talking ai

林永青 原创 | 2023-03-12 06:01 | 收藏 | 投票



talking-person1 00:02
Got it. Thanks again for setting china values, interview and dialogue. You have very strong experience and expertise in several relative sectors. First topic, we would like to discuss the conventional wisdom on how technology will change the future. It is wrong. You mentioned in your book and we also agree with that. You lay out radically different and uptick, mystic vision for what really coming. According to your opinion, the in innovation will come not from any single big invention, but from conference of radical advance in three primary technology domains, like a microprocessor.

By the way, ii use work for intel, who is the one of the biggest cpu maker? The second is materials, actually new materials. The third one is machines and ay now, can you please explain and why you are come to this conclusion?

talking-person2 01:30
Sure, I you first can do you can approach technology forecasting two ways. One is you can use histories lessons because history tells you something about patterns, right? People always ask like act like this time, it's different, right? And computers today are different than adding machines of a century ago. So obviously, but the patterns of history tell you something about that. There's an expression that history doesn't repeat. It rhymes, which is there are patterns. And if you think about how the 20th century of all, but what happened? I it's very common to look at a product as an invention by itself, the car, the airplane, right? This. And obviously the airplane changed a lot of things, but the airplane or the car was made possible the confluence of other developments that it occurred, roughly contemporaneously.

So that's really the sort of the key . of my book is to first frame how innovation happens. You needed to get to stick with the car as an example to make the modern automobile age happen in the form it took place. You needed the invention of an internal combustion engine, obviously, but that was only made possible by advanced steel, which was independently invented. It also required the invention and development of a refining industry around oil, which was totally independently developed. And you needed the idea, a refined idea of how to make a mass production line. The idea of making lots of parts of something is not new. It's very old. In fact, arguably, the earliest mass production lines were in china 2000 bc the british had mass production lines at 1,000 ad for arrows for their long bowman. But the kind of mass production we're talking about the modern era is characterized by requires special machines and timing and controls clocks, control systems, which are all being developed roughly the early in the early 20th century.

Those things together made it possible to make high quality, inexpensive automobiles wasn't just one thing, was confluent to things. Same was true of the train in its the development of the era of the steam engine happened because of, obviously, the steam engine itself.

But again, high strained steel and the telegraph, you couldn't coordinate the movement of trains on a single track and bunches them without a fast communication system to move faster than the train.

The point of that is that today we live at a similar time, what's been going on is these confluence of technologies. They typically fall into three domains, information systems, broadly speaking, acquiring information, controlling it, the materials available to build things. Right? And then the underlying logic behind behind systems. If it, the physical machines that you can make to us to fabricate something, change the game, those things happen episodic over history, and then we engineers could make everything better. So you invent the things. And then for a period of many decades, the products and services get better, they get refined. I think it's cheaper, they get more efficient, they develop more speed or capabilities.

Then you have a long period where nothing seems to change. That's sort of what's happened for the 20th century. My contention is that now we're at aa time where there's a confluence of revolutions and the same three domains, we have radically different kinds of machines for manufacturing things, three d printers. We manufacture things, the molecular scale, intel, your former company, I worked a long time ago when rca before intel became a big company, rca with texas instruments and ibm with a big semiconductor company.

So I manufactured it worked with the process engineer and semiconductor fab that the machines we use then are what made it possible to manufacture. The feature sets are so tiny. Those machines didn't exist 30 years earlier. Right? And they've gotten better and better. But the class of machines available to do the photography, right? To do the etching, to do the assembly. Those things were remarkable, sort of unsung advances that combined with the advances in material science were growing at the same time, really led to the computing aids that we have. The same things going on.

Now we have new classes of materials. People don't have really good names for them, smart materials. These are materials that are reacted to the environment there. They're designed to be inherently reactive and adapt themselves, be self healing, really, quite remarkable. Biocompatible materials, the biggest challenge physicians have in surgery. It's finding materials that your body doesn't reject. Your biocompatibility is important. But we now know how to make. In fact, we have fda approval in the united states for biocompatible electronics, implantable electronics, biocompatible. This is really quite remarkable when you think about the things we need to understand about the human body and track things.

And course, the biggest change is why the books called the cloud revolution is in the information acquisition and information processing and information sharing infrastructure. Because you have to three features of information, always have to acquire information. Somehow. You have to store it process that you have to redistribute it back towards useful. So those three domains sort of advanced, the all events sort of independent. The telecommunications domain is a dissemination sensors, whether they're manual or automatic sensors get advanced information processing, whether it's manual, people reading, spreadsheets doing things by hand or with computers. There are three independent domains which converged into the cloud. I the cloud uses the internet, telecommunications uses computing, uses edge sensors, but to have them converge into a single sort of seamless global system, which is really what the cloud is. It's really quite remarkable. I think it's much high preserves been about cloud and ai this is one of the cases where the hype doesn't adequately capture how profound the change is.

But all these changes go through. The gardener quoted the created the phrase, the hype cycle. All these things go through hype cycles in the sense that early on when something is new, like three d printers, you see a lot of publicity about how everything's gonna be three d printed in the future.

It turns out it's a lot harder and you don't replace everything with one new technology. It would be like saying when the helicopters invented, when it was first invented, there's a lot of hype about. That was the end of cars. Everybody would have a helicopter. I it was in hindsight, that's obviously silly because helicopters have a lot of utility. But the utility is specific to certain work functions. It's always true for technology. And the new stuff when it comes along, always goes through these hype cycles. And it goes into these phases where it's out of the public mind. And I nobody's paying attention. But what's going on as engineers are actually improving and perfecting the tools.

And the machine, broadly speaking, what I tried to map out was what not only what is happening in each of these domains and how they converge. But what are the implications of it for transportation, for health care, for jobs and for manufacturing? Because that's for entertainment. Frankly, those are the things that matters. In the end. The implication of the car when it was invented, what mattered wasn't who made the car that was a big deal. Industries that didn't exist before the first 30 years of the automobile. There were 4 or 500 automobile companies in the world. There aren't 4 or 500 companies anymore, but there are lots of them making cars and trying to perfect the technology figure out what people would use. And the that's a lot of business is really interesting to investors. But what really changed the world wasn't the fact that there are now car companies. What changed the world is the fact there's lots of cars. They are very affordable for lots of people, for all kinds of services, not just automobiles, for personal transportation, with all kinds.

So when you think about these technologies and how they're changing, what really matters in the end is not the businesses that make them like drones and robots. That's really interesting. One wants to be an investor in the next intel. If you bought intel stock in 1974, and held on it for a couple of decades, you'd be a very happy guy. You bought it in 2004 and held onto it. You're not as happy, because it's at the mature phase of the business.

There's a lot of new businesses that are going to merge. It'll feel like intel of 1974. It's already the case, but some of them will go away. I you and I both know that there were dozens of chip manufacturing companies that existed at one point that don't exist anymore in computer companies from debt, digital equipment. To wang ming, wang was one of the most successful along with digital equipment. Computer companies in the 20th century, nobody's ever heard of them today. They're just gone.

So we'll go through the same cycle. There will be robot companies and drone companies will be paul, many material companies and new classes of sensors, new classes of biological materials and devices.

And the companies that really succeed, obviously, we'll make the products. It's a standard adage easiest to use, right? Lowest cost and have scalability, because some things could be relatively easy to make, but they can't scale. This is andy grove. The boss of your former company wrote terrific pieces about the scaling problem in manufacturing. You task engineers in a semiconductor company like intel. This is going on today in artificial intelligence chip companies and computing companies and robots to task an engineer build something at. You just can do some pretty amazing things like if you give them a task.

But then the task to take the thing you've designed and design it for scalability is a whole other problem, which is where the delays come. This is why we go from the hype. Somebody in invents aa stick back with the three d printer or my favorite subject is robots. You see an anthropomorphic robot. That is really quite remarkable. It does incredible things in youtube videos. That's an impressive stunt. Designing a manufacturable robot that can be made at scale. It can be reliable and made affordable it by the thousands of millions. This is a very, very different engineering challenge. That's the one you tackle after you figure out whether you can make one at all. Good.

talking-person1 12:20
thank you.

The next question is about your definition of crow revolution, accelerating or enabling all of the technology emergency is crown. We may all agree. And history is biggest infrastructure. Yeah, we seen the it is and which is itself based on the building block of next generation micros processor. And ai please show us several topic. Characters are of your defying for this crowd revolution.

talking-person2 13:04
Yeah, I think the first thing is to have a definition that may be more simplistic and meaningful of what the cloud comprises, right? It's a sort of an advice, infrastructure and advice, utility and inference utility. If we think in terms of what computing is done for most of the computing history is to do exactly what that word says, computation. You have a problem, and it has a specific answer. And there's no variable, right? Simplicity, a spreadsheet gives you an answer. It has to be right or wrong. The computing does. That's what computing does. What we really use the cloud for increasingly is not for computation, but for inference or advice.

So mapping is the easiest example of advice. Right? If you have an edge of an edge device, which has gps and obviously another location capabilities, and you ask, for advice, it's what you're asking for. If you're looking for a a plot to go drive somewhere or walk somewhere is advice on the best path. That advice is not the exact correct answer. The answer is usually based on traffic patterns. It may might even incorporate increase and whether or other features of things that are going on and gives you different options. You're not getting a computational task performed. You're getting advice. Turns out that's the easiest one to do. And the reason that happened first is because it's easiest.

If we extend that to, The obvious thing are all of us have experience with advice being given to us when we shop online. As we well know, what's going on is not computation, is looking at data, but your purchasing behavior what's available out there. Then if you like this, you might like that. All of us have seen this tool. That's advice, right? It's not a computation. It's not calculating knowing you want that for sure. And you haven't asked it to what you want for sure, advice. That's easy. The hard part is, if you're aa doctor and you have a set of data from a patient, what you really would like is what a lot of companies are trying to work on is provide ai and data driven tools that don't compute or give a specific answer to what the patients problem is.

But the diagnostic support is advice. Had you consider this? There's research on that. The the whole point of the advisory tool is that it's doing inference, not computation, and doing it in real time, which requires high speed, low latency networks. High bandwidth requires a lot of computational capability that goes beyond simply addition. It's back to inference. If you work as a manufacturer, right? What are the things you do when you're in a manufacturing business? Is you're always planning for the future. You're thinking if I at what doesn't matter what you're making, whether you're making bread or making computers, you have supply chains, labor constraints, cost constraints, all kinds of variables, all those variables you use in your planning. Right? When you're planning what to do next with your business to hire, what kinds of products to make, what kind of sourcing to have in a perfect world.

This world is arriving with, again, the cloud because of the edge network, because of sensors, because you can vacuum up information. What you'd like to have is an app, if you like, for each person in the supply chain to give them advice. I if if you choose to do this, now this is likely what will happen based on what we now know about the supply chain upstream. It impacts downstream consumer behavior. Be good to get that kind of advice in real time. Those are the kind of things we know are feasible.

In fact, there's hundreds of apps emerging giving all kinds of businesses that class, that class of advice, whether it's in the oil and gas business, or whether it's in the electric utility business, whether it's in the pharmaceutical business. What you want is inference based rich information being collected to give advice that in real time, that utility function that the clouds providing is profoundly different than just communicating purchasing decisions.

I go online to buy a book. I'm not looking for advice that facilitating that where you use in the internet with web pages was a big deal, right? It took friction out of commerce. It made life easier and more convenient. And companies like amazon and others came into being based on that model. But that sort of only one tiny part of what cloud function is. If you think about called function in research, if you're doing basic research, you have the same issue. What you'd like to have is the instrumentation tools you want to have them. Not only have I use some examples that some biological researchers are using. You want put cameras in the wild, but these are sensors. You want them motion detected, so you don't have them running all the time. You're taking pictures of wildlife behavior. You don't want the pictures happening when they don't need to happen.

It's a smart camera right up loading information in real time. And you have could have hundreds of those cameras. You could have thousands plus other kinds of sensors that kind of data collection put in real time through wireless and wired communication systems into an inference engine. That's cheap. This is the key as the inference engine has to be cheap, has to be very low cost and very fast. These are really, profoundly different ways of gathering information in giving advice to scientific researchers, whether in biology, whether in physiology or whether they're doing basic science and chemistry.

I think, as I map out of my book, all of these functions are so different from what researchers and manufacturers have had available to them, overall of history that it'll take time to adapt to them. This is no different than when electrification came along. Electrification didn't just replace belt drive. It allowed you to redesign how to factory. So it took a decade or two for factories to get redesigned around the efficacy of electric drive, as opposed to belt drives. It's gonna take a decade or more for manufacturers to re equilibrate and adapt how they train and operate around inference capabilities that are done in real time.

I and I that's already beginning. We see hints of it. A lot of the the hype in companies that people get excited about are not the companies that solve all problems. You can't do that for any business, but an app that solves a specific problem for a specific employee at a specific kind of job. But that makes their life easier, faster, safer, that those apps, those tools are all made possible by cloud functions because they don't run just in the device, right? They run through the intermediation of a cloud as a utility. And this is it's so obvious when you state it, because everyone has doing these things increasingly in their business and in their personal life.

But I think it's the level of penetration that is the percentage of functions that have really benefited from cloud based inference in real time, the percentage of activities that benefit from it so far, probably not the single digit percentages yet. Right? The number of things that we do in our lives that on a daily basis that have that benefit or let you know the map, the mapping app is one of them, the shopping app is another.

But if you go into things that are as important to all of us as our personal health, the number of really useful apps that provide inference about our diet or behaviors that are genuinely useful. That is, I don't have to do complicated things for the computer, the inference engine to tell me something.

But it works in semantically. If you like the semantic web. Those things are only now becoming possible. That's why chat gpt was such a popular hit and stunt, because it's so good at semantics. That's why it's called chat gpt right? It was semantically trained. It demonstrated how much better we could make our interface with a computer rather than alexa siri, right? Or the different. The very clunky voice recognition systems don't work that.

Well. They're pretty good. They recognize your voice, but they don't, they're very pedantic. They don't have any inference capability.

Now chat gpt is everybody's now figured out is also pretty makes a lot of mistakes pretty clunky compared to what our expectations are. But we also know that the generation cycle for chat gpt to get better now and its other competitors is now very, very fast. We're just a few years away from a really good semantic tool where I can ask useful inference questions.

A a of a cloud engine based on I what i'm involved in as a business or a person, right? This is we're close we're a few years away from it. We're not a few months away from it. I I think it's if I were picking a time, I think we're 35 years away from really a highly efficient semantic web.

talking-person1 21:56
Semantic web from as far as my memory. The very beginning was come from tim burners, lee. At the time he tried to be a sort of a meaningful web, right? Rather than just keywords searching. Actually, we also have some specific ai project in hand, maybe later ii I can show you. And you can see my background. This picture was created by open ai yeah, good.

You just keep it in, but I found a very interesting thing. The chat gpt and all other ai of robert can do only the existing knowledge. It cannot discover. The new thing, correct? That's right. If you ask them some uncertainty or no discovered knowledge, they just seriously speak some nonsense. Yeah, that is my finding thing.

talking-person2 23:16
You're right. And it has two limitations. One is it can it looks backwards because it's learning from history. It doesn't have predictive capacity, although one could write algorithms that suggest that an inference engine can run scenarios. We do this all the time with computing modeling. But this requires astonishing amounts of compute horsepower. It's already the other limitation chat. Gpt has this is not doing is learning in real time. It does it. You teach, you do the learning phase of any of these machine learning algorithms. You learn, and then you finish learning, and then you run it to do inference.

Of course, the holy grail here is that to have it be able to do learning and inference simultaneously in real time. And that the machines aren't doing to your point is that the not only are they not predictive, they can't do novel things. But what they can do is they can come up with interesting in what appears novel to us, combinations of known things like the images in your that you've generated. It's true for poetry or stories. This is very useful, right? Because it stimulates the feedback loop with the idea generator, which is the wetware of our brain. That's why I think of it a as a classic example of a tool.

I used it a piece I wrote recently a very simplistic example. When you get a really good tool, people make the mistake of thinking the tools useful for everything. I to be very simplistic. A hammer is a much better tool for hammering a nail than a stick.

talking-person1 25:01

talking-person2 25:02

talking-person1 25:04
the hammer can consider everything as meal.

talking-person2 25:08
Exactly. And you can't use it the weld metal. You have to design a different tool for that.

talking-person1 25:13

talking-person2 25:13
I chat gpt is a very specific tool. It has utility broader than a hammer, but it's very narrow. But that's not a bad thing because if my goal is to make tool that will help me generate narrative outlines for a training manual chat, gbt is very good for。

talking-person1 25:31
that the can improve enhance your and dramatically enhance your efficiency and just get you read of as sort of like a daily, boring job, but people can keep there more in imaging.

Imagine aa task and another. Ii can show you once one information articles. And that's another limited of ai we also are discussing we with the it is a scientist and mathematical science from professor from you. Universal cambridge, they approve their paper, approved magical mathematical paradox dormitory, the limited ai because the human can are usually pretty good at recognizing when they get things wrong. But chat gbt or ai never know. No chat is no.

But so this is very interesting.

talking-person2 26:48
There's an expression for that humans generally have in their field, especially good bs detectors.

talking-person1 26:56

talking-person2 26:56
we know nonsense when we see it. Right?

talking-person1 26:59
Whereas the training,

talking-person2 27:03
training, machine learning, algorithm training ai to reflexively suspect its nonsense is difficult. Could you have you as a person have to then define in advance the boundaries of what's nonsense? But you can't do that, because you don't know it's nonsense until it's produced.

talking-person1 27:24
So not green.

talking-person2 27:26
But here's the thing, though, to your ., that if you think about the efficiency benefits of the drudgery removal, because what ai is very good at right now, increasingly good is removing the drudgery and task or making easier for you to more quickly get answers that are knowable, but are not intellectually difficult to add the answer to.

This is whether it's filling out a form, right? Or whether it's calibrating an instrument. If you're you what you work a a in a fab, you're calibrating in sprints all the time. But there's very specific protocols to that. And if you haven't used that machine that day, you have to look up the man you go online for the calibration. This takes time one of the most time consuming things we had in the fab. Still true is you're always calibrating instruments, make sure they're correct for quality assurance, quality control. It make my life a lot easier if there's a semantic invoice. That's natural language.

If the cloud knows My ideal location, I'm next to that machine. I can ask how to calibrate this mean, it's gonna know what the machine is. It's gonna know the manual. It's gonna know what the machine is done. It's gonna short circuit all the things that I would spend, whatever time, an hour, half hour trying to figure out and do it in 5 minutes with me. That kind of thing is sort of a lubricant, but it didn't use the machine to discover a new chemical. I I'm the humans doing that, but anyways, you're absolutely right. It's but it does raise some interesting problems. It already has both in the art world, in particular, where the doesn't do novel art, it borrows other people's art, right? To you can literally say in the style of Matisse make a picture of me. And you will, well, Matisse isn't around to complain, but if he might consider it an intellectual property violation.

talking-person1 29:26
one of our currently research project is about try to integrate human brain, AI artificial Intelligence and human Intelligence cost.

If we cannot prove like the Cambridge professor, if we can approve AI or algorithm, is limited a paradox, so which means the human being still have its value, even in the later 3 to 5 decades.

That's now, that's my belief right now. We also find you are out. You're also very of steaming about the technology. So there will be our next question. But before that, we are very interesting for understanding more about your personal background. Cause it means why you are here today. We found you right now. You have a ropes in Manhattan in institute. A it's a research institute. Please tell us what are your main philosophy and your pursue of this for fun, fundamental technology, policy?

And also, you have experience in starting in investing is for several startup companies. So what insight do you share with us about a different role of entrepreneur, investor, and researcher in the context of crowd revolution. And it's a big question.

talking-person2 31:26
I my career in been involved in those three areas. As you point out, I the Manhattan Institute is a academic oriented think tank that looks at your educational mission is outward facing, not inward facing at students. I focus on technology policy and energy policy in my own philosophical view is that the human in inventiveness, people, humans are naturally inventors and tool creators. And that capacity is hard to predict and steer and control.

So the the greater the extent that people are given the freedom to do pursuit of ideas and allowed to fail, if you like, and that there's a whole lot of political and social baggage around how you do that. But my bias is towards more markets that are freer and open, greater trade and exchange of ideas between businesses and companies and countries.

So that we can fertilize all these innovators because they do better when they interact with other people, other businesses, and other markets. It's a Free market. It's not a Free market in the sense. It the expression of libertarian and no regulation. It's it's a Free market orientation as opposed to a Top down approach to technology development. So that's so the man has to orientation. It were my sort of my experience and head is on the investor front. I've been involved with hedge funds and private equity funds and venture funds. Currently, I'm AA co founding partner in a equity fund and venture fund focuses on investing in software. It only software, we only do software and we only do software.

Early stage companies are involved in the energy industries broadly. That is all the energy industries and in the industrial sector. Because the basic thesis we have for this fund is that all of the industrial activities in the world are under digital eyes. There's a far more digitalization and software can be applied now, but the tools are better. So most of the opportunities are in small early stage companies, entrepreneurs, and they need capital. So the key with that is that they can't get their capital just from governments, partly because most governments around the world just don't have the capacity to make those decisions. They do some of it, all governments do. It doesn't matter what country you're in. But by and large, the places that function the best and the times at a function best, to get really interesting new ideas. The money, typically there's government money involves a lot of countries, a lot of times, but the real fuel, the capital fuel, typically comes from private markets and private investors, whether the corporations themselves invest in their entrepreneurs, or whether financial equity funds like the one we are invest in the entrepreneurs.

So one of the reasons I wrote the book and one of the things that I'm convinced that based on both the research I did in my book and my work that I do is I think we're at a very, particularly important and even exciting sort of pivot in history, but there's a lot more innovation possible than most people realize. But it's not often the innovation that the politicians want. So the problem is that the disconnect between what governments want to have happened versus what people want to do themselves as innovators. There's a lot of alignment. If the Venn diagram expression, the Venn diagram is the overlap, there's overlapping interests, but there's a lot there's conflict between those. So I i've been interested in a lot of my career in studying and writing and testifying about where the roles are appropriate, where are governments when these governments involve the most useful? When is private sector spending the most useful for the best possible outcomes? And that mean useful politically in terms of getting getting the outcomes we want, which is like new science, new technologies and new capabilities.

Now, the fact that I'm optimistic that we live at a time of a profoundly exciting new technology capabilities. Does it mean that I'm not unrealistic about the challenges that the world faces politically? I we all know there's, as we speak, there's a war going on in Europe for the first time, in a long time. The Russian invasion of Ukraine. These are not good things. We have tensions to all kinds of countries in the Middle East. We've had tensions, which I hope get relaxed between China and the United States. I these are things that go on politically. I'm not naive that politicians are capable of killing innovation. They have a very hard time political leaders creating innovation. They can't do that. They can create conditions in which innovators will thrive, but they can't create innovation, but they can create conditions to kill innovation in it, but high taxation or over control.

And those things do worry me because there's an awful lot of drifting in most countries, not just in China, but the United States as well, where the political class believes they know what future innovation should be, where the time and effort should be focused. It's sometimes they're right, sometimes in at times of active crisis or war. If there's a natural disaster, obviously, the government's efforts to focus on something can be important. But by and large, I expose my bias, it's really easy to see. I don't think that the reason the governments can't pick and choose easily what the new technologies are is because we can't predict the future, obviously perfectly. And it presumes that the people in the government, the bureaucrats, politicians are, in fact, smarter, wiser, and know better about the future than everybody else in the economy.

Obviously, so as you state that we know, that's not the case. So that's why I'm sorry, Lean on the side of government restraint, but it's trying to create innovation and the role of government is to create, if you like, an environment in which it can happen. Yeah. I'm optimistic if that happens. And that's a big that's a big hope, right? In some place. But if we do that, given the state of technology development, given the state of what we're doing in computing capabilities, what we're doing in new classes of materials and BIO pharmaceuticals. Right? Then the the new classes of machines, even my favorite machine I keep coming back to is the robot in the sense that we all mean robot, we don't mean a 1 hour machine bolted down welding car parts. We mean a machine they can walk around and roll around and work with us。


talking-person1 00:00
Okay, please. Please go continue your talk about, but according to my according to my opinion, cause how to say, I don't know why the according to my opinion, because in innovation are very heavily rely on aa more free open in environment rather than the government or one central organization control everything.

We a lots of historical case already approved this.

talking-person2 00:51
And the problem we have is that the exceptions are exceptions. They don't make the rule. The so thing, things like putting a man or woman on the moon or getting to mars, these are the things that probably will require significant government involvement already do and have in the past, building aa national highway system. If you like, I there are there are very specific things that we've found. Either only the government can do or can do with some reasonable degree of of speed and efficiency. But those are the exceptions. But the majority of things that go on in in our society, most of the innovation that happens doesn't happen that way. So it's not like a aa cafeteria where you can pick and choose the outcomes you want. In some cases, you can say, I we can use it to go back in history that the fact that the government was very much involved in computing early on.

The first computers were built by governments, the colossus built by the british government during world war two, secretly and aniac by the military during world war two. But the computing age wasn't made possible because of government spending or government innovators was made possible by companies like fair child and intel. They had some government contracts, but that's not where the innovation came from, and it's not where they corporate growth came from or their ability to scale.

talking-person1 02:22
So in, I don't know why。

talking-person2 02:31
that is rain.

talking-person1 02:48
Sorry, I don't know why not testing.

talking-person2 02:53
That's all right.

talking-person1 02:54
Yeah, no problem. Let's go on. So in terms of government, the role of government, you have testified before uuuus congress of the us government many times. So in, as particularly in, you are an expert of as an expert witness on an energy economic issues.

What do you think the most urgent energy economy issues? And do you have do we have reliable solution now? But I also from another opinion, sounds like this kind of new energy issue. Sounds like a big conspiracy. I don't know. So can you give me your opinion? Well.

talking-person2 03:57
the idea that the world is now undergoing an energy transition, that phrase it's been used to completely eliminate the use of hydrocarbons, oil, gas, and coal. I've been pretty vocal in the public record that that I it's the data show that's not happening. The world is not abandoning oil, gas, and coal, because there are very, economically efficient ways to operate civilization. And i've also been on record of saying that we're not going to abandon oil, gas, and coal in the coming decades, because we don't have technologies capable of matching the matching the capabilities in either economic or scale terms.

But that's not the same as saying that there won't be lots more windmills and solar rays and electric cars. In the future, there are and there will be. What we have is a pattern in history that's pretty clear and predictive new sources of power and energy are typically in through all history, additive that we add to old sources.

In fact, a fact that I like to point out that most people aren't aware of is that after 20 years and globally and something of the order of 5 or $10 trillion of spending that is in the europe, united states, and china, and wind and solar, as of right now, burning wood provides 3 times more energy to the world than all the world's wind and solar systems combined.

Burning wood, the oldest source of energy on demand is still 3 times bigger contributor to global energy supply than all of the world's wind and solar arrays combined. The point of that, we're not gonna build more windmills and solar rays and more electric cars. It said it's a very big system. It's very slow to make significant impacts on a system as big as a global energy system. The addition of more wind and solar is not causing oil, gas and coal use to go away. It's is added, additive. It's slowing the growth rate of oil, gas, and coal, but it's not replacing them, which again, is the pattern of all of history, which is what I think will continue.

And I I think the data show that's exactly what will happen, whether people think that's good or bad. I obviously, the reason the energy transition has been put forward is to avoid hydrocarbons. We don't burns. We don't emit carbon dioxide because of the climate change concerns, but that framing is irrelevant to what's actually possible.

So most of my work in this area focuses again on technology forecasting. I it's not that I don't believe there aren't new technologies for how we use and produce energy. It's that the time it takes this to get the scale, it's just like, again, if i'm going back to the analogy that would be good for people who are familiar with computers, who believe that we can replace our energy system would be the equivalent of saying with the technologies that we knew in 1974 for making computers, microprocessors, we couldn't possibly build a modern data center. I don't know how much money you threw at it. You couldn't make a kind of data centers that exist today. Could not have been built in 1980, could development with any amount of money.

And if you tried to the amount of money, you would spend to emulate today's data center, would bankrupt the economy would have if you try to make computer that produced a peta byte capability with computing capabilities. In the 1980s, it would have cost trillions of dollars. The computer literally would have cost trillions of dollars and would have not worked very well. So what we're trying to do today that a lot of politicians and analysts are trying to climb that we can take of today's efficient oil, gas and coal energy system and replace it with what is inherently less efficient. Technologies, wind, solar technologies are inherently less economically efficient and less energetically efficient.

That's in a lot of different ways, different forms. When I test the light of this, you find people skeptical and saying, you you're obviously not a technology office. You don't realize these technologies get better. I I beg to differ. I do the issue with all technologies. Again, it's just as it was with the computers at the early age, is the time for scaling and scaling time to society levels to operate all of civilization is very different than the scaling time. To, let's use an example, replace telephone communications with from wires to wireless wires, the wireless fall. It was a big job. It took a couple of decades to do that. But that's just phoning. I the telephone activity of civilization doesn't even comprise a few percentage points of the whole economy.

talking-person1 09:04
Good. It's very interesting observation and your insight.

And also, I as your self mention in your book, you are technical optimism and believer in the innovator and policy makers. We also wants talk with the founder of singularity university. The founder, president named peter diamond is he's also a technology are optimist based on his belief that human being can harness technological evolution to advanced human wellbeing in terms of your role of policymaker and technology inventor.

So how how about your belief come from? I how your all optimistic this feeling and we all know it's very, very important, even much more important than then the specific technology. No.

talking-person2 10:23
it's I guess party you can say it comes from looking at history and what's happened so far? I it's indisputable that two things are true, right? We haven't gotten rid of human suffering everywhere. We haven't stopped wars. We haven't stopped our ability to disagree with each other and do silly and bad things, right? Human. We're human beings, right? We're not automatons. So humans have those capacities. At the same time. It's indisputable that the human condition for billions of people is far better than it's ever been in history. We have improved the quality of life, not just the length of life, but the quality of life for billions of people because of innovation, because of technology. Into the extent that human beings get along, we are fighting and trying to kill each other in wars. We've made great strides. It's obvious, right? That you don't have to guess that.

So then to not to be a pessimist, you'd have to believe one of two things. You have to believe that we're gonna have only wars in the future that wars won't be episodic to be permanent. I there's no evidence in history that we are permanently at war. We episodic fight. So I don't see the destruction of the human race coming. It's only in science fiction novels. It's kind of fun. The other thing you have to believe to be a pessimist is that there's no new technology that we have invented. Everything is ever to be invented, because all of the improvements that have occurred, the ability to protect ourselves, from nature, protect ourselves, from disease, but protect ourselves from the natural disasters, to recover from disasters, to make people healthier, to make life more comfortable.

All those come from technology, right? Technology is a driver of productivity and wealth. It's what creates the entertainments that we have in the world and our comforts. And tourism is all from technology. So people who our optimists are essentially believing that all those changes, all those improvements are done. We've invented. Everything is possible to invent. And so going forward, all we're gonna face is more of the same, which means no improvements to believe that it takes. First, it's not possible to believe that that we have invented everything that's possible that we've discovered everything that's possible. We all know that we don't understand human biology. Well enough, right? We were always making discoveries about diseases and how to cure diseases and how to solve these problems. There's a lot to discover, there's a lot to invent. In fact, the tools that we have for discovery invention are much better than they've ever been in history. So I wouldn't say i'm an optimist. I'm saying, i'm a realist. I'm saying the pessimists believe things that aren't real, which is the innovation has stopped. And since innovation brings improvement to the human condition, even when we are annoyingly fighting each other, so brings.

I i'm saying if you're a realist, that is by definition, what we're now calling an optimist because there is plenty of things to innovate, to learn. And those lead to a better human condition. It gives us the opportunity to, let's say, we can have fewer things to fight over. Is this more wealth? There might be fewer things to fight over, will still find things to fight over. People always do, but that the framing is what's odd, right? The optimists are not pollyanna, the expression that they think everything's gonna be perfect in the future. There's no perfection, but optimists believe, I think, is that progress hasn't finished. We are not at the end of science or the end of technology. If you know that technology brings you these benefits which it has for centuries, then the future is gonna be better than the past. This is just realism, but we label an optimism, because I think there's sort of aa reflexive preoccupation with the problems here.

Here's a good example. The correct preoccupation of the problems is an expression. No one would deny that the invention of the automobile, but has been a freedom generating machine for the world. It's a perfect people buy them because of their social democrat. They're free. The freedoms that gives you right to how you live and what you do. But when you invent the automobile, you've invented the automobile accident as well. By definition, the negative consequences on a model bill sticking just to that. And you've invented the accident. You've also invented the environmental impacts from the upstream things you have to do the manufacture, automobile, mining, and manufacturing. It's not that there's no negative effects from new technologies. Obviously, there are, but again, you use technology to minimize and ameliorate the negative effects.

So it's a long way of saying the reason that I evolved into an optimist is because when you study the history of technology and the state of technology today as a realist, it leads you to conclude that we've only just begun to solve difficult problems and to invent new things.

talking-person1 15:40
Yeah, good. It is a very good, realistic answer. I like it. So come to our last question also about aai we may all agree today is the the age, all time of AI according to your very rich and strong background and expertise, can you tell us what is in the key roles of entrepreneur, research in investor and their relations in the con context of this aiai time how this kind of different people, but in in personally, in your background, you are very good integrity, a different role in one.

So for many other people, they do specific different roles, how to integrate or how they collaborate with each other.

talking-person2 16:52
You just defined the three elements of how we end up with the things that always look like magic to us in the present.

But when they arrive flying in airplanes at the speed of sound, hundreds of people pretty magical, right? I so that expression that are Arthur c Clark created this famous science fiction writer, any sufficiently advanced technology will look like magic. Right? So the magic comes from those three domains of making new technologies and products and services. The researcher, right? The entrepreneur and the financier, the money and you have to have all three. And there are people are particularly good. And usually it's only at one of the three people are really good at understanding when something should be finance and how to finance. Investors may not be. They couldn't repair their own car or build one to explain how it works. They're not they're not an engineer necessarily.

Then some of them might be. There's obviously a relationship between those things. You need basic understanding of how things work. In order to make the things work better. That's research, by definition, this kind of research doesn't have to be research into the origins of the universe. But the research goes across a very broad spectrum, but research usually means answering questions like, why does that work?

So that's the focus of research. You need to know that why did that break? That's usually a research to ask the why question the entrepreneur does something different. They answered a how question? How can I build something from that? How can I make that? How can I make it? How do I build it? So that's a different function. And it's a different kind of skill set, and it's a different kind of business. They overlap each other. There's a relationship. How person often talks to the white person as they're trying to figure out a building? Why did it break in the researcher? It can be the same person. You're doing research to figure out why something broke in order to make it better. But usually a there's a different team that helps you. Because in the world, the real world, the currency that makes everything possible is currency, money. Money is the way which we not only invest in researchers and pay their salaries, but building any Enterprise requires capital.

And the people who are involved in the capital domains need to know enough about the other two to know when they're making good bets. I whether you're loaning somebody money. It's a factory. You're qualifying the factory, whether or not they are a good bet. That's very, very different and much more rigid and easier task than if you're an investor trying to invest in an entrepreneur. Do you really believe they can build a business? Very different metrics are applied, but you're doing the basically the same thing. You're trying to gauge risk, putting capital to work.

But so that's the, if you like, the holy Trinity of how technology has always happened, people people figure things out. They want to solve a problem. It may not be an obvious problem. Anybody else? They have to figure out how to build a machine in a company around it. They always need capital. It's because they're very fluid things. If you think it's obvious that those things all are interact and are fluid, this is back to something else we talked about earlier. This is why it's so difficult for command and control system to be effective. On average, at making that system work well. It's very dynamic. It's in flux. It's not fixed. The boundaries aren't fixed. The skill sets aren't fixed in even issues with respect to do you take those risks at all? Governments tend to be low risk enterprises. Much, much innovation is inherently high risk. If you want really profound innovation, you have to create a system in which people are willing and able to take those risks and try things and fail.

I sort of bring me back full circle of where we started that, what the freer the system of which those three domains can function, the more likely we all have good outcomes. But they are guaranteed, they may not be outcomes that will occur in time frames that politicians and policymakers like. But we do know to go to come back to my other point. Do know that over time. We figure stuff out. We build new things to believe that we're not gonna discover new things in research and finance, new entrepreneurs to build profoundly different kinds of businesses and services. That's not gonna happen, I think, not just pessimistic. It's profoundly naive. It's the opposite of of what really happens in the world.

talking-person1 21:43
Very good. And thanks again for your wonderful sharing and your insights.


每日关注 更多
林永青 的日志归档