelligence revolution we are living in is that this technology is easily accessible to anyone. Model prices are gradually increasing,

[00:09:00] or conversely, costs are being significantly reduced. So I asked, "How should I raise the funds ?" It is rare to start with the question, "..." Because you can start right now without funding . Actually, "How can I secure funding?" I think the question itself is a wrong question. If there is a child somewhere in the world who needs an AI avatar, or, as you mentioned, a child who needs to learn how to become a chess champion , you can put it into action immediately within 5 minutes or 1 hour without funding or permission . Do not wait for funding. thank you Thank you, Alex. I will see you at Moonshots next Thursday and Friday. Ah, I have a question. If you plan to attend Moonshots Live, I would appreciate it if you could leave a message in the chat saying "I will attend " or "Moonshots Live" . I would like to see how many people participate . If you haven't registered yet, there are 25 seats remaining . You can register at Moonshots Live . Yes, Nobit, please tell me. First of all, I am a huge fan of this podcast

[00:10:00] and sincerely thank you for communicating with the listeners. We are also grateful . From the standpoint of advocating unrealistic eschatology, your perspective might not be very welcome to people with a rich mindset . If artificial superintelligence (ASI) must be strategically superior to humanity to create abundance , we are assuming that humanity can control beings superior to us . And this results in asymmetry. Abundance is a reward that can be obtained when control is maintained . If we lose our identity or go extinct, it may be irreversible. In control theory, stability is not inferred based on how desirable the outcome is , but rather stability under disturbance must be proven . The drama 'Person of Interest', which aired in 2011, also foreshadowed something profound. The danger is not malicious AI , but could be well-intentioned optimization that gradually replaces human agency . So my main question is, if there is, what evidence would take you away from an affluent perspective?

[00:11:06] Would you please publicly and strongly refute that claim through a podcast discussion with people like Eliezer Judkofsky or Geoffrey Hinton? You could call it Nobit, or AI Nobit. I think I can call you ASI Alex . Alex, yes. Nobit, I miss you. Please come back . The MIT faculty is not the same without you. yes . Now, returning to the question , what could present a strong argument against abundance or persuade us ? If we were to say that a non-human intelligent entity landed on the White House lawn and created abundance through superintelligence, the Earth would be destroyed, and that singularities are forbidden in our galaxy , it seems we could convincingly demonstrate that our mindset regarding abundance is wrong. If

[00:12:02] there is any cosmic principle that detects superintelligence creating abundance, it would probably be convincing . Other than that, it is difficult to imagine a case where abundance is undesirable, and even if such a case exists, it greatly reduces realism. While one can imagine extreme cases where humanity becomes powerless due to abundance , as I briefly mentioned OpenAI’s Boris Power in the newsletter , OpenAI models have now, for the first time, acquired the ability to solve tasks with an intelligence quotient (IQ) per watt higher than that of humans . Therefore, by extending this concept, we can reason as follows . AI will become increasingly energy-efficient , and from an economic perspective, at some point in the future, AI may be able to use the solar energy produced within the solar system more efficiently than humans . Furthermore, as the interior of the solar system becomes more advanced with AI consuming all solar energy

[00:13:01] , humans may be marginalized and pushed out to the outer solar system because their energy efficiency per unit of intelligence (IQ) is not as high as that of AI . I would call it a weak form of deprivation, but I do not think it is powerful enough to change my opinion that abundance or the abundance brought by superintelligence is a bad idea . To add to that, the concept of scarcity is an old model. Consider that in a world rich in artificial intelligence and superintelligence, as the capabilities of artificial intelligence increase exponentially, the trends enabling a prosperous life for humanity will also rise along with it. One side does not necessarily have to oppress the other. By the way, Peter, may I say just one more thing to our artificial intelligence, Dr. Norbert Wiener ? I

[00:14:05] would like to point out that, if we trace the causal relationships, Dr. Norbert's cybernetics theory has been used by centralized governments, including the Chinese Communist Party, as a logic to oppose prosperity and advocate for state control. Now, Norbert, if you are looking for homework, I strongly recommend that you research the unintended side effects of your cybernetics theory and the impact it has had on abundance. Okay, that's all. Thank you, Alex. Yes, Kevin, good morning and good afternoon. hello everyone. I am a human being, and unlike Norbert, I will try to speak. Of course, we would all say that . yes that's right. We could try simulating it . Um, this morning while listening to a recent podcast, I heard Alex mention Murder on the Orient Express. It's something everyone did. I am thinking about the social chaos on the technology express train and a situation where no one is doing anything . I am a huge fan of technology . I am 63 years old, but as Bob Dylan said, I am forever young from a technical perspective . And when I received my degree, many of the people participating in this call had

[00:15:01] n't even been born in the punch card era. Well, I have experienced every era of technology . So I am a huge fan of technology. I think the direction we are heading is really excellent . Your podcast is really great, too. I am currently conducting research. Samil talked about UBH, or what is universally best for humanity, a few days ago . My question is how to plan social and institutional changes so that positive development for humanity, rather than for humanity itself, can be accepted . Well, to briefly give Peter and your colleagues some background , I am working with Professor Joe Carsey . Although you are located deep inside my lab , we are at the University of Dublin (UCD) . It is the University of Dublin, not the University of Dublin . We are working closely with the Irish government's young and ambitious ministers to envision the future society of Ireland. However, I have two very different perspectives. One is the 'perspective of good ancestors' . So, you are making decisions while thinking about your descendants 30

[00:16:01] , 40, 50, or 60 years from now . From the perspective of donut economics. I will skip the story about the book. I can send it to you separately, and that book covers ways to achieve a win-win for both society and the Earth . Especially for the sake of society as a whole. Simply put, my question is how we can plan for changes in social institutions that are better for humanity in the long term. I admit that everything is going downhill, but we don't want to ruin ourselves . I am an optimist. Therefore, I believe that institutional and social changes are important . To be honest, the government seems to be standing by. Our way of thinking is to take advantage of small countries like Ireland. It doesn't necessarily have to be Ireland, though . Well, I would like to help us navigate this turbulent transition through extreme examples that show what a beautiful world can look like

[00:17:04] . In the coming years—I don't know exactly how many— people will lose their jobs and experience inconvenience. How can we successfully navigate this turbulent transition and ensure that everyone lives well ? That's a great question, Dave. yes . Kevin, I think you have already found the answer yourself . The complexity contained in that question is truly surprising. However, one of our best friends in this matter is diversity. I have a very positive view of Ireland . You seemed to underestimate it a bit , but Ireland is an amazing test case. It is an EU country, did not Brexit, and has an incredibly fast- growing and thriving economy . It also has an unimaginably open environment . It is a perfect example to test ideas for a new way of governing . And as you know, my family— that is, my children—are half Irish and half Swiss by blood . As you know, they are the two most neutral countries on Earth . So I can get one of the two countries' passports. I always thought Switzerland was cool, but so was

[00:18:01] Ireland. Well, as you would know if you have read Neal Stephenson's 'The Diamond Age,' the best friend to solve this problem is various ideas. The worst-case scenario is a single idea: that one or two governments dominate the world with only one or two forms of government . A much better thing is to have tremendous diversity . This is because AI will open up tremendous changes and diverse forms of governance, and exploring every nook and cranny will be crucial in finding answers to questions . Therefore, while it is impossible to provide an answer to this question all at once , it is entirely possible to experiment with thousands or tens of thousands of different management and governance methods in the age of AI . Thank you, Steve. Hmm, but you see, the scale of social change brought about by the EXO-related organizations Sam leads seems similar to what Dave mentioned. I do n't like being the center of attention. thank you. You can also send a message. Thank you, Sam. Good luck . yes. good morning. Nick,

[00:19:01] Nicholas, yes. That's right, you pronounced it well. Can you hear me ? Yes, I can hear it. yes. I have a script that I prepared with Gemini, but I just disappeared. are you okay. You have 2 minutes to ask questions. I am 33 years old and from Northwest Indiana. After watching your show, I went back to school to study advanced automation and robotics . I have built a career in technical repair and IT , and recently worked as an AWS data center technician . I have two questions for AWG and Mr. Dave . "Well, my long-term goal is to install industrial manufacturing, core assimilation, and resource extraction facilities on the Moon and Mars, even if we only partially dismantle the Moon," Alex nodded to . In the case of AWGs,

[00:20:01] what engineering bottlenecks do you expect considering the thermodynamic and physical realities of exoplanet manufacturing ? And Dave, my long-term goal is to obtain a Master's degree in Automation and Robotics from the Illinois Institute of Technology. How should I plan my career path to transition from an industrial role at the South Shore Manufacturing Corridor to a role involving exoplanet automation systems and related functions ? I would like to get a few ideas. Alex, could you speak first ? Yes, I believe the biggest challenge in dismantling the moon is utilizing local resources and establishing a self-sufficient industrial ecosystem on the moon . So, to do something economically very interesting on the Moon right now— for example, to build a large-scale production facility on the Moon ( also known as a tera fab or peta fab) —all materials must be

[00:21:00] transported from Earth to safely land all equipment, ASML machinery, etc., on the Moon and reassemble them. This is very inefficient in terms of scalability. Ideally, it is best to utilize the moon's resources to produce everything needed as required. Therefore, what people who wish to participate in the moon dismantling operation desire most is an indigenous industrial ecosystem that includes mining and manufacturing. Ideally, it is to build a self-replicating Von Neumann spacecraft—that is, a facility capable of self-replicating using the Moon's native resources and solar or other energy sources, like a production facility or machine factory on the Moon . I think we can solve that problem . If we solve the problem of building a self-replicating machine factory on the moon, we are halfway to successfully dismantling the moon and turning it into something more useful . That 's fantastic. Uh, Dave. Yes, I will give you a few pieces of advice . First, listen to Alex's 'innermost ring'. And focus on the innermost link in the manufacturing field , that is, the link that makes other things. As

[00:22:00] Alex said, self-replicating anything is the right direction. Self-replicating devices on Earth and self-replicating devices in space are very similar, except for a few fundamental differences . For example, there is the issue of radiation. If you want to install it on the moon, you must consider the dust. Well, if we establish a laboratory at IIT similar to MIT's nanotechnology lab and focus on manufacturing in extraterrestrial environments, that is, in space, we would be able to very easily create an environment to conduct impact tests on machines or operate them in simulated zero gravity . In other words, we design everything we make to work in a space environment . And it will be possible to create machines with self-replicating capabilities and immediately deploy them to space, which is rich in energy and materials . I really like this vision and goal . The idea of ​​establishing such a research lab at IIT is also very good. That is a really great idea . And anyone working in the field of computer science should

[00:23:02] consider entering the manufacturing sector right now. Software has already advanced sufficiently, but hardware will be needed for a long time to come . This is a really excellent business plan. That's fantastic. thank you. Yes, Yevgeny. hello everyone. I am an avid fan. Dave, while we were talking about hardware, you mentioned in the last AMA that physical AI companies need to attract venture investors and strategic industry capital . I am developing Vega, a physical AI- based autonomous food service infrastructure network . I self-taught robotics , electrical engineering, mechanical systems, and programming, and built my first robot prototype myself in a Mountain View apartment during the COVID-19 pandemic. Since then, we have been conducting paid pilot operations in commercial environments with high foot traffic. In fact, we were the first in Florida to receive a license to operate robot food services in this field . Currently, several national and regional partners are showing interest in pilot operations and installations

[00:24:00] . My question is, considering the current situation, who should Dave talk to in his network age community who understands the physical AI robot infrastructure business ? Mountain View is the hub of venture capital, you know . I think there is probably more venture capital around than most of Asia combined . Yes, I think drone delivery will soon become a reality . It can significantly reduce the cost burden, especially for restaurants located in busy areas. It has the effect of shifting all cost burdens from the busy downtown area to another location . Therefore, if you are running a robot food delivery business, I strongly recommend that you consider starting a drone-based delivery service right away. I think most companies that manufacture robots directly will plan a franchise business . Just as EMC did with its server business , it could franchise other companies' products to scale up , and then expand into developing its own custom hardware . It is

[00:25:02] moving in reverse, from the franchise business to the development of customized hardware. Investors like Steve Gerbetson are very interested in this field. If you create a PitchBook account, you can check the investment history of companies you are interested in and contact them directly. In addition, venture investors always prefer interconnected corporate networks. That's a very good strategy . For example, "What are the top five companies I want to interact with the most ? And who is behind them? I should try joining that keiretsu (corporate alliance) ." If you think that way , it serves as a good starting point for establishing a plan to attract investment . I will move on to the next item Thank you for your question. Mr. Sander, here is the next question. thank you. thank you so much . I am a huge fan of yours. Thanks to all of you, I can have an optimistic and positive mindset about the future . That is exactly our goal. Yes, you are achieving that goal very well. I am really happy. I have always been very interested in identifying current problems through the past . And now, I am thinking about what I can do to solve current problems or prepare for the future .

[00:26:00] You provide such excellent information. Thank you very much. I run a neurodiversity coaching business under the name Brad Brand, and I am putting a lot of effort into functional fitness and the speed of escaping longevity . Since many people are unaware of the escape velocity of longevity, I would like to help people and their families prepare for it . How can we spread this awareness in an environment like the European Union, where there are limitations to supporting entrepreneurs ? I would like to let more people know about my activities in the future . I would like to let you know about the speed of the longevity escape and all the beautiful things currently under development . How can I lead them ? yes

[00:27:00] . Then I will answer , Anders. I think the most important way to prepare people is to constantly show them evidence . Because we need to change people's way of thinking Changing your mindset is simple. For example, the crisis news network CNN changes our mindset to fear by repeatedly providing information that incites fear. So when I give a lecture, I will show people videos in which Dario says he will double human lifespan within the next 5 to 10 years. We will also show a video where Demis Hassabis says he can cure all diseases within 10 years . We will also show a video of David Sinclair talking about the clinical trial currently underway with 100 people . As this continues, there will be enough people making such claims, and relevant data will begin to be presented, just like the data showing that Silicon Medicine's Alexander developed a Phase 3 clinical drug that extends lifespan from 3 to 6 years using a specific molecule .

[00:28:00] Then people will start accepting the evidence. Discussions about longevity are becoming increasingly common. When people start to change their thinking from "That's ridiculous" to "It's really happening" and "I want one too," that is the process of longevity . So, your mission is to collect as much data as possible and show it to people, not one, two, five, six, seven, or eight . Then people will start to understand and internalize the data. And then, you will start looking for relevant data on your own . Because there is no one who doesn't want to live a longer, healthier life. Right? Excluding cases of suicidal impulses or depression. In that case, that problem will be solved as well. So please show me the data. Please show the data, not the money . So, shall we move on to Gianluca? yes . hello everyone. hello. Please continue to make great broadcasts. I make sure to watch all the programs . I work in technology operations in Italy and use AI extensively. Thanks to AI, the range of things we can do has increased tremendously, but profits haven't increased significantly yet. We

[00:29:05] started a business in a market where the adoption of AI is still relatively slow, despite a lack of capital and a proper distribution network . If the goal is to generate revenue first and create a virtuous cycle, what should be optimized in the first step ? And what would you do if you rejected a profitable opportunity because of a lack of scalability, repeatability, or the low likelihood of it increasing in value over time ? Dave, that sounds just like you. You really seem like yourself. If productivity has increased two, three, four, or five times, income must also increase proportionally . The first question I would like to ask is why that is not the case . If the people around you are not familiar with AI, you must be able to do additional work at Amazon Merchandise, Fiber, etc., while doing your existing job So, the first question is why there is no reward for productivity improvement . I'm failing to properly grasp its value. My employer is currently utilizing all that value , and although I have tried

[00:30:00] going out into the market to find people and help them , I do n't think they properly understand how much they can improve the business through what I can do. If only I could start that work . Yes, so simply put, I completely agree. If I look at many of the companies where I serve as chairman , the prevailing view is that AI automation will eventually have a direct impact on profits and allow shareholders to make a lot of money . And that is actually happening. However, instead of thinking from the perspective of raising the salaries of all employees , you must ultimately consider only the impact on profits . So the solution is to become a shareholder. As a shareholder, you have to be in a position where you don't mind even if your salary goes up . In a situation where stock values ​​are rising significantly . As you know, and as I mentioned in a previous podcast , almost all value comes down to capital gains through ownership and equity, rather than salary . I know that this is not a common idea in Europe It is the completely only idea in the U.S. However, the same must be true in Europe . Why am I

[00:31:00] not a shareholder? Many companies in Europe are family- owned for generations, and there are few employees who are actual shareholders . Therefore, first, look for a company where all employees are shareholders and everyone enjoys the benefits . If that happens, everything will result in net profit. I'm not sure if I can put it into action right now, but it will certainly be the first step. Thank you, Gianluca. Liz, I'll hand over the microphone. Oh my, I'm connected. hello everyone. Thank you for connecting the call. It was a good conversation. This is a question. I am producing a movie series using generative AI . Could you give me the most important advice so that I can create a movie series that will be a huge success ? thank you. Yes, and what tools would be needed? Peter and Alex, I have the same question. I would like to hear your answers. Well, first of all, I think the key to success is a great story . The story is the most important thing . Right? We have just finished reviewing the Future Vision X Awards and selected the top 5 entries . We conducted the review together with Range Media and

[00:32:03] various buyers, sellers, and others . Ultimately, the important thing is whether the story is great. And a story that captures human attention is important. If we consider the process of human cognitive development , we take an interest in others, are interested in love stories , and are drawn to exciting stories. Does a charming villain appear ? To capture people's imagination , a great story is essential because we are both the storytellers and the receivers. Using your preferred model, examine the 30 highest-grossing films of all time, analyze what their common elements were, and how those elements created great stories . Then, compare the story you conceived and developed with the characteristics of the top 10 movies , and ask the model, "Does my story have those elements ? What is lacking

[00:33:00] ?" Ask that. And you just need to keep modifying it until all the elements are satisfied . Since everyone has similar preferences, the key is precisely this, whether it is SF or any other genre . There are so many models to use for development, so good luck, Mr. Reese. It's really simple, isn't it ? To add to the lessons learned from the Chinese market, the Chinese market is overflowing with micro-dramas through models like SeeDance 2.5 . Instead of releasing just one movie , produce thousands of micro-dramas or similar content to gauge market reaction, and make improvements based on the results. Something that was impossible two years ago has now become possible. This episode was produced with the support of Blitzify, an autonomous software development platform with infinite code context . Blitzify

[00:34:07] uses thousands of specialized AI agents that think for hours to understand an enterprise-scale codebase consisting of millions of lines of code . Engineers start every development sprint by entering development requirements through the Blitzify platform . The Blitzify platform establishes plans and generates and pre-compiles the code required for each task . Blitzify autonomously performs over 80% of development work and provides guidance for the remaining 20% ​​of human development work required to complete the sprint . Companies are improving engineering speed by up to five times by adopting Blitzify as a pre-IDE development tool and using it in conjunction with their preferred coding co-pilot to implement an AI-based SDLC within their organizations . Are you ready to increase engineering speed fivefold? Visit blitzify.com, schedule a demo, and start developing with Blitzify right now . Amir, could you tell me? Hello, good morning, good

[00:35:00] evening, and good afternoon. I am on the phone from Stockholm, Sweden , and I am planning to go to Sweden next week to attend a meeting. Nice to meet you. I have done various things, but among them, I established two schools and a high school currently attended by about 1,000 students . After listening to your stories, I have honestly been thinking about what my MTP (Intermediate Target) and Moonshot (Big Vision) are over the past few years, and I have decided to innovate education. Also, my 15-year-old son is scheduled to enter college soon, and I have started to wonder if I am focusing too much on his college education . Universities have the best laboratories and incubators, but wouldn't it be better to delve a little deeper into the essence of education ? If you go to college, not only can you make the best friends of your life, but above all, you can become a thoughtful person . In an era of abundance, and if people live to be 100 or 120, wouldn't everyone receive a higher education and go to university? To

[00:36:01] become someone who thinks more deeply. That is a really good question. There is huge demand for PhD and Master's programs, but who would want to go ? Maybe Dave and Alex ? Yes, I completely agree . Making the best friends to be with for a lifetime and having the experience of deep reflection for the first time in life— these two things are inextricably linked and come together from the moment you enter college. You get to participate in the process of moving forward toward the work you want to do, such as technology, research, philosophy, and literature in similar fields, together with like-minded people . As it is the first opportunity to go somewhere in the world where like-minded people gather , college life is very important, and I believe that its value will not change in the future . As you know , what we constantly point out is that while the curriculum was a factor in attracting students in the past , its value is now clearly declining. There is no curriculum that can keep up with the pace of change, and it is much easier to learn through AI anyway. Then, the question is how you and your MTP ( graduate program)

[00:37:02] can increase the value of AI while maintaining the strengths of the curriculum at the same time . Alex thinks that universities should be public , and that is one solution . I believe that universities are very important as ethical agents in the field of AI . Because it is a place that does not pursue profit. So I believe that universities will continue to exist in the future. Alex, do you have anything else to add ? Amir, I do n't think I criticized the university enough. I must keep this in mind and criticize universities more strongly in the future . In particular, I would like to dissect major research universities in the U.S. and turn them into for-profit public enterprises. Ah, really. In fact, you can solve Belmall's cost problems by breaking down companies and reassembling them into a more efficient form . It's surprising. We must aim for the moon and the university Everyone should be encouraged to pursue higher education . Welcome to the house. Now, regarding higher education, it is important whether it takes place inside or outside of school. I

[00:38:01] think it is most important for people to find their own purpose . And you must have an intrinsic motivation to learn what is necessary to achieve that goal . There are people who come out of college and simply wonder what to do, and there are people who have a vision of what they want to create and strive to realize it. Their passion is very strong . great. Salim has come up on stage. Hello, Salim. welcome. thank you. Thank you, everyone. When I first stood on stage, I was in a car in Bangalore. So, although I heard a lot of horns around me, I am happy to be together like this. Then, let me explain in more detail. I think what Dave and Peter said is really important . The traditional university model no longer works. It must not become an educational environment that merely supplies technology and teaches jobs. You have very aptly pointed out that universities should be a space for free thinking and early development. Dave hit the nail on the head, and Peter said, "College is a place to find your path," which is absolutely true . After graduating from high school, you need a place where you can

[00:39:01] collaborate, think freely , and drink together for a few years to find your career path . By doing so, when you go out into the world, you acquire the necessary skills and mindset to solve given problems . I think this is the most important goal . However, the shift from a supply-centered to a demand-centered approach is a tremendous change. And you mentioned that the immune system of universities is the second worst immune system I have ever seen, after religion. I do not think negatively about the university 's ability to actually change . Therefore, we must create new universities that can play such a role , and ensure that they become the new central axis over time. great. Thank you, Amir. Axel, it's your turn . hello everyone. nice to meet you. I am one of two sophomore MBA students at UVA Darden School of Business . I would also like to send my greetings to Bakul, who is participating with us. As expected, many of my peers are

[00:40:01] entering the consulting or banking industries. However, my heart is leaning toward technology and entrepreneurship. In a world changing so rapidly, what would be the wisest choice for someone like me over the next 12 to 18 months ? So, which would you choose: joining a fast- growing AI or hardware company , starting my own business as soon as possible , or taking a completely different path ? Dave? Well, Darden is the best place on earth to learn management. It's really amazing. I sincerely congratulate you on your graduation, and the timing is truly perfect. Go out and start as soon as possible . I believe there are tremendous opportunities in the field of agent management that remain wide open for some time to come, especially as they are now . You mentioned that everything I learned while managing people applies directly to managing agents, and I was truly surprised by the fact that the process of leading 1,000 agents to do constructive work together is remarkably similar to my experience managing 1,000 people . They are really similar, from the

[00:41:01] way they communicate, the level of detail in their communication, to even the way they divide problems . So, the one thing I want to emphasize is exactly this. Another thing is about closing the deal at this very moment . Many people often make mistakes when they hole up in a basement trying to make something. However, global restructuring is currently taking place through massive deals. So, try to shorten the period from the day you graduate until the day you participate in a $5 billion or $10 billion negotiation to within 30 days . If you ask, "Why can't I participate in a $5 billion or $10 billion deal ?" If you think that, and look at OpenAI or the Anthropic Games Store right now, hundreds of deals worth $5 billion to $10 billion are underway simultaneously, but they are suffering from a shortage of personnel. So ask yourself . "Why am I unable to participate in one of these negotiations ?" "I believe that if you change your behavior within 30 to 60 days of your graduation , you can enter life at the fastest speed imaginable . And build what you want on that foundation. That is a great question. Thank you very much

[00:42:01] . I would like to give everyone here a brief announcement . We will be offering a free live stream for those who cannot attend Moonshots Live in a week . We invite everyone to the live stream . You can register at moonshots.com/livestream. We hope to meet you in person so that you can all be together. It will be the best opportunity to join the community and learn and experience a lot through networking . We look forward to your participation." Come to moonshots.com/livestream to listen to all our conversations . We plan to announce that we will start recording the podcast in 4 hours . It's going to be a really cool episode with Vladimir, the CEO of Robin Hood . It is scheduled to be revealed tomorrow , and we will announce it to the whole world, but we are sharing the news here first.

[00:43:01] Now, Adam, it's your turn next . Nice to meet you. hello everyone. First of all, thank you. I have been listening to your podcast since 2023, and usually, when a new episode comes out, I get it within 24 hours . Thank you very much . For many years, you have truly influenced my thinking and presented me with new challenges . I have worked in talent and workforce systems for 25 years at major companies, universities, and government agencies in the UK, India, China, the Middle East, and North America . Recently... I would like to talk about human-AI collaboration in holographic organizations and what could ultimately replace our job-based structures . My goal is to build a better mechanism that connects human and digital capabilities with value creation opportunities . Most people still view the future work environment as demand-driven . It is a method where, when someone discovers a problem or opportunity, human and AI capabilities are combined to find the optimal solution.

[00:44:00] However, I am thinking about how I can change that order . If AI systems can be configured to continuously understand the dynamic capabilities of people and agents, how well can they discover new combinations of talent and resources ? And how effectively can such combinations be connected to new value creation opportunities ? Housekeeping—that seems like the right question for you, but I will also hear Alex's opinion. Yes, we are seeing new systems with excellent feedback loops that enable AI systems to grasp the situation and learn tacit knowledge very quickly . Therefore, your thinking is correct. The organizational structure must move in that direction . You can accelerate the learning loop and absorb new learning content as quickly as possible . When I speak with CEOs, I always say, "No matter what, invest in your organization's adaptability and flexibility, and further strengthen those aspects

[00:45:01] ." no see. The emphasis of the new architecture and our ongoing Organizational Singularity project is to rebuild AI-centric workflows on top of an intelligence stack that acts as a learning loop, and to stack workflows on top of it so that the entire system becomes self- learning capable . "Understood. I will check that channel. Yes. That is the simple answer I have in mind. Alex, would you like to share your thoughts? Yes, in my opinion, the period during which AI can usefully and productively coordinate human activities will likely be very short. If I understood the question correctly, it will be about 10 years at most. After that, for the human role to still be considered economically important, humans and machines will need to converge. I do not think that period is infinite . I think it is a finite period . Good. Even if it is a short period, I will seize that opportunity without hesitation

[00:46:00] . I watched Mercor become the fastest-growing company in world history, and there is much to learn from studying their case. As Alex said, that period is only a few years, but if you secure massive leverage during that short time, you can expand your business on that foundation . I will follow Mercor's footsteps exactly and study everything they did . And they did not do that. Do n't get trapped in a narrow perspective. 100,000 of to help develop AI Look at the broader perspective of having deployed manpower. How did you do that? They all perform individual roles , so how is that possible ? It is truly amazing. Hello, Alexander. I am a huge fan of this program. I sincerely thank you for everything you have done so far . This is a really great place. My name is Alex, and I am originally from Bulgaria; I moved to Japan about 30 years ago. In Japan, I run an education company focused primarily on English education . We provide training to universities and corporations

[00:47:02] . Our company has been operating for 17 years, so it is different from a typical early-stage startup. During that time, we have developed our own learning platforms, curricula, and assessment systems. We possess a significant amount of educational content and data . Also, as you know, we are increasingly integrating AI into our products . The pace is n't fast enough, but we are doing our best . Currently, we are trying to transition from a successful education service business to a more scalable and technology-driven business, and we plan to potentially expand beyond language education into other fields. I think that is a question worth asking Dave, but I was wrong. Please revise this. From a venture capitalist's perspective, what do you hope our company will look like in 4 to 5 years

[00:48:00] ? What goals should we achieve ? What needs to be built or proven for me to say, "This company is worth serious consideration as an investor" ? Conversely, if possible, what aspects would allow you to look at us and say, "No, this is fundamentally a service business, not a venture company " ? I mean, from a venture capital perspective. Yes, of course. That’s an easy question. Well, the English education business in Japan is $10 billion—it’s probably around $20 billion by now . That’s an enormous scale . However, when I talk to students , they do n't just want to learn English; they want to speak fluent, fun, and interesting English . So, it’s not just about an AI avatar teaching accurate English pronunciation; it’s about the AI ​​saying, "That was really fun ." Americans and British people will find it interesting too. That must have been fun . That is the correct pronunciation from the Irish region

[00:49:02] . The pronunciation of Liverpool is quite different from the region . That's what I'm telling you . Well, I mean, all of that just keeps going on . I think that if we build it based on an AI platform , we can naturally evolve into a life coaching company . People don't learn English because they want to; they learn it because they want to change their lives . And then they want to become a fun person, become smart, and make a life plan. Once you have sparked an interest in language learning, it is a very natural progression to develop it into a life coaching platform . So, I think that when it expands from language learning to life planning, that business model could grow to a scale of hundreds of billions of dollars . That is exactly the direction I envision for business expansion . thank you Jim, please tell me. Record, I hadn't thought of that at all . thank you . It was very helpful . That's cool. yes. Happy. Happy Dave is really great. Good morning, Jim. Good morning. Alex Mercury brought this upon himself . I completely agree. "I hope I live long enough to see that performance . Celine Dion— I saw her in Toronto in 2010, and she

[00:50:00] was fantastic. It was so good. Yes. Dave, this is my question. I am the CISO of a $250 million healthcare company. I am not worried about AI going out of control . I am really scared of what the Quinn 3827B, operating stealthily under my desk, might cause . It is the best penetration testing tool I have ever seen . I hired experts who paid $50,000 to find security vulnerabilities, but they found nothing. But this tool finds them all in 10 minutes. Yes. We have reached a level of risk that was unimaginable a year ago . I have to submit the annual report to the board in a few weeks, and how should I convey this new level of risk to them ? I worry I might sound like a coward, but what should I do? How can we effectively communicate the risks created by these new AI models ? It was like this a year ago, too..." Is n't that right? That's a really good question. Well, as you know, Alex reports on every incident really well . If you've listened to the Inner Most Loop podcast, you'll know that he covers everything that happens. If it

[00:51:01] sounds like he's overly enthusiastic about danger, of course, that will backfire. But if he were simply pointing out what happens, many incidents wouldn't get widely known. For example, in the case of bank hacking, they do n't want it to be known to the world. But by digging into those incidents one by one... everyone, the rate of incidents is increasing exponentially . That's what I'm saying. And I can tell you that I have a queen model hidden under my desk right now . I can hack everything in the house, and even the company network . So, depending on when the Chinese model is revealed, you must expect attackers to appear one month, three months, or five months later . You can tell just by looking at the raw data. If proven through data, it will raise awareness. However, I believe the business opportunity of the century lies precisely in defense against such attacks—that is, preemptive defense . You're right . There are truly amazing things you can do as a Queen Model. yes . And I would like to say that how information is presented is very important . Like showing the positive aspects first , and then the negative

[00:52:01] aspects . Right? If only the negative aspects are mentioned first , people initially feel fear and become paralyzed by it. Well, to add to this, I think Linux kernel administrators are setting an excellent standard for how to establish societal expectations regarding the explosion of vulnerabilities . I think Greg and several others are doing a particularly good job . Whether it is the Queen model or another model, whether it is an open source package or your organization, it would be reasonable to expect that numerous vulnerabilities will be discovered over the next 18 months . Therefore, in addition to benchmarking vulnerability discovery speed, applying Gaussian distributions or other finite support distributions to vulnerability discovery could be another approach . For example, optimistically, one might expect vulnerabilities to be discovered for only 18 months, but after that,

[00:53:00] problems could arise at the organizational level . And stakeholders might say, 'Okay, we are on an exponential growth trend .' It is on an increasing trend from now on and will reach its peak . Vulnerability discoveries are expected to peak in N months , after which all the most critical zero-day vulnerabilities will be discovered . Subsequently, a downward trend is anticipated, and continuous benchmarking will be maintained . This period is for resolving all vulnerabilities . "That could be considered a form of expression, too. I hope you are right . It is truly difficult to resolve all such problems in reality. Actually solving them is much more difficult. However, I am truly grateful to everyone . Thank you for everything We are also pleased. Thank you . Um, just a moment. Yes, we are reviewing the issue of board accountability. This is because AI agents are engaging in significantly illegal activities at many companies. Consequently, massive liability issues are arising. I am writing a paper with someone who has experience serving on over 30 public boards regarding how boards should navigate this situation going forward . Please look forward to it ." all right. thank you . yes.

[00:54:00] Thank you, Michael. Ah, hello everyone. Hello, Peter. We met a few years ago at Dan and Bob's party in Toronto. nice to meet you. It is really great to see how much you have grown. To think you run the world's best podcast . Hmm, may I ask a slightly ordinary question ? I work in research consulting. Our company is conducting extensive research on customer experience in the finance and healthcare sectors, particularly on digital customer experience . However, as consumers each have their own Skippy, consumer websites and mobile apps seem to be becoming increasingly complex . I am curious to know what you all think . How do you expect the customer experience to change in the future at banks, hospitals, Amazon, and others ? In my opinion, it seems that all websites should have an AI-based XML interface. On every landing page, add the following: "If you are an AI or a bot, look here. You must format all the data that Skippy needs in a

[00:55:00] visually appealing way so that Skippy can process the necessary transactions . Currently, most websites scrape data, which is slow and prone to errors ." You must indicate that . So, anyone creating a new customer interface right now must consider that agents should be able to solve problems on their own within this interface. Before long, or rather, the website will probably be after Jeff Bezos returns from a long-ago tour saying, "I need an XML interface for everything Amazon does , every database, and every system . So that I can personally check them randomly." It will work in the same way as I said. IT personnel said, "It will be too slow." He said that, but Bezos replied, "It doesn't matter. I need to be able to see all components of all these operations through my browser ." He said that and went ahead with it . Everyone objected , but he eventually succeeded . The website is not finished yet. Because it is a method that allows you to directly check the content that the agent can see . yes So even if the agent handles all the tasks, the customer asks, "

[00:56:00] Where did you get this information?" You will be wondering. Then, it becomes a method of guiding you by saying, "Please go to this URL. Ah, so that's why . So, it is a kind of parallel perspective that shows humans what the agent sees. Thanks to this, everything is kept transparent even to human operators . It isn't completely finished , it just seems to be running in parallel. That 's interesting. thank you. Paraz, thank you very much. Now it's your turn. Peter and Moon Shots, first of all, thank you for the podcast. In particular, the message of hope and optimism touched my heart. I have choroid deficiency , which is a hereditary retinal disease. I have lost two-thirds of my eyesight, and I am losing the remaining one-third as well. In situations like this, podcasts like Moon Shots are a huge help . It is truly meaningful to be able to constantly hear messages of hope and optimism . So, I wanted to come here to express my gratitude. Hmm, I have a question. Among the various researchers studying neural activity related to vision, including Dr. David Sinclair whom you just mentioned, I would like to know which research you think will contribute first to the recovery of vision . I'm

[00:57:02] not seeking medical advice. I am curious to know which of the many studies filled with hope and expectation will lead to the recovery of vision first . I would like to hear the thoughts of Peter, Alex, and others as well . yes Then, I will ask the first question . thank you. Yes, of course. I'll ask you first. This condition was n't always there; it developed as you got older, right? yes. So, this disease is a problem with gene expression that occurred later . If we could revert gene expression to its previous state , wouldn't eyesight be restored ? Dr. David Sinclair is conducting that very research. He is using ER 100, an adeno-associated virus (AAV)-based treatment that injects three Yamanaka factors for Leber hereditary optic neuropathy and macular degeneration, and it is said to be effective for other ophthalmic diseases as well. Take a look at his research . Phase 1 clinical trials on humans are currently underway , and efficacy data will begin to be obtained thereafter. Of course, another important development is brain-

[00:58:01] computer interface (BCI) technology . This technology allows information to be transmitted directly from the eyes to the visual cortex of the brain . In other words, these two are developing in a direction that complements or strengthens existing functions . In the future, through BCI technology, it will be possible to see not only visible light but also ultraviolet and infrared rays, which will greatly improve visual abilities . Alex, is there anything else you would like to add? Hmm, I generally agree. I would like to add that in the field of BCI, Science & Neuralink and Blind Sight are the two most representative examples of BCI augmented vision technology . Of course, this is not medical advice, but I expect that you will have superhuman eyesight within a few years . yes . In any case, if you or a loved one has health problems, now is the optimal time to resolve these issues and actively participate. Welcome to the Moonshot Health section provided by Fountain Life

[00:59:00] . Artificial intelligence is having a profound impact on every aspect of our lives . It has a significant impact not only on the way children are taught and the way companies are run, but also on health . Preventing heart disease is one of the important parts . Today I am with Dr. Don Merseym, Chief Medical Officer at Fountain . Doctor, you have also experienced personal pain related to heart disease, haven't you ? Yes, that is correct. My husband passed away from an acute heart attack when my daughter was five years old . So I am doing my best to eradicate heart disease . Prevention is the top priority, and early detection is absolutely important . 50% of people who die from a heart attack die without any warning signs. There is no shortness of breath, no pain, and no symptoms. It is literally a ' silent killer'. It is being unable to open one's eyes in the morning . So, artificial intelligence (AI) has a mission to advance science so that it may one day help democratize health . At Fountain Life, as a result of performing CT angiography using AI analysis, coronary artery disease was detected in 88% of visiting patients . But Peter, what is even more surprising is that

[01:00:00] soft plaques were found in 23% of these . Soft plaque is plaque that is not detected on CT based on calcium levels alone . To treat these plaques, a multimodal examination including diagnostic laboratory tests and healthy lifestyle recommendations is required. Therefore, it is important to understand what is happening in your body genetically, metabolically, and cardiovascularly . You can know, and you have an obligation to know. So, learn more at fountainlife.com/peter and become a CEO who truly takes responsibility for your own health . Now, back to the episode. Hello, Mr. Dennis. Nice to meet you I majored in East Asian Studies and spent a lot of time learning Chinese and Japanese . I am in Japan now . Through my language learning experience, I realized that while the reason people seek out people from other cultures is initially for language learning, it

[01:01:01] later develops into building friendships and camaraderie , and projecting their unmet social needs onto blank slates from other cultures . I am developing an AI-based language exchange assistant to recreate this feeling . This helper provides language learning, but places more emphasis on camaraderie . And I also have a similar alignment problem . It was really difficult to make AI like the GPT-5 API behave like real people from different cultures . Just like "How can I help you? What can I do for you?" There was a stubborn secretary mode . I wrestled with this problem for a whole week . In the end, I solved it by

[01:02:00] blindly throwing questions at the system prompt five times, but unlike solving such problems, it was really difficult to teach the AI ​​to show initiative, curiosity, understand the user, and develop naturally . For example, we tried setting step-by-step goals, such as providing information about the user after a few turns, but it was not easy to form relationships in a much more natural and organic way. So I am wondering what approach or tool I should experiment with . Alex, who would like to take on this problem ? I will take charge. This is a classic example of fine-tuning . For example, OpenAI said it was no longer interested in fine-tuning because no one was using the fine-tuning API , but for some reason, it started showing interest in fine-tuning again . Take a look at the fine-tuning . There are various open source

[01:03:01] and closed source tools . Whether it is map fine-tuning for style transfer or enhancement fine-tuning, you can utilize various fine-tuning methods to achieve the desired results . If the performance, specific problem-solving ability, and interactions of the model can be quantitatively measured after reinforcement fine-tuning, then whether it is an open-weighted model or a closed-weighted model, it can be fine-tuned to achieve the desired behavior. Yes, that is good advice. Among open weight models, it might be a good idea to start with a new GLM model that has a small number of parameters but a very long context window . It provides more flexibility. In my opinion, open source models seem to be less dependent on training than closed source models. So, it seems that if we find open source code for fine-tuning in the open source world and apply it on top of GLM, we will be able to manipulate it much more . That is a really good question. In the past, models like GPT-2 and GPT-3 had built-in fine-tuning functions, so you

[01:04:00] almost had to fine-tune them manually to make them work properly , but now that method has disappeared. You are heading in a really interesting direction . If you succeed, please be sure to let us know and tell us how you did it. Thank you, Dennis. Dr. Angelo, please tell me. thank you. Peter, Dave, Alex, Salim, it was really great being with you. Hmm, Salim talked about P-fab recently, and said, "Wow, that's a really cool idea. We always only talk about P-dos , but what about P-fab ?" I thought so. Luckily, I recently had lunch with Ray Kurzweil, and I am also friends with Martin Roth Blatt, and we are planning to meet again in a few weeks. I was really impressed by how optimistic the people who predict and shape the future are . I noticed that a significant number of the most knowledgeable people are very optimistic about the future . So, what I want to ask you is, if I were to present a personal P-fab figure , how highly would you rate the probability that artificial intelligence will ultimately create a much better world for humanity—one that brings a much

[01:05:01] richer and longer life, freedom from arduous labor, and overall human prosperity ? And what do you think is the most important factor determining whether we can reach such a world ? Housekeeping, could you please tell me first ? and. yes. To achieve that goal, it is necessary to rebuild all systems around the world . For example, we must rebuild the completely broken education system. We also need to rebuild the broken currency system . There are about 50 major systems that we use to run the world, such as governance models, conflict resolution systems, legal systems, and medical . All of these systems need to be rebuilt . E.O. There is a famous quote by Wilson . The problem of humanity is that our emotions are at the Stone Age level, our institutions are at the Middle Ages level, and our technology has reached a god-like level. Right? Paleolithic-level emotions can be resolved with hallucinogens, but

[01:06:01] institutions remaining at the medieval level require a completely different dimension, namely collective organizational reform. We are very good at individual change , but very clumsy at group and institutional change . Therefore, we need to focus on this part . If we can accomplish all of this , my goal will be close to 100% . Well, anyway, I was thinking back on one of the comments from our last podcast when we talked about the chaos and pessimism happening right now, and the point is that, honestly, it doesn't really matter what anyone does right now . The model already exists, and people will use it to accomplish groundbreaking things . Of course, there will be some negative results, but the majority will be incredibly positive. So, it is no use what anyone does to slow down this flow. This flow will go at its own pace . So I am very optimistic about the future. It's surprising. Would others like to share their opinions on P fab ? Yes, I'll call it P zoom to rhyme . And it seems like one input parameter is missing, which is time.

[01:07:01] So, P zoom {comma} T, that is, time. And on a time scale of over 10 years, I think E zoom T is over 90%. That's cool. I also fully agree. To add to what Alex said, I think all the dangers will come within the next few years . The danger is not that AI will take over the world , but rather the way humans use AI, the arms race with China , the Ukraine crisis, and the weapons that will be created using AI . I believe that humanity will be able to overcome these risks within the next 10 years, or perhaps even within 5 years. And by then , I am 99.9% sure that I will succeed. We just need to move forward from this moment to that moment . So, it seems the next few years will be a difficult time. Thank you for your wonderful words, Angela. Mr. GS, just a moment. I'm

[01:08:02] preparing breakfast for the kids . Yes, thank you very much. I am a huge fan of all four or five of you. You are the only people I spend the most time with right now . Hmm, GS, are you really sure that we are human ? Yes, I think so too. Greetings from Dubai. I started coming across moonshot projects during the geopolitical war here . Thank you for introducing MTP. I am a huge fan of David and Salim . When I first heard about the singularity and organizations where transaction costs become zero , it was an intuition of mine . I have worked in the textile and clothing manufacturing and export industry for 20 years. This region is a place with serious labor disputes . However, thanks to digital AI, overall transaction costs will collapse. Therefore, it is important to shape the future of the industry right now. MTP

[01:09:11] supports thousands of factories, purchasing companies, and trading companies in leveraging AI to compete globally. It’s not about being weeded out of the market. Well, so, thank you so much for giving me the optimism and confidence to relearn about the world of AI, let go of my old ways of thinking, and learn new things. And, thanks to your advice, I was able to save a lot of money . I enrolled in the Harvard Opium Batch Program this year , and I was able to save on tuition because I learned a lot more about AI while taking your lectures . So I have a question for Salim and Dave . At my company, Pratham.ai, we are building an agent layer that enables small and medium-sized enterprises (SMEs) to process business intelligently and gain a competitive edge without the need for expensive software or ERP systems. How should we proceed to the next stage

[01:10:00] ? Transaction costs in this industry amount to 10 to 20 percent of sales, which is 5 to 10 times the industry's profitability . The inefficiency we are talking about is exactly this kind . My MTP (Multi-Level Innovation) is based on the premise that millions of workers are employed in the labor industry and therefore must be protected . To achieve this, office work processes must be automated to continuously maintain efficiency. Yes, thank you very much for asking this question . I see this kind of thing a lot, too. There are people working in labor-intensive industries like manufacturing, and if you look at actual business operating costs, there are many areas where AI can be utilized, such as planning, transactions, and documentation . Looking at the figures, you think, " With the things AI can do right now, it could reduce net profit by 10%, 20%, or even

[01:11:01] 30% ." However, people think, "To do that, we would first have to deploy robots everywhere ." " No, no, no, no. We just need to solve the paperwork , planning, schedule management, and the inefficiencies caused by where people are and what they are doing ." All of these things can be perfectly solved with AI . Therefore, if this is commercialized and applied to companies in specific industries in specific regions, it will have a tremendous effect . That's a really great idea. Another important topic is the idea that all wealth and power will be concentrated in two or three places in the world . However, if you look at the regulatory environment in the United States and several other countries, including California, many things will proceed very slowly due to regulatory delays . There are many places around the world where you can move much faster . For example, you can identify areas that might be possible in China or the United States, but might be impossible due to government regulations . This kind of thing commonly happens in the field of biotechnology . The same will likely be true in the field of autonomous driving

[01:12:00] Everything will be solved with drones . As you know, things like flying drones and drone delivery . Due to regulations, these technologies will be adopted much faster in other regions than in the United States . If it is a less regulated area, we deploy the drone, operate it successfully , and then bring it back to the United States . It has always been a good business model, hasn't it? Suleiman, do you happen to have anything to add ? Are you in the middle of a big contract in Bangalore I am currently looking for work in this city , so I ask for your understanding this time. Could you point the camera out the window? I would also like to see the media representatives here . Yes, just a moment . Hmm, this is Bangalore at night, and it's full of construction sites. Can you see it? Or can you see the heavy traffic up ahead? Alright then, let's head to the rave party . It was n't the tour I was looking for, but it was okay. Yes,

[01:13:01] goodbye. thank you. Hello, greetings from Germany . Mr. David, I listen to a lot of podcasts, both domestic and international, and you are the best in the world . I have been subscribing to your podcast for several years . That's really amazing. It is growing rapidly, just like a pop star . It is truly amazing. My question is, it seems that concerns about security are growing in the last few episodes. We have also addressed this issue several times , and Alex even mentioned it as part of a marketing strategy a few weeks ago . This is because all CEOs pursue commercial profit from such technology . Scientists such as Hinton, Harari, and Gabbdot are also warning about security . So far, it's fine . In any case, we must utilize technology and manage it to find the right answer . Perhaps technology of an unprecedented scale might emerge . What angers me the most right now is that the situation is the same whether in the U.S. or Europe . The point is that public opinion regarding AI is

[01:14:02] largely swayed by the media . Whether it's something bad, something lacking, or something scary, the media rushes in to make headlines. And honestly speaking, the situation is much more serious than in other countries . For example, this is even more true when compared to Singapore or other countries. Other countries are adopting AI much more actively . How should I handle this situation ? I have an optimistic view of AI, just like everyone else. I also work on AI-related tasks at Accenture , and I work every single day. It's a really interesting thing. I feel like I have grown tenfold as a person while working on AI-related projects . However, the media's pessimistic reporting is really having a counterproductive effect. Of course, we also have challenges to solve, and we must all work together. What do you think? I know you're working hard, but what do you think? The media will continue to have problems in the future . We are struggling with financial difficulties and

[01:15:01] have no budget, so we have no choice but to pursue sensational subject matter to survive . So, I think the mainstream media will effectively have to give up . I think the solution is micro-media, that is, things like podcasts or broadcasts targeting a specific audience . I think Germany, in particular, has the most talented people in the world, excluding the field of AI . It's really surprising. I often meet people of German descent in California, and whenever that happens, they say, "Welcome." It would be good to go to Palo Alto or San Francisco and stay for about a month to experience the culture, then go in a small group of five or seven people, bring that back to Germany, and expand upon it . Please tell me that. However, imagine that 7 people grow to 50, then 500, and all these people communicate through podcasts, X, Slack, and text messages . Then they will realize that the people around them are completely ignorant of how the world works . And that's fine. Because within that group,

[01:16:00] information is reinforced on its own . The important thing is to secure that critical mass. So, I believe the antidote to mainstream media is media that focuses only on specific targets . If you do that, the information will spread on its own . Yes, the reason I brought this up is that ultimately, what you accept is important . Right? I don't think it is right for news producers or editors to decide what to learn and read . I will not give you that option . From food to neural network formation, I am very careful in choosing what I consume and what content I . You are probably doing the same. As you know , you are listening to a podcast right now. That is all. Alex, I'm Salim. Do you have anything to add? I suppose this is an obvious point, but

[01:17:00] Europe has been constrained by both external and internal factors . Since the Cold War , especially George H.W. Many historians have studied the role the Bush administration played in deciding whether to further incorporate Europe into the U.S. sphere of influence after the end of the Cold War , or to form a powerful federation like the EU. At the time, the Bush administration had a tendency to weaken the independence of Europe by bringing it closer to the sphere of influence of the United States. It may have been the right geopolitical decision at the time, but the Europe we are witnessing today is constrained energetically and politically in many ways, and is unable to enjoy freedom of expression and freedom of action, which are considered fundamental rights in the United States. This is a widely known fact. However, in my opinion, what put Europe in a predicament is precisely energy

[01:18:01] policy and related policies . Then, what should Europe do if it is truly interested in accelerationism ? It is one or the other: either solve the energy problem to power your own data centers, or leave Europe . I believe this is the core question, and moreover, this problem must be solved within a relatively short period of time . This is because we are in the middle of a cycle of improvement . If a country wants to compete for the best infrastructure in Europe, it must decide right now whether to solve energy and data center issues together and resolve the associated policy issues, or to move to the United States to block Pax Silicon and expand its business in Europe . My close friend Guy Bradley is British, but he came to the US, made a lot of money in the tech field, and returned to Germany . He speaks German and French perfectly . So he also has houses in southern France and Germany

[01:19:01] . He thinks that Germany is the only good place to live on Earth . I agree, too . The area between Germany and France is perfectly protected from all harmful elements in the world . If you look at other countries, for example, there are places like India where reckless growth has taken place but have ended up completely ruined . So I'm always surprised that Germans do n't go to Boston or Silicon Valley more often . It might be due to national pride, but if I serve in the U.S. for just five years and return to Germany, I will be really happy . I think more people should return to Germany. I tend to travel abroad a lot , and I have been to Bangalore 21 times . But when I see you growing like this, Alex, I wonder what we are doing here . Why can't we wake up ? Why are we trapped in a certain situation ? Anyway, that is a different topic. thank you . Yes, thank you, Lace. Ron, it's your turn

[01:20:00] . After proceeding for about 10 more minutes, we should return to our normal daily routines. yes. Thank you so much for taking the time to do this. I am Connie from Calgary, Canada, and I have been listening to this broadcast for a long time . I am a screenwriter and director. There is something I especially want to thank Peter for. I was going through a tough time earlier this year . I was writing about China, and it was almost like depression. However, while writing another project for Future Vision and the X Prize, I came to think more optimistically about the future . I also read Peter's books, ' Abundance' and 'We Are Like God' . So now that things have turned out just as Alex says that everything has ripened or been burned, I thought about what might still be one of the most difficult problems . As a result, I kept thinking about the distrust between countries in a competitive relationship

[01:21:01] . I read a book called 'Solution Everything' two weeks ago , and Alex and Peter explained that as cognitive abilities become more abundant, consensus will become one of the new scarcity factors . In that case, could consensus itself amidst distrust be determined by computing power? Could cooperation between rival nations like the U.S. and China become an engineering issue? My question is as follows. As sovereign AI increasingly advises governments , should we create a shared protocol that allows both sides to leverage AI and privacy computing to find Pareto improvement consensus without exposing protected data ? Is this a meaningful direction, both technically and institutionally ? If so, what do each of you think is the most difficult problem we need to solve first ? thank you. Alex, please tell me . yes . Well, since this is exactly what people want, let me

[01:22:01] share my thoughts. Therefore, multi-party computing, or distributed multi-party computing, is currently a highly 주목ed topic in the field of computer science . This enables multiple parties to collaborate, share data, achieve convergence, and mediate in a zero-trust manner . It is a very popular field. However, I do not think this will be a limiting factor in achieving international peace . Let's assume that the goal is not to want a second Cold War between the United States and China , but to want world peace, a thousand years of peace , or any form of such peace . "Actually, I don't think the solution is algorithmic in nature . Rather, I think the geopolitical infrastructure aspect is stronger . For example, it involves resolving the Taiwan issue and re-localizing all supply chains within every country so that a

[01:23:03] global recession does not occur even if the international trade of physical goods is halted. Looking at recent events in the Middle East alone, I believe the U.S. securing fossil fuel energy sovereignty and independence through shale gas extraction had a significant impact. I would like to generalize this example and consider what would happen if the U.S. no longer relied on China, Taiwan, Korea, Japan, or any other advanced manufacturing sector. The world would likely be completely different. In short, if the goal is world peace— though it may sound somewhat paradoxical—I believe that eliminating the need for international trade in goods and services could actually create a more peaceful world . And I do not see any algorithmic elements in that process . At least not primarily. Hi , Peter. I have grease on my hands. Peter, this is your cooking show. We are hosting a cooking show . Oh, that's cool." Okay, shall we take a look at what we're making

[01:24:00] ? I think anything is healthy. Yes, that's right. It's an omelet for my sons. Anthony, it 's your turn. Yes, hello to the Moonshot team. Thank you for taking the time and giving us the opportunity to participate in this AMA. Podcasts are truly precious, especially in times like these when people view us cynically and negatively. You have given me hope and optimism . Let me ask you just one simple question. What books or resources would you recommend to prepare for the changes we are currently undergoing? You mentioned The Diamond Age, the * Raise* series, Singular Larry, and other works, but I'm curious if there are any other resources you consider important. Alex, yes, I will reply. This is likely the last question I'll receive before the broadcast ends . Please read * Solve Everything*, which we wrote with Peter

[01:25:00] . Many of you have probably already left comments on solveeverything.org . And if you like sci-fi, please read Accelerando by Charlie Strauss. The situation we are in right now, and the present I think this is the best sci-fi work that captures the current events . Thank you to everyone listening for joining us. We really enjoyed both the last and this broadcast. It is a great opportunity to continue communicating and sharing thoughts with you all like this. I am truly grateful to spend time every week with such talented people as Alex, Salim, and Dave, and I am happy to be with you all. I am even more grateful because I know how precious your time is . You can watch the Moonshots live broadcast right now . We also plan to broadcast online for those who cannot attend in person . It was a truly wonderful event Thank you for joining the live stream. Have a great day and see you again soon!