GPT (Generative Pre-trained Transformer) large model has become one of the most important technological breakthroughs in the current field of science and technology, and machine intelligence has exploded in the form of emergence. Through the rapid growth and popularity of applications such as ChatGPT and Midjourney, C-end users can also feel the technological leap with a low threshold. AGI (General Artificial Intelligence) is emerging as the next industrial revolution. At this moment of great changes in the technology-driven industry and society, Chuxin Capital has precipitated its thinking on the changes and immutability of the post-GPT era, and hopes to maintain long-term sensitivity and in-depth discussions with you.
Straight to the point, based on the needs of human nature and the historical development of the industry, we believe that some things are constant:
1. GPT large model will become the infrastructure of the new era, and will be deeply integrated into people's lives in the future and become the new normal. Similar to the popularity of public power grids, people in the future will become accustomed to AI applications that grow on large models, and large models become the support behind moisturizing and silent.
2. The long-term competitive landscape of the large model can refer to the industry development process of cloud computing. Due to heavy technology investment and high scale effects, the number of players in the market will gradually converge, the first and second places will dominate, and the three players will consider differentiated competition such as open source.
3. Insights will contribute the most valuable value in application-layer business competition. With the help of GPT large model technology, the realization of ideas becomes faster and easier, and insights will lead to more precise and sharp entry points, which will become a source of important competitive advantage.
4. Elite-driven innovation brings great social value. The GPT era will better empower every creative individual and productize ideas faster.
As OpenAI founder Sam Altman said: Everything deep in humanity is not changed by AI. As human beings, we still pay attention to the interaction between people, the reward mechanism of the human brain has not changed, we still pursue happiness, have the desire to create and compete, and desire to form a family... What humans cared about 50,000 years ago, humans will care about a hundred years from now.
But there is no doubt that the GPT big model will bring huge and profound changes to human society, and this process is already accelerating:
1. Specific industries have ushered in dramatic changes: Industries such as gaming, design, and education have been strongly impacted by the introduction of GPT large model technology. In the future, most knowledge workers' work tasks will be influenced by large models.
2. Changing user groups: With the support of a new generation of AI-enabled tools, digital native user groups will nurture new product opportunities.
3. Changes in organizational forms and labor relations in the new era: new jobs such as prompt engineers have emerged, and there is a trend of integration of positions (such as creative technologist). The trend of distributed office collaboration has been further promoted, such as Tome AI's adoption of distributed remote work model, with two weeks of IRL (in real life) offline time each quarter.
4. The quadrant model of competence of excellent entrepreneurs has changed: the experience and resources accumulated in the past, including capital, will no longer be a strong barrier; Many new forms of entrepreneurship are done by small, lean organizations (such as Midjourney, an unfunded company with just a dozen people).
01 | Infrastructure needs remain the same, but new features have evolved
The LLM big language model has become a new infrastructure and entrance, driving the rapid explosion of upper-layer applications
Over the past 20 years, the percentage of people working from home in the US has grown from 3.3% to 5.5%, and this percentage soared to 58% during the pandemic in 2020. In May 2021, Google CEO Sundar Pichai announced plans to put 20% of the company's workforce into permanent remote work. Remote and hybrid work have gone from emergency solutions during the pandemic to the new normal for people's future work.
The basic layer and the model layer will be integrated, and in the future, the large model will become the entrance, and the underlying cloud resources will be freely scheduled, and users no longer need to pay attention to which cloud vendor's cloud resources and services are used. Just like when you come home in the summer, turn on the air conditioner to blow a cool breeze, open the refrigerator and take out a cold drink, turn on the TV to find a sci-fi movie to watch, and take out your mobile phone to brush your friends. People will be thankful for the facilities provided by modern technology, and thank them for the changes in air conditioning, refrigerators, televisions and mobile phones, but not for the public grid.
Cloud computing vendors are facing market changes, and the entrance may be preempted by large models in the future, resulting in their own ecology being squeezed and reduced to server vendors; Therefore, there is also a strong incentive to layout and catch up with large models.
The model layer may become the thickest layer of value in the GPT technology revolution, not only becoming a new infrastructure, but also extending to the application layer. OpenAI's ChatGPT, along with Microsoft's New Bing and Microsoft 365 Copilot, are expected to be toC's killer apps. Applications that only mobilize large model API interfaces will face challenges in long-term differentiated competition and building barriers.
Infrastructure characteristics change: LLM large model training cost breaks through the limits of Moore's Law, convenience and penetration rate are unprecedented
As a new generation of infrastructure, the characteristics of the big model have completely broken through the characteristics of cloud computing as the previous generation of infrastructure.
1. Emergence. In the era of cloud computing, human beings have good predictability of the development process of cloud computing, such as the development stage and performance of products, and the iteration speed of scale. But large models are characterized by the emergence of intelligence, and the rate of emergence is also rising rapidly, which is unexpected and unpredictable for humans.
2. The cost decreases quickly. The speed of AI underlying computing power optimization is more non-linear than the cost reduction of cloud computing.
3. Convenience and penetration. In the past, cloud computing vendors have enabled traditional enterprises to be cloud-based and online. But this GPT is the first basic layer technology leap in human history, starting from Day 1 to take the individual as the main body of empowerment, Day 1 users are global, all industries and all ages.
Changes in infrastructure characteristics: The product iteration speed of LLM large models is becoming more and more amazing, and technological innovation has entered an explosive period
GPT, GPT-2, and GPT-3 are all released at intervals of about one year, but the launch time of ChatGPT and GPT-4 has gradually shortened to 3 months. At present, there are many voices in Silicon Valley calling for the suspension of the subsequent development of the GPT-4 model, because the discussion of technology ethics and safety has lagged far behind the breakthrough speed of GPT technology.
02 | The application layer is blooming, and specific industries are ushering in dramatic changes
Generative AI technology is already having a significant impact on some industries, and long-term disruption is expected
Taking the game industry as an example, a game company with avatar generation as the core used to spend a year on a technical team of 300 people to produce 100 avatars, and now, with the blessing of AI technology, it only takes a team of 30 people a month to produce 1,000 more vivid avatars.
Through the empowerment of AI models, a fresh college student can now be qualified for a position that once required a design major and has 5-8 years of design experience, and a college student can generate 700,000 yuan of design order output value per year.
AI big model technology is showing a broader and more profound impact in high-value areas, such as knowledge workers and highly educated people, which is very different from the expectation that AI will replace blue-collar jobs first. In the long run, with the further improvement of GPT technology, more and more industries and occupations will be affected, and people need to rethink their competitiveness relative to machine intelligence.
Change is expected to continue to evolve: Although it is still in the early stages of industry development, the overseas application layer ecology has blossomed
Although the industry is still in the early stage of development, the overseas generative AI application layer has shown a trend of blooming. The application maturity level in different scenarios is different, and a large number of tools and products are still in the demo and waitlist stages; There have been some products that have been successfully applied in actual scenarios and have been favored by users.
In the future, with the continuous advancement of technology and the continuous expansion of application scenarios, the ecology of new AI applications will continue to evolve, bringing more profound changes to people's lives and work. Here are some AI products that we think have new ideas:
1.Adept "AI teammates" help humans do their jobs. In addition to reading and writing, AI can also use tools such as Airtable, Photoshop, ATS, Tableau, Twilio and other tools to complete tasks such as "generate this month's reading report";
2. Descript edits own audio and video like a document, removing tone words and errors;
3. Runway A description or a style intention map magic change video; Synthesize creative videos in one sentence;
4.Character.ai Personalized ChatGPT, which can talk to virtual Musk, Steve Jobs, or personal characters;
5.Rewind AI can retrace everything you have seen, heard, and said on the computer in the past with one click, such as finding where and who mentioned a new concept before;
6. Lindy daily work assistant, manage calendar, mail, travel arrangements. You can find the right person according to JD intelligence;
7. Hebbia can understand your question and highlight all relevant information in the web page Chrome extension, making search more efficient;
8.Galileo AI Copilot for interface design Create editable UI designs with simple text descriptions, enabling users to design faster;
9.Scenario AI-generated game assets uses AI technology to generate high-quality game materials, including characters, scenes, animations, etc. Make game development faster and more efficient;
10.civitai.com AI Art-Generated Model Community Sharing Center. Free to use, open source, and constantly improving.
03 | Entrepreneurship develops change
Take Midjourney as an example to see the growth trajectory of the next generation of AI applications
GPT has brought many record breakers in the history of human business, of which Midjourney is a phenomenal project that must be mentioned. It was founded in August 2021 and is currently at the top of the track, founded by a co-founder of Leap Motion.
Midjourney is positioned as an unfunded company that has never raised money and has no plans to raise funds. The organization is extremely lean (plus the CEO has only 12 full-time employees), and iteration and growth are rare: ARR reached $100 million in less than a year of existence.
Take Midjourney as an example to see the growth trajectory of the next generation of AI applications
OpenAI's growth also broke records: it reached 1 million registered users in just 5 days, significantly faster than Netflix, Facebook and a host of other great companies. ChatGPT reached 1 million paying users in two months of launch, assuming that user retention reaches 50%, which is equivalent to reaching an ARR of $100 million in two months, which is an amazing achievement.
Some technical KOLs believe that if OpenAI does not adopt the strategy of obtaining the market at a low price, but fully unleashes its business value, it can actually reach an ARR of $1 billion in a very short time. Binance CEO Changpeng Zhao once shared that Binance is the fastest company in human history to reach a net profit of $1 billion, in just 18 months; Binance's secret sauce is to free up freedom of money, allowing money to flow freely. GPT has increased productivity by leaps and bounds and is constantly pushing the limits of human intelligence, so its long-term business value can be imagined.
04 | Organizational transformation: AI quickly replaces mediocrity, creates super individuals, flattens the organization, and the special forces model
Quote netizens' views: We are about to usher in an era of "eliminating mediocrity", and will also usher in the era of "one person is a team", becoming an eliminator or becoming a boss who scolds Fang Xuan is a topic that every low-level worker must face squarely.
In the new era, what is the relationship between capital and entrepreneurship? What kind of organizational form does innovation require?
We believe that innovation in the new era requires an organization with agility, rapid iteration, and efficient execution, similar to special forces, where everyone is a hexagonal warrior and can react and adapt quickly in a rapidly changing market environment. In addition, as the application of AI and automation technology matures, product development only requires a very small, extremely lean team, or even "one person is a team".
Innovation is likely to become more and more democratic, and the threshold continues to be lowered. Creative people can achieve innovation through accurate insights and landing efficiency, and for many application innovations closer to users and scenarios, the leverage of capital is no longer a key success factor. In the future, we will see the emergence of more entrepreneurial organizational forms and business models based on open, shared and independent innovation.
Traditional jobs are also merging. The original intention to see overseas startups recruiting creative technologist positions is actually the integration of design and technology. Midjourney runs the Discord community with a large number of part-time workers. In addition, prompt engineers have now become an important new type of work (although there are different opinions that this is a phased work, and in the future large models will be smart enough to rely only on natural semantics to achieve better interactions).
05 | Changing user populations: Digital natives will lead the rise of a new generation of AI-enabled productivity tools
In the face of technological change, the original investment strategy is "new", that is, new technology and new demand. From the demand side, we see the rise of a new generation of AI-enabled productivity tools led by digital indigenous groups.
The digital native crowd is always on the web, willing and good at trying new tools, and has a very high acceptance of highly flexible, user-buildable products. In part, this feature also explains the rise of the No Code tool.
Digital indigenous groups advocate individual expression and love to use social products. For example, Notion quickly became popular through Tiktok.
Post-95s and post-00s have also embraced new forms of work, and they are very adapted to the era of hybrid remote work, and more advocate the status of digital nomads and slash youth. Discord community part-time operations are a prime example. Many of the productivity tools needed for remote work and the creator economy will also become powered by GPT in the future, with better forms of interaction with users.
06 | The post-GPT innovation paradigm: insight is a constant source of value; Entrepreneur Competency Model Update
The organizational form of the company will undergo fundamental changes, and the future entrepreneurship will be more of a special forces model. Multi-layered, rule-based management means inefficiency, and founders no longer need to spend too much effort managing large, bloated organizations. Some scholars believe that if the founder spends a lot of energy on managing the team in the future, it is likely to be unhealthy and abnormal.
We believe that the real value comes from high-quality insights that keep abreast of changing needs. D5 Render, the original invested company, currently has 40 people and no overseas operation team, but 70% of the revenue of its products comes from overseas. They rely primarily on YouTube for distribution and sales to users around the world who need a renderer. We once asked the CEO how to explore opportunities, D5 Render CEO Niu Zeping mentioned that when it was immersed in the older generation of renderer user communities (such as Lumion), NVIDIA graphics card performance changed a lot, but the older generation of renderers required designers to choose between performance and efficiency, so the team combined the two values to make a high-quality rendering product.
Xiao Yixiang, CEO of Watermelon Maker, another invested company of Chuxin Capital, has an insight that investors pay attention to outcome indicators, but outcome indicators are lagging, entrepreneurs pay more attention to process indicators, such as NPS (net promoter score), when NPS is close to 100%, the product has a good reputation and a high referral rate, you can see that the product has gone out of the Black Forest, but investors rarely pay attention to this indicator. Where are the great young entrepreneurs? Commonly found in university laboratories.
The original team has developed the ability model of excellent founders: imagination, insight, determination, and continuous learning.
In this regard, Chuxin suggested that entrepreneurs must do relatively retreats two days a week and not get stuck in executive business, because in turbulent times, retreats are very important, and they should think in pieces, not fragmented time.
Another original investment team recently proposed an almost "crazy" goal in OKR: to achieve 70% of the company's work done by machines by 2023. When you embrace the post-GPT era, new paradigms and thinking are always born.
Write at the end
Chuxin Capital will continue to pay attention to new technologies and needs in the post-GPT era.