Dialogue with "Steve Jobs", representatives of Silicon Valley star venture capital talks in detail about the potential fields of AI

Author: New Newin

Editor: Tencent News Hao Boyang

AI is changing the way we live, work and play, from driverless cars and voice assistants to AI-assisted medical diagnosis, financial forecasting and ChatGPT, how will it change our social, political and economic life, across various industries, industries and geographic fields?

Earlier this month, the Milken Institute invited star representatives from the Silicon Valley venture capital community and CNBC hosts to launch a global dialogue on the next generation of artificial intelligence. The guests included:

Elad Gil, serial entrepreneur and investor, co-founder of Color Health;

Ashton Kutcher, "Jobs" actor, investor of OpenAI, Anthropic and StabilityAI; (the picture below is a still photo)

Peter Lee, Chairman, WorldQuant Predictive;

Nicola Mendelsohn, Head of Global Business Group, Meta;

Alexander Wang, CEO and founder of Scale AI;

Each panelist shared in detail what they see as the biggest opportunity and areas of greatest interest right now.

Draw key points

①Among the fields where AI may bring about changes, medical care, education, and law will be the fields with the most potential to benefit in the near future. Automation will greatly reduce the cost of use in these fields, and create a personalized experience to realize social Deep fair.

② One of the topics that is really ignored or bypassed in the real-life discussion is about the use of advanced analysis technology to help existing business processes. Overwhelmed by data, executives want to be able to predict outcomes, have data-driven decision support to simulate different scenarios, and ultimately optimize decisions to drive revenue growth, reduce costs, and manage risk. This is the area where executives place the highest priority today.

③ At the level of Transformer basic model construction, the winners will be those companies that can obtain low-cost energy, high-quality computing power at low cost, cloud storage and extraordinary talents.

④In this technological revolution, 80% of the benefits will go to existing giants, and 20% to start-ups. But because there is so much value creation here, these startups have huge potential.

**What new changes does AI bring? **

Julia Boorstin:

Elad, what are you most excited about right now? Is it the health field? Or is there a different enterprise tool involved?

Elad Gil:

In general, I've been involved in the tech industry for about 20 years, and I've worked on AI at Google and Twitter. I've founded multiple companies and invested in many past breakthrough companies such as Airbnb and Stripe**. This is probably the first time in the past 10 to 20 years that there has been such a huge platform shift, everything is changing. Every time this shift occurs, the underlying technology changes. **

**We made two breakthroughs in 2015 and 2017, introducing new models that allowed us to do many exciting new things, with new user interfaces, and new application scenarios that changed. **You will see application scenarios like law, where for the first time you can upload a 30-page litigation document and have the AI output a 20-page response with cited legal cases. Companies like Harvey are doing that right now, and you're seeing this massive shift where AI is now able to beat doctors, pass medical exams, bar exams, etc.

So it's been claimed that the people who designed these AIs are constantly putting pressure on them to be doctors and lawyers, when they just want to be an artist like Midjourney. So that's all changing. I think every field, at least every aspect of the field that the technology touches is changing, although that sounds a bit hyperbolic because 10 years ago, people said the same thing about AI, but the technology wasn't there yet, That's why you don't see many startups flourishing. **And now, I feel like we've finally hit a tipping point in technology where it's good enough to change all those things. **

Julia Boorstin:

So there are multiple areas involved. Ashton, you just launched your fund today investing in OpenAI, Stability and Anthropic. What are you most excited about right now?

Ashton Kutcher:

Oh, I guess I mostly just wanted to reiterate your thoughts. I think we've got these Transformer models, and they're starting to get good enough to be very effective for people. So, **many people in the past have thought of AI as an alien object that affects you. And what we're finding now is that it's a tool that people can take advantage of. **I think this is a very beautiful thing, over time, these Transformer models will continue to be strengthened by data and get better and better.

Many cloud computing companies are associating themselves with these Transformer models because they realize that these models will generate a lot of data, thereby increasing the overall market size of the cloud market. Then we'll have application layers plugged in and fine-tuning those models. **If we look back at the last wave of software from the world wide web to the mobile internet, it was largely a reinvention and disruption of the unskilled labor market, whereas what we are about to see is a reinvention of the skilled labor market . **

**What is really exciting is the opportunity this opportunity brings to the masses. **Taking law as an example, if you had a company in the past and you wanted to reinvent a software company in the legal space, you might need to hire 1,000 paralegals to serve your community, whereas most of you here today probably need Lawyers can be found when lawyers are available. But most citizens can't because they can't afford it. **And when you have a paralegal capable of managing 1,000 AI agents, it dramatically lowers the barrier to entry for everyone in the field. **

**So I think AI is a fair and inclusive development, a huge opportunity like having access to pediatricians, doctors, lawyers and personalized tutors today, all of which will now be possible. ** These applications will be built on a foundation of fairness and inclusivity, black cars are no longer just for the rich and famous; doctors, lawyers, educators, etc. The same goes for easily accessible careers.

Julia Boorstin:

It is worth noting that although this fund is your top three investments, your investments are mainly at the fundamental level rather than use cases. Do you see opportunities to invest in other areas as well?

Ashton Kutcher:

Yes, **I think under the tuning model for each specific application case, there will be these large Transformer models, because it is difficult to surpass the output of these models. **For example, if you look at Claude 2.0 or GPT-4, their output is excellent. It's much easier to build at the application level, just concatenate it with an existing Transformer model, because it's more cost-effective, and then fine-tune your own model to your specific dataset and vertical.

So, **I think on this basis, there will be some large Transformer models, some of which are multimodal, some of which are specific, and they provide support for the application layer. **Interestingly, there may be regional differences at the application level, since practicing medicine in California requires a different license than practicing medicine in some other states. Because of the small differences, there may be geography-based firms that cater to legal work in different regions. Therefore, there will be some specific models adjusted for specific regulatory environments. Based on these Transformer models, different values can be generated in different applications.

**Which areas will generate new business models? **

Julia Boorstin:

Peter, you have a lot of experience in the corporate space and in finance. What are you most concerned about right now?

Peter Lee:

I think there's some risk right now, and while this generative AI boom is exciting, as Gachan Li mentioned, I think there's clearly some transformative opportunity as well. **But I think one of the topics that is really being ignored or bypassed in the real world discussion is about using advanced analytics to help existing business processes. **

I thought it was interesting, and before I came to this discussion group, I did an interesting thing, if you take the Bing chat feature (which is powered by GPT-4, of course) and ask it about industrial objects Networked generative AI use case, it returns zero. Yet we have billions of connected devices. We are faced with real problems in the field of manufacturing, such as providing high quality products at important prices and so on.

**I think AI can be applied to these types of use cases, which are really the problems executives face on a daily basis. The executives we talked to were so caught up in the data, they were literally overwhelmed by the data. They want to be able to predict outcomes, have data-driven decision support to simulate different scenarios, and ultimately optimize decisions to drive revenue growth, reduce costs, and manage risk. So I think there's a real opportunity for AI to bring the discussion into existing business processes, areas that executives value most today. **

Julia Boorstin:

Nicola, how do you think about applying AI as a tool to your customers, the big brands that spend money on Meta?

Nicola Mendelsohn:

We joked before that we should actually rename this conference the AI Conference because whether you're involved in the future of finance, the future of health, or the future of philanthropy, AI really has a lot going for it. But if you think back to the early days of Meta, back in 2006, when we created the newsfeed, machine learning and AI played a big part in making sure that users saw what they wanted to see from their friends and family content.

In response to your specific question about how we leverage AI today, last August, we did continue to provide users with a service where they see what they want to see while filtering out what they don't want to see , all through AI and machine learning. For advertisers, we have a whole suite of products that we call the Advantage Plus product. What it does is offload many of the human-processing tasks to AI to better predict our target audience and find new ones so advertisers can reach them more effectively.

The deployment of AI is also interesting in terms of metrics. We're now starting to experiment with some adjustments in brightness and darkness, etc., which may sound a little more low-key than some of the larger use cases you guys just talked about, but it's important for advertisers to reach With the right audience, these tweaks can really work. So we're very excited about the whole field. As Mark shared on the recent earnings call, we see a lot of potential across all products as well, including the deployment of generative AI.

Julia Boorstin：

Alexander, what are you most concerned about right now? Scale provides a wide variety of tools.

Alexander Wang：

We do see developments across the board, let's take a look at each industry. The first is the technology industry, which may have one of the most significant transformations on the horizon. If you think about the balance of power among the tech giants, it will fundamentally change because of this technology. Chatbots are catching up as a new form of consumer interface. People like to spend a lot of time on it.

**As we look ahead, in what is to come this year, there will be a battle between the tech giants about who can offer the most compelling product. This may be one of the few moments this year when the big tech companies are actually fighting. We are still in the early stages of commercialization of generative AI, and there is no multi-billion dollar revenue stream related to generative AI outside of the possible multi-billion dollar revenue stream related to the chip business in video. So, I think over the course of the year we'll see that evolve. **

**I think it will be very interesting to watch the commercialization potential take shape, whether through advertising or otherwise. For the enterprise market, we see a range of use cases, as Ashton and I mentioned earlier, the professional services industry is fundamentally disrupted. You can see a lot of professional services firms, like PwC, that are investing heavily because they feel an existential threat. **They announced a $1 billion investment in AI and generative AI last week.

As we've discussed, the legal profession will be fundamentally changed by this technology. And then we look at every business function, whether it's advertising, as we show you in our San Francisco office, or sales, or customer service, every function is fundamentally changing the economic model.

**I've personally spent a lot of time lately focusing on the geopolitical implications of AI. As you mentioned in a previous discussion, as we saw in World War II, countries that can quickly integrate new technologies into their warfare and defense intelligence processes will eventually come out on top, and AI is a capable Technology that changes the balance of power in diplomacy. **

AI regulation and corporate competition

Julia Boorstin：

Yeah, I want to dig a little deeper into this and make sure we get back to talking about the competition between the incumbents and the upstarts and the shifting of power. But I think it's also important to delve into the idea of this new-age space race. Elad, what are your thoughts on this?

Elad Gil:

**I think a lot of the recent calls for regulation of AI feel very one-sided, not really hearing voices against it, everyone seems to be calling for more regulation of AI, which seems premature. **

In some respects, there are indeed some things that should be regulated, such as export controls on advanced chip technology, as Alexander said, China may benefit from it, or other potential opponents. However, there are many things that are mixed up, and someone is concerned about the safety of AI, that is, will the robot offend you? Or say something you don't like? Or spread misinformation? Is there still a risk of someone misusing or misusing this technology? Could you hack a train and derail it if it was controlled by a computer system? Or can you do other things with AI? Finally, people worry about whether AI will become a species that competes with us and so on.

Obviously, people are conflating these issues, but they all lead to calls for more regulation and slower development. Elon Musk and others published an open letter recommending that we put AI development on hold for at least six months and possibly longer so that I can catch up and build my own technology. Some of this may be for self-interest. But I think it's too early to do that. We still have to see how the technology actually works, and in many ways it's still being proven.

If we look at the history, there are a lot of tools, as you said in the previous discussion, they were misused. **But if we look at the entire history of the 20th century, the history of the 20th century has been a history of people doing horrible things to other people. We don't need AI to cause two world wars, to commit genocide, or to cause mass famine. **Thus, I think our reading of the relative risks and benefits of this issue is wrong. I think it's premature to call for more regulation.

Ashton Kutcher:

I would like to add one more point, just to be clear. We haven't reached general AI yet. General AI is definitely something that needs a lot of regulation and consideration, but we're not there yet, so that's the number one priority. Plus, three of the companies I know take these issues very seriously, and they think ahead about the secondary and tertiary consequences of the technology they're building in a way I haven't seen before.

**In fact, these companies are more involved in this problem than some of the big tech companies, and they are more forward-looking and willing to apply their technology to solve problems, rather than cover up problems that exist on their platforms. **They are actively addressing these issues, actively thinking about how these technologies can be used in all kinds of bad ways, and mitigating the impact of these problems by building models themselves. In fact, they may lead the way for regulation because they will suggest best practice in this field, which is what we should do.

China's current state-of-the-art model has about one trillion parameters, while the current state-of-the-art model in the United States is GPT-4, which has about 800 billion parameters, but there are about 2,700 characters in Chinese, and only 28 characters in English , so fundamentally the GPT-4 model may be more powerful.

Looking at the bigger picture, for the model to output unpredictable things, you need to keep some nodes in the neural model, otherwise that will reduce the quality of the model output. For now, models such as GPT-4 have advantages. If these models and companies are too restricted, we will be at a competitive disadvantage.

Elad Gil:

**On the question of regulation, I think there is a pivotal moment coming, and that is the upcoming presidential election. This will be the first AI-driven election. **We will be using generative AI to create ad copy, robo-dialing and perform various tasks. At that point, the dynamics to regulate will become very strong, and we may need to resist that dynamic. It's going to be the first AI election, and I think it's going to be very dramatic, it's going to have a lot of impact.

Peter Lee:

**From a business perspective, I think AI has huge benefits for organizations. Especially post-COVID, companies are going through digital transformation, when you really think about your organization, there is an insatiable demand for applications that can leverage embedded AI or embedded analytics, how many Do people in the organization really have the ability to understand the full life cycle of these AI models, including development, deployment and maintenance, and how to deal with issues such as bias and discrimination? **

**So, I think that in the digital transformation process of the company, it is necessary to cooperate with a trusted partner. **Over time, this feature will be highlighted, whether it is a company like scale AI or WorldCom, predictive, etc., this is a very important feature in the upcoming AI revolution.

Nicola Mendelsohn:

Looking back at history, whenever any new technology comes out, from the printing press, to electricity, to the automobile, we always go through a curve, from great excitement to great fear, and then enter a phase of normalization where we move on. What's different this time around is that we're going through both excitement and fear at the same time, but I'm very encouraged by the sense of responsibility that many companies working in this space are showing in the build process.

From a macro perspective, we have an entire responsible AI department, and we've worked with different regulators, civil society, both internally and externally, to co-create some tools, such as a tool called an "open loop" that allows Policy makers and regulators test and try out the ideas they come up with in real time, as we find that for new technologies, regulation lags because of a lack of expertise.

So, by creating a product like this early, from the very first moment, I think it presents a good opportunity to be applied in many areas, from elections to everyday use of these products. I'm therefore encouraged by the seriousness with which the company is taking this issue.

Ashton Kutcher:

**I think one of the most important considerations as this technology rolls out will be rebalancing the privacy and security perspectives. Because when you have something with this much power, we as a group might be willing to give up some privacy to be safe. I think this will probably be one of the most important considerations over the next few years, how much privacy should we be willing to give up to keep us safe? Because there's no question that people will do terrible things with this technology. The question is can we find it? Can we stop it in time? And use the same technique to find it and stop it? Do we have the necessary gating mechanisms to do this?

Nicola Mendelsohn:

Yes, I totally agree. One of the things we're doing is finding deepfake videos, analyzing them, understanding them, feeding them into a machine, and letting the machine train itself to be able to understand and recognize other content that uses the same technology. So I do think it's an area that needs to be thought out,** an area where we need targeted regulation, so that we don't get hamstrung and not making really significant progress in curing diseases and dealing with other health conditions. **

Julia Boorstin:

Peter, you mentioned the potential for AI adoption in many different industries, and I'm curious if you could tell us what the barriers are to driving more enterprise AI adoption?

Peter Lee:

Yes, I think there are several different types of obstacles. For high-value use cases, teamwork is critical. Data science is a team sport,** a company like Telecom Italia won't automate the incentives for its 40 million mobile subscribers if a model gives a recommendation but doesn't have some form of governance lifecycle. **

I think that high-value use cases that require collaboration, communication, and buy-in from a representative group are a real barrier today. For ChatGPT, which is very exciting, we've had a lot of executives say to us that they really want to leverage advanced analytics, but our data isn't ready, not there yet.

I think ChatGPT demonstrates a tremendous ability to derive competitive and predictive insights from data around the world, for which we have spent a lot of time sifting through and monitoring public, private, commercial and derived alternative datasets.

Revolutions in education, healthcare, law, and business decision making

Alexander Wang:

When I meet with CEOs around the world, I ask them, have you tried this technology? I would advise them to try, to experiment, to learn, to arouse your curiosity, and then you will see the opportunities that lie before you, and you will be able to make your business more efficient, more effective, more productive, and gain greater achievement.

Ashton Kutcher:

**If you are a commercial company and ignore e-commerce, then you are probably out of business. If you're a company and you turn a blind eye to AI technology, you could go out of business, it's so good and powerful from an exploitation standpoint. **

**When we talk about commitment and potential, it often starts with education. Just in the field of education, having a personalized tutor who is always with you is a very advantageous position, and these large language models are best at analogies and metaphors. So if you're looking for an analogy to explain homomorphic encryption, or an analogy to explain privacy-preserving machine learning, you can take an analogy on a topic for which you already have domain expertise, and apply it to the What to learn. We're going to have students, kids and people from all over the world able to use this technology to learn about a field in a way that hasn't been done before. **

We've all heard about personalized medicine, personalized law, and personalized education, and we're going to personalize every personalized media experience, and you'll be able to take that song and make it your own. My friend co-authored a book with ChatGPT on the basis of his past six books, and he also asked what will be the next book in AI?

So if you're willing to wait a week, I'm going to make you a personalized version of what we've had in the book and that's the content experience we're going to have and that's the consumer experience we're going to have , which is very promising for humans.

Julia Boorstin:

Alexander, now that you're talking to enterprise customers, what do you think?

Alexander Wang:

Yes, I think in terms of the greatest growth in human potential, that is education. Most people in the world do not learn through one-on-one tutoring, but through classroom learning, but most people do classroom learning because it is affordable. Now that we have one-on-one mentorship for everyone on the planet, this will have a huge impact on human potential for decades to come, which is very exciting.

**The really big use cases we're seeing are businesses fundamentally changing the way they interact with customers, whether it's an e-commerce company, they're thinking: How do you get everyone to have a sales assistant that's basically AI driven , to help you find what you need to buy and help you find the perfect gift for your mom, or what to buy for that BBQ party you're hosting. **

I think this is a completely different way of delivering products and services. The same analogy exists in the medical industry, how to provide a more personalized and tailored experience. Technology companies are also thinking about this issue. For example, Apple is thinking very deeply. Whether the way we interact with our phones now is fundamentally different because of this technology. Therefore, **I think all business leaders should think carefully about what the ideal customer experience will look like in the future AI era and how to achieve this goal. **

Elad Gil:

I think Color has done some related work in the field of population health care,** from a health care and cost reduction standpoint, you can cover a part of the job of a doctor or a nurse through AI, so that the cost can be greatly reduced; if you can use A 10-fold reduction in the cost of certain aspects of medical care expands the pool of people who can receive medical care and achieves a profound level of equity. **

**Therefore, care delivery is one of the areas that is about to change. In the long run, the 5-year version is that anyone anywhere in the world can get medical care on the level of Stanford Medical School or UCLA through their mobile phones. **You can take a picture and upload some text description and you will get some ideas or hypotheses no matter where you are in the world or what you can or can't afford in terms of diagnosis. We've seen some adoption of this in the U.S., lowering cost and increasing availability, but I think that's only going to accelerate, and what's really exciting is the global adoption.

Ashton Kutcher:

I think one of the biggest questions is in the application of AI in the enterprise, it is not "what is AI?" but "what is human?" and become unique.

Julia Boorstin:

This will change education and what kids need to learn and think about, and the relevance of philosophy and ethics and more. I want to go back to the question of who will win from these AI opportunities, Apple, Google, Meta, Microsoft, and OpenAI, but the question is not just which of these tech giants will win, but young people like Scale Can the company, or other companies you invest in, stand out in this new competitive ecosystem?

**Large companies and start-up companies, who is rising? **

Elad Gil:

I think everyone's still sorting this out, which is why it's an exciting time to be in this market. Many believe that the underlying model layer will be consolidated into a handful of companies, often in partnership with a cloud service provider. So you have Microsoft working with OpenAI, you have Google with both cloud services and models, and then you have Anthropic as another key player, and some emerging companies like Scale, CoHere, and 21.

For now, they appear to be one of the potential leaders in the field. As you go up the stack, you see some tool companies, Scale provides not only tooling, but also various other functions, and they have become a very important part of the ecosystem. There are also new things like LangChain or Llama Index, which people are more and more interested in.

Application level, I remember before mobile, people were asking, “What’s going to be Salesforce for mobile? What’s the new CRM system? Or what’s the search engine for mobile? Turns out, Salesforce for mobile is Salesforce for the iPhone, and search for mobile is Google for Android.

**An important question is how much of the value goes to incumbents vs. start-ups. If I had to guess, I'd say 80% goes to incumbents and 20% goes to start-ups, but with so much value creation here, these start-ups have huge potential. **

So one of the really exciting things about this era is that the incumbents are going to reap huge benefits and of course you're going to see a whole range of exciting new use cases and experiences like tutoring your kids And healthcare at a wider society level, which was previously provided by fledgling start-ups.

Peter Lee:

**I think we should also direct some of the audience's attention to the data providers, who will be winners too. **I do think that for many companies, there is an opportunity because they have a huge amount of untapped data. Not just from a company perspective, if we use data as a lens, not just companies, those companies that are closely connected to data and can interpret data to help customers understand how to use data will be the winners in this space.

Ashton Kutcher:

**At the Transformer base model building level, the winners will be those companies that have access to low-cost energy, high-quality computing power at low cost, cloud storage, and extraordinary talent. **These basically will determine everything. There are many alliances being formed right now, changing daily, and these alliances are good leading indicators of where the value will accrue.

** On the application level, I think similar to the mobile revolution, I agree a lot of the value will come with the incumbents, but the winning companies will be the ones that have a great product and understand and value the human experience, understand human pain and The company of what frictions exist in the journey to happiness. **Companies that understand this and intuitively build products to solve these problems, and do so in a cost-effective manner, will be the winners.

Nicola Mendelsohn:

I think the masses will be the winners, and the little girl who was told she wasn't good at math, the little boy who was told she wasn't good at creative work, will no longer be the case in the future. When we solve some problems, we as a society will have to rethink that people will no longer need to do jobs that did not satisfy them before and provide more opportunities for innovation and creativity. For me, it's very exciting.