In 134 days, users broke through one million, a breakthrough for an AI entrepreneur.

Editor’s note: This article is from WeChat WeChat official account, with a new in (ID: new in data), by Cang Yang and Half Buddha, and is published by Entrepreneurship.

Tome is an AI native story-telling platform. It is hoped that users can generate a complete visual display in a few seconds through simple prompts, similar generative AI demonstration products, and Gamma and AI enhanced document product Notion.

Last week, Tome officially announced that its user scale exceeded 10 million, which was only about six months before Tome announced that it had exceeded 1 million in January, and in February, it received a series B financing of 43 million US dollars led by Lightspeed. Other investors include Coatue, Greylock(Reid Hoffman is a member of Tome’s board of directors), Emad Mostaque, CEO of Stability AI, and Eric Schmidt, former CEO of Google.

Keith Peiris is co-founder and CEO of Tome. Prior to this, Keith worked as a product leader at Instagram, Facebook and Citizen.

The content of this issue comes from Keith’s online sharing, covering the personal experience of transforming from product manager to entrepreneur, the conception and development of early products, and the thinking on growth, financing, commercialization and AI. The full text is as follows:

• From product manager to founder

• The original intention and advantages of establishing Tome

• Iteration from Idea to MVP

• Vision? How to guide product iteration

• What is Tome’s competitive advantage?

• 134 days, 1 million growth course

• Thoughts on financing sharing and commercialization

• Team size and portrait of members

• Financial and business planning

• Can Tome train big models?

• AI is not a substitute but an enhancement.

Keith Peiris

Many technical schools hold robot competitions every year. At first, the competition was about a large robot that can play football, just like a human-sized robot, and then it became a smaller robot, and then it became smaller. Then one year, they thought, we should hold a football match for a robot, which is only as thick as human hair. At that time, I was studying nano physics. That sounds great.

I’ve always wanted to build a robot, and I only learn things on this very small scale. I want to build a very small robot. It can play football. They can grab this small silicone disk and move it. In fact, in this process, I learned a lot about fundraising. Because when I want to start this project, the school doesn’t want to fund it. They say your grades are not good enough. We don’t think you have a good plan. So I have to find sponsorship from different companies, which are trying to hire computer science students from Waterloo. So I spent several years studying it.

In our first year, we did badly, but in the third year, we finally won the game. Then, in the year after I left, the team broke a series of records about this tiny robot that can play football. So this is a very interesting entrepreneurial lesson.

Grace Gong

So how are nanobots related to your successful products? You learned some cool lessons in that era, which helped you start your product career. Obviously, you also worked in a large technology company like Facebook. You know, any small change of yours will affect millions of people. And now you have built something from scratch. What have you learned from your early experience that has influenced you?

Keith Peiris

There are several things. First of all, I can recruit these very smart students to work for this robot. The first thing I learned is that when you work with truly talented people, you can’t bind them. You must give them some space to think, and you must give them space to experiment and play. I think for people with motivation and talent, if you give them autonomy and goals, you can inspire real creativity. However, whenever I try to interfere with their work excessively, or I try to be a little radical, the result is always worse. So they learned something about management from it very early.

Now, the second thing I learned is that when you are building a very, very tiny robot, or anything tiny, you can’t handle too much complexity. Because if you make a small mistake, you must throw away what you have done and start over. My first version was very complicated. We can’t even test this idea because we can’t build this thing. This is what I learned compulsively from it. Do simple things first. When I joined Instagram, I believed that it was valuable, and it was also applicable to products. It was easy to come up with some complicated things, but it needed a lot of hard work to make it simple and simplified. This is an important experience that I brought from the physical world to the digital world and products.

The last thing is, just like starting a business, I think the CEO’s job is to accept multiple rejections every day, and find a way for you to recruit people, investors and customers who don’t like your products, and continue to do the same thing after getting up the next morning. In our example, we built 40 or 50 different versions, and one of them was really successful. So this is an accelerated study of the value of perseverance.

Grace Gong

These are great lessons. I could hardly believe it when you said that you were rejected. Because from the product point of view, it looks very beautiful, and look at the investors. You really have Weig. You have Dan Rose, one of the founders of LinkedIn, or Eric Schmidt. They are all top operators and transformation investors. Out of curiosity, who are in your personal advisory team when it comes to career development?

Keith Peiris

To some extent, I am very lucky because I was able to meet many thoughtful people when Facebook joined. Then a few years later, they all left and decided to create something else and become investors. This certainly helps. Regarding your question, who is my personal advisory team? I’m not sure if I have a personal advisory team. I have my fiance, and I get all kinds of advice from her.

Most of the time, I think there are many people with unique expertise in a certain field. It is very important to be able to contact these people to get professional knowledge or advice in creative fields. But I think it is difficult to find a person who is excellent in every field. But most importantly, I have a group of people. When I have crazy ideas about machine learning, I will ask them for help. I also have a group of CEOs. Sometimes I call them to ask about the decision of the leadership. Then I have a group of product designers or investors who give me advice.

I began to build such a model. When I need to learn something new, this is my current job. I always learn new roles, do it myself, and then finally hire or persuade others to do it for me. I found that I created this model, formed my own point of view, and then communicated with three to five people to improve this point of view, then stopped communicating with others, formed a new point of view, and then tried. So I have been thinking about the micro-network of these professional knowledge, as you move forward in this journey.

Grace Gong

As far as skills are concerned, you have changed from an engineer to a product manager and then to a CEO. You think there is a skill that you are constantly striving to improve. It can be technical skills, such as coding, or soft skills, such as sales.

Keith Peiris

A hard skill and a soft skill. For hard skills, I am now trying to catch up with machine learning. I think I have some understanding of the possibility of sorting, recommending and searching 10 years ago. I really like the progress made in the past few years. Once I catch up, I fall behind the next morning, so I am constantly editing. This is my view of the world, and it is very useful for me to constantly improve it.

Another skill I’m trying to improve is to be able to give a real presentation, that is, you know, if I look at my meeting schedule, today or this week, some of it involves finance, some of it involves recruitment, marketing, product review and so on. And I found that if I take these ideas from one meeting to another, I can’t play a role. Because this is a completely different competition, for me, it’s almost like a vet. You see a dog, then a cat, a chicken and a bird, so I’ve been working hard, trying to leave my thoughts at the end of the meeting, and then concentrate on the new meeting without thinking about the previous meeting.

Grace Gong

What do you think is your personal advantage when you start a company or incubate a company in Greylock? How does this translate into the company you are creating?

Keith Peiris

This is a good question. I can think from several different angles.

First of all, if you want to start your own company, one of the obvious things for me is that I will use the same thing to promote this company ten times a day until the rest of my life. When you think about it, you need to be deeply curious and fascinated by that problem area and the people you serve in order to promote this thing with passion. Because this is something you will devote all your energy and thinking to, a big part of your adult life. If you are really passionate and excited about your field, then I think other things will become easy.

For me, I have always been fascinated by communication. This is the main line of my work. Even on Facebook, I first engage in searching for products. I think searching is a kind of communication, by letting the machine give you what you want. Then I work on an Instagram camera or post messages on Instagram, and I am always fascinated by why people like to talk to each other. Why are they afraid to talk to each other? What will prevent them from sharing their faces in the morning? If you like someone, what way will you break the deadlock?

So when I think about different ideas, I know I want to work in a communication company. And I know in my heart that I know something about communication, because I have worked in this field before. But I also have this strong desire to engage in the communication of ideas. In other words, together with all the people involved in camera products and social networks, we have made it very easy to share your face, what you do on weekends and what you eat. However, if you want to share some thoughts in your mind, we don’t actually have a good medium. You must either record a podcast, make a slide show, or write a long document, and no one will read it.

I’ve been thinking about it. I want to reinvent the way people share their ideas and make them clear. I’m very, very fascinated and full of passion. Moreover, those who I will provide services for, that is, those who have ideas to share, want their bosses to give them opportunities to realize them, and they want investors to give them money so that they can realize their ideas. They hope to write a story to attract people.

It sounds like these people I want to serve for a long time. This is the origin of Tome, and then once I know what I want to do in the next 10, 20 and 30 years, most of my thoughts are, how can we turn it into a company?

I come from social networks, so my first thought is whether we can build a media of visual ideas. I soon found out that this is a very difficult company to build, and it is a business model of an enterprise. Not all incentives are the same. If you build an advertising business, you only need to let people keep creating, because you want to put advertisements to the audience, or if you only build something for consumers, you may not be able to build connections with different data sources or use a lot.

Grace Gong

So how to build a product that can eventually develop into commercialization?

Keith Peiris

First of all, when I start a company, the first person I want to bring in is a technology co-founder. I have some unique views, that is, the first person I want to start this company with should help me imagine and shape the products we want.

I think once we determine the exact product we want to build, then we can hire a team to build it, instead of just looking for someone to think about it too early. We are not sure what to build.

I found my co-founder Henry through a mutual friend. In the first few months of Tome, we worked closely with Graylock to sketch and imagine what we wanted. When I say "we", I mean Henry, who can draw and has design skills, but we both worked in the mobile field for a while, so he designed the mobile prototype of Tome.

Imagine if you use Notion, Canva or PowerPoint, or see such a tool based on mobile devices, it seems incredible to many people, so we showed it to our friends. It’s just a non-working prototype, just an idea, but we showed it to those friends who work in these companies, and they all said it was really interesting.

If you can find out how to build it, we may pay for it, and then we start looking for our founding engineering team. They are very, very excited to build the first version, and then we all have the belief that we should build simple things first and wait for customers to tell us what else they need, so that we will know that we have not built anything extra.

In this field, one thing may be scary, that is, you unexpectedly spent five years rebuilding PowerPoint, but never really realized the adaptability of the product market, so our idea is to build a tiny thing that we know will work first, and then iterate, so in Tome version, you can only put two tiles together, which is a text and image tile and the like.

We said to try a page with only two tiles. Of course, people didn’t make much progress. Every team said, I need to put three things on the page and start developing a system so that we can put more things on the page. Then we built the system and got feedback, but we still need to theme and brand Tome, because this is what the company needs.

So, we keep iterating and working with a small group of people until we start to create something that many people want to use.

Grace Gong

Have you ever imagined that we will solve the biggest problem in the world? At the beginning, what is your vision for your future work? How does it shape you to iterate over your product by following this vision?

Keith Peiris

So actually, I’ve been talking about if we set up this company in 2030.

I mean, we will use neural links to build something, because in a perfect world, you should be able to complete every transaction with complete loyalty like connecting your brain to the person you want to tell a story, but we haven’t reached that stage yet.

So, before we reach that stage, what can we build with the software on the screen? Then, he is really fascinated by the idea of a magical atmosphere. If you play many RPG games, many of them have an atmosphere, and this atmosphere has a spell. We imagined anything like opening a book, choosing a hologram to jump out of the book, and then it tells stories and shares everything, so that’s the vision we want to achieve.

Then, we ask ourselves, there are many participants in Google, Microsoft and so on, and since these tools were conceived, the world has changed a lot, both culturally and technically. Is there enough room to have a great impact on startups? The reason I say this is that traditional companies and new technologies usually don’t match well, right?

Although Microsoft has contributed to cloud computing, you can say that they never really turned Office to the cloud, or when I was working at Facebook, although we knew that mobile devices were very important and tried to think about how to become a mobile device-based company, but we could not build a mobile device native experience like Instagram or Snapchat.

We sat together and thought, and it seems that there are three things that are very different today from the release of Google Slides in 2000 or 2006:

1) It is almost obvious that mobile devices have become a fact. Whether at work or in life, we all use them to watch stories, transmit information through Slack, view them on Twitter and view them by email. In my opinion, if you want to build a 16-to-9 rectangular screen, it is very difficult to re-imagine it to adapt to mobile devices such as touch or iPad. And we think there is a long way to go, because many people no longer have laptops, especially young people or people in developing countries. So this is a key point.

2) We have a belief that in the modern world, the best stories are multimodal. I mean, there are modes in media, such as video, image and text, but in the Internet age, you should be able to quote different data sources from different places. Even when I work on Facebook or Instagram, my presentations are all screenshots of different network services, screenshot data, screenshot customer feedback and screenshot design files. We think we should build a tool to connect different data sources, because I think the best stories will quote many different data sources. I seem to think that traditional selection companies have not thought about this problem, they are thinking, such as drawing a graphic canvas. So we’re like, oh, it seems very basic, involving the way people work and think.

3) We believe that if you read a lot of presentations, I read about 20,000 presentations before I started, you will notice that many stories follow familiar patterns, but the expressions are different, but many presentations of product reviews follow very similar structures, which proved to be very true.

The way you fill in the details is a unique part. I like to joke that I think every Pixar film follows a similar narrative structure, but they are very unique in their own way. At that time, there was a technology to connect, so that I could read every story and every presentation in the world, and then you could connect them and build a structure for your story. That would be a change, right?

As a thought partner, it would be a crazy change. At that time, GPT was such a thing. We are not sure that it will get better in two or five years, which is a time range, but we know that it will become an important part of the company. There is enough foundation in the progress of mobile devices, different modes and network services, and large-scale language models, but it will be very difficult for a big company to change their practices.

We are thinking, if we start from scratch and save their time according to all this, can we build such a good thing? In this way, we can take the lead for a few years and then move forward as fast as possible. This is the main content of our thinking. Of course, you can’t accurately predict when all these things will land, but so far, many of them are correct.

Grace Gong

I think your initial product was really great. Just like Tome had 1 million users within 134 days after its release, you are the productivity tool that reached this milestone the fastest, surpassing Dropbox, Slack and Zoom. Curious, how did a rapid growth happen? How did you adapt so quickly? Because you mentioned that there were 40 ~ 50 versions, was that before you released your first product? So how did people quickly accept and use this product at the beginning?

Keith Peiris

One thing I wish I could tell myself two years ago, the Internet was a very efficient communication channel. If something you build is really ten times better than the products or solutions people are currently using, it will spread quickly on its own, and it will spread on Twitter, Reddit, Discord and TikTok.

I think as a founder, your primary goal is to build something really better. We have done something that a good marketing team should do, that is, whenever we launch a new feature, we always send an email to existing customers first, and we always make a short video and post it on social media.

When we first launched the first version generated by Tome, we thought it was not bad, so we released it. When we released this video, it quickly became viral. I think Henry and I both released this video a week before Christmas, and then it has been forwarded, forwarded and forwarded, and then celebrities began to find it on TikTok. Customers are also pouring in.

It is such a snowballing process. During the holiday, our service encountered a bottleneck, and the usage exceeded the access limit of OpenAI. It can be seen that this is an improvement. We realize that our job now is to stick to it and shape our image.

We will now contact celebrities and invite them to cooperate to help us launch our next product. I think the craziest thing that happened in the past few months is that our user base has completely changed. From the initial thousands of teams, these teams are very good at using our products, such as problem teams, engineering teams and design teams, and now we have all kinds of users. One of the biggest challenges we face internally is how to determine the priority of the next step, because we have various needs from authors, students, CTO, founders and product managers. So we are on this difficult priority path.

Grace Gong

You didn’t think about products at first. People didn’t want to create things from scratch. People hope to work on this basis, perhaps with a correct existing format as a guide, rather than starting from scratch.

You are the first company that takes AI first. Compared with the big companies in the world, they usually started a long time ago. They add AI to their teams or do similar work. I am curious. I believe investors will also ask us, what is your competitive advantage?

For example, when Microsoft or other companies suddenly plug into the technical department, such as inserting a camera or inserting Microsoft, especially for Microsoft, they have invested in OpenAI, and they definitely have more advantages in integrating AI into products. How did you solve this problem? And compared with these big companies, what are the advantages of AI-first companies?

Keith Peiris

Regarding AI-first or native AI, I think it is an idea, that is, how to consider capabilities, limitations and limitations when designing and conceiving products.

I think one of the obvious parts is that when you ask a large language model or an image generation model to create something for you, it’s a very iterative process. You know it’s almost like working with a person. You talk back and forth to edit and shape, and finally what you get is very different from the instructions.

When I use the drawing tool, I will put a rectangle here and then an arrow here. I am very instructive, so when we build AI native products, we think a lot about how to build something that feels like a dialogue, but it is not necessarily a dialogue in the chat window, but a dialogue in a sense. We generate something for you, then give you tools to drag, shape and perfect it, and then we will accept feedback. Then we will eventually reach a different position.

We will learn something, we will learn that you like to write in a very direct and logical way, or we will learn that you like to generate images with Cyberpunk aesthetics. Then we’ll start giving you more things you want. When I think about this problem, I think it is very difficult to integrate a product that has existed for 20 years.

I think a good analogy is that if you drive a Tesla, or they need to imagine the limitations of making an electric car with batteries and electric motors, so you need to put the battery at a low center of gravity, you need to redesign the transmission to consider torque, and then you need to redesign the dashboard, because you will drive less and less, compared with a classic car.

When I think about Tome, compared with some traditional products, they almost feel like a refitter, just like installing an electric motor on a burning car, while Tome is very, very different. Anyway, now we are beginning to realize these benefits gradually, that is to say, we are beginning to consider enterprise pilots.

A very common request we hear is that those who are interested in Tome say that we are happy to build on your large language model, but we also want to have our own data layer. We hope to upload everything from our company’s repository, such as documents, charts and data, and then we hope that you can use these facts and ideas and integrate them into the structure of GPT-4. This is the third thing that seems very difficult to do, unless. So we are now weaving all the things in the company, such as the company’s internal website, into a large language model and adjusting it, and finally giving you something that can write like you and someone in your company.

Of course, a big company can do it if they want, but I think it is very challenging for a big company to take it as a sideline.

Grace Gong

You have recently obtained a very successful financing, raising $43 million, with a valuation of $300 million, attracting many famous investors. Before that, what was your interesting financing trip like? And what is the business model like? I know that you plan to charge 10 dollars per user per month, as well as enterprises and so on, so I think many decisions may be based on the adoption situation, that is, who gets the fastest cash, what do you think of this?

Keith Peiris

Early financing, I think, is still in a very early stage. We have been operating for more than two years. Whenever we raise funds, we all realize that we will never convince an investor to believe something they didn’t believe. I believe other founders may have different experiences, but what is really important for us is to find those investors and board members who share our beliefs in technology and market scale.

Frankly speaking, we should find them as soon as possible and associate with those who hold different views as little as possible. For example, we deliberately adopted a very consumer-centered marketing strategy, that is, we believe that we have the best chance to build a lasting company.

For users, it is always free, and it can grow organically by people sharing Tome on the Internet. We are sure that if you bring it to work, you will be willing to pay for it one day. This bottom-up way for consumers to market is the best way to establish an engineering company, because it gives us the opportunity to overcome the competition from the huge sales teams of some listed companies.

We have raised funds many times, and found that the truly insightful toC investors know something about this. They will say, you are talking about the network effect, we can talk about the dissemination, participation maintenance and how you think it can promote the growth of this product. This is a very important but easy way for them, because you can say that there are many lightweight formats suitable for mobile devices, and you can say that you have a lot of such content; Because of this, we have many people who share our beliefs. For them, this is very quick and obvious.

Another thing is that we have found investors who really use our products, not those who try to read the opportunities. This is an interesting discovery. I think one of the difficulties faced by a company like ours is that you need to build horizontally and consider vertical marketing at the same time, which means I don’t know who the text editor is designed for. I don’t know what kind of people the chart system is designed for. They can serve all kinds of users.

Therefore, you need to build a lot of people horizontally and have confidence to enter the enterprise market vertically in the future. We found that not everyone in Silicon Valley believes this, so it’s about believing.

We will start charging soon. We will always provide a free version. In this quarter, we don’t want to charge students. We want them to use it and give full play to their creativity. One of the reasons why we started charging is that we have some exciting events coming. A question asked by many customers is how to get unlimited computing power. I don’t want to worry about running out of quotas because of different operations. I just want to have computing power when I need it. For example, when making important presentations, I feel the same way. I don’t want to think about quotas when I work, which is one of the reasons we will take.

Another reason we build it is that it helps us to know who really uses this tool seriously. Obviously, some people use it every day, and they may generate some tomes, but this is different from those who pay monthly, which will help us focus on the correct staffing and pay attention to the road map. One thing we discussed internally is that we want to grow into an enterprise company, but at the right speed and pace.

If we focus on the enterprise and team too early, we will lose the magic as a startup company, lose the opportunity to build a product that people love and can expand, and focus too much on things like enterprise-specific functions, so we are considering the enterprise version, and we will start some experiments. We will never forget that we are essentially a toC company.

Grace Gong

I’m curious, just like when you started, when you raised funds, has Tome become what it is today? Or did it not achieve today’s function? What was the AI industry like at that time? Because you started about two years ago, now AI has become a hot topic.

Keith Peiris

When it comes to financing, I think you must be very honest about what you have and what you need to build next. At the earliest stage, we had excellent prototype design. Henrik was an excellent product person. He drew the whole blueprint on Figma, made various prototypes, and showed the possible appearance and demonstration of the function. Then I conducted a lot of customer research. I made a survey on Survey Monkey, called some potential customers and recorded on Zoom. Ask them what they think of PowerPoint and what conditions they need to use the new tools. We try to present the situation that we know exactly what we are dealing with.

To be honest, we just need to build it, which I think resonates with many investors, and it turns out that this method is effective; As for AI, it’s very interesting. I want to quote a quote from the CEO of Stripe. It’s difficult to predict what will happen in the next 1-3 years, but you can predict what will happen in the next 10 years more accurately. I really feel this way. I don’t know what the economy will look like next year, but I have an expectation for the economy in the next 10 years.

When we designed Tome, we knew that large-scale language model and machine translation would be important components, but we were not sure whether it would be possible in 2022, 2024, 2026 or 2029. Based on this, we decided to form a small team to build what we knew needed to be built, and keep an eye on the pulse. We paid attention to all kinds of development and stability and all kinds of important days. At some point, this direction will become obvious, and it began to become obvious last year.

Grace Gong

What are you like in terms of talents? For example, how many are engineers and how many are designers in proportion? How many are sales and business people, which are very important for business people like us. What was your idea when designing the team? Then how do we arrange the budget? For example, "how many dollars will we sell in 2023, and then we will achieve this goal by forming 10 teams", is it bottom-up or top-down?

Keith Peiris

I think the answer is always both. Of course, we set up a very simple Google form. We realize that we can hire a certain number of people, and then we evaluate it day by day and adapt to our own needs.

Our philosophy is that when you start a company, try to hire an excellent person every month and try to hire someone better than them next month. At some point, it will become more and more difficult to hire talents, because you will have an excellent team, which is a control mechanism.

As a startup, you have to convince people to take risks and choose you, so usually you are limited by how many people you can find, not how much you can pay. Therefore, we always have a limit, we like small teams, and we keep our scale; You can always do more unless you have to (according to Linkedin, there are currently 71 people).

We need to be very precise about the next 1-3 people to be recruited, and then we will constantly re-evaluate, because as a founder, you will get a lot of new information in the initial stage, your user base may be completely different next week, all customers may be different next week, and you may encounter infrastructure problems next week, because many things are changing, and I don’t think it is beneficial to make long-term planning too strictly. I am more concerned about whether I am making the best decision for the next recruitment, which is the most important thing.

Grace Gong

You have raised a considerable sum of money. What do you plan to do with it? For example, a team that develops the latest technology? Or what are your plans?

Keith Peiris

We are a technology and product company, so we have invested heavily in developing machine learning and in developing design, product and engineering teams.

As you know, this has always been our core, and I think this is always a place where people come here to build something they are proud of. The work you build on Tome should be the best work in your career, compared with the previous company, so it will be a huge investment.

Having said that, we are transforming into a company with more users than a few months ago, dealing with payment problems and exploring the enterprise market. Therefore, we are building a very strong sales team to help improve all aspects, but our goal is to keep it as small and lean as possible and keep it as long as possible.

I think everyone has the same background and knowledge, and everyone owns a part of the company. Because of our salary method, we are all very focused, and we all really feel the same about our mission. For me, this is one of the most interesting parts of building a company, accepting this scale. I think we will try to keep this scale as long as possible.

Grace Gong

I heard that many companies are developing with technologies like GPT-4 provided by OpenAI, such as GPT-4 on AWS. Why is everyone developing? How difficult is it to train your own model? Suppose I have a unique demand. You mentioned that Tome can help you create with your unique writing style. For example, I don’t want every presentation on the Internet. I just want to be like a YC document. I’ve got 100 million dollars. How can I add this layer to my Tome? Can it be done now? Or are you trying to develop it? How does it work in terms of technical composition?

Keith Peiris

First, we consider using these basic models as a platform. I think it’s actually more like a comparison between AWS, Google and Azure. I think they may be more like oligopoly than monopoly. I think there will be three or five main participants, all of whom are developing rapidly. I think many companies will become very good, and many companies will also tune for slightly different applications.

Therefore, our point of view is to build the best experience for customers. We have built our system to completely replace the model accurately to get the best experience. Even though we are using a model to generate images now, there are new models coming out from different suppliers, which are much better. We never want to be caught up and not use the best model to serve our customers, because it is a potential space, and our iterative and adaptive ability is crucial to survival.

Our focus now is how to obtain data sources for you and your company. These data sources involve the fact that you want to use them again and again, which is the most important thing. If you work in a bank, we want you to use your bank’s research and facts, or if you work in a Web3 company, we want you to talk about your technology instead of some fantasies generated from the language model.

In terms of style, it is interesting that we often argue that we should try to make Tome sound like you. Or should we try to make Tome sound good? I think the current situation is to make Tome sound good, convincing and true, because I think the individualization of style may not appear until later.

For your next question, how to train? How to train only the correct presentation, not the wrong presentation? This is a good question, and we began to explore it. If you enter the Playground of OpenAI, you can extract the content and use it as part of the context window, and then you can train it for a short time and then output something based on it.

One of the cool things about GPT is that we can get a large context window immediately, which can reach 25 pages, so it has a background and can train specific things locally. We are doing some work, and I think they will be very noticeable in the next few months.

Grace Gong

What will your future job look like for you? Now, a hot topic is that whenever I post AI-related content on LinkedIn, there will always be people who like it and they are very excited about the future. The other half think that AI will take away our jobs. What do you think of this? What impact will it have on future work?

Keith Peiris

Tome’s point of view is that we don’t really believe that AI will completely replace human work. We really believe that AI is a tool for human enhancement, which helps us realize our dreams better, faster and more efficiently. This is one of our excitement, because no matter how much you train the machine, the machine can never create a presentation for you without human intervention, because many of the presentations are your vision, your imagination and the company you want to build.

The best assistants can only give you some successful examples and give you a general and effective structure, but shaping it into your own thing is what you need to do as a human being, and this is what excites us.

When I think about my future work, the reason why we focus on storytelling is that I think storytelling is the basic building block of productivity. When I persuade you to work with me by telling a wonderful story, sometimes it involves facts and sometimes it involves visual effects. This is the way I work with you. Therefore, I think storytelling is a prerequisite, and we need to coordinate our work. We need coordination to deal with big and small problems together.

We know that Tome is an epoch-making company. If we can help anyone tell a fascinating story, we think the best idea will win. We believe that society will cooperate with these ideas. When I think about the wrong information on the Internet, you know, we talked about the filtering bubble, which is part of it. We only hear people who share our views, but I think the other side of the problem is that we have a good communication platform to discuss complex ideas with each other. I hope Tome can move in the right direction and help people discuss complex problems in more subtle ways.

This article is published by Entrepreneurial State authorized by the columnist, and the copyright belongs to the original author. The article is the author’s personal opinion and does not represent the position of the entrepreneurial state. Please contact the original author for reprinting. If you have any questions, please contact editor@cyzone.cn.