Image Source: Generated by Wujie AI
The popularity of large models and AIGC behind them is a journey of technical experts.
In the era of mobile internet, the division of front and back-end responsibilities has "sealed" programmers, engineers, and other technical personnel behind the scenes. They build product architecture, fix bugs, and optimize experiences, while product managers stand in the forefront, creating "temple-level" products based on their insights into user experience.
Today, the technical staff fighting on the front lines have become senior managers in companies, either CTOs or responsible for a specific business line. From accompanying the company's growth from 0 to 1 to maintaining the status quo, they have long left the front lines, but their hearts are always "technical."
The explosion of ChatGPT may be a redemption for "technical" people with dreams.
Unsettled Dreams
"I really miss the days of staying up late writing code, and I hope to regain the 0-1 state," said Jia Yangqing after leaving Alibaba.
Jia Yangqing, well-known in the industry as the "framework guru," majored in automation at Tsinghua University. During his postgraduate studies, he worked/interned at the National University of Singapore, Microsoft Asia Research Institute, NEC Labs America, and Google Research. In 2008, he went to the University of California, Berkeley to pursue a Ph.D. in computer science. During this time, he created the open-source deep learning framework Caffe, which was adopted by companies such as Microsoft, Yahoo, NVIDIA, and Adobe.
In 2013, he joined the Google Brain team and, under the leadership of "Turing Award Three Giants" and others, participated in the development of the TensorFlow platform. In 2016, he joined Facebook and created the prototype of the first open model format, ONNX. In March 2019, he joined Alibaba as the vice president of technology, responsible for the technical, product, and business aspects of big data and AI.
On March 20, 2023, accompanied by absurd philosophy, he left Alibaba. "My team and I announced the news of graduation. March 20th is the last day of winter. Borrowing a phrase from Camus: the end of every winter is the beginning of spring."
Some speculate that he is starting an AI architecture venture, while others speculate that he is working on large models. In April, Jia Yangqing stated in "Tech Early Know" that he is not working on large models, but rather on building AI application layers in the B2B direction, bridging the gap between small and medium-sized enterprises and AI, and believes this is a rare opportunity.
Similarly, on March 7, Li Mu left AWS and co-founded Boson.ai, an artificial intelligence company, with "father of parameter servers" Professor Alex Smola. In July 2016, Professor Alex Smola joined Amazon AWS as a vice president-level scientist.
On March 8, it was reported that Li Yan, head of Multimedia Understanding (MMU) at Kuaishou, established an AI company called Yuanshi Technology, mainly focusing on the research and development of multimodal large models.
Li Yan graduated from the Institute of Computing Technology of the Chinese Academy of Sciences and is a key figure in Kuaishou's AI technology development. In November 2015, he established the internal deep learning department DL (Deep Learning) at Kuaishou, with the goal of building algorithm models to identify illegal and non-compliant video content. As Kuaishou's demand for video content understanding grew, the team was reorganized into the MMU group in 2016.
On March 24, Wang Changhu, head of visual technology at ByteDance, left to join Longfor Group, where he is building an entrepreneurial team focusing on a visual multimodal algorithm platform for generative AI.
In 2017, Wang Changhu joined ByteDance's AI Lab as director, focusing on computer vision, video understanding, multimedia retrieval, and machine learning. With internal position changes, he became responsible for commercial visual technology at ByteDance, overseeing several major product lines including Douyin, TikTok, and Toutiao.
As executives, Li Yan and Wang Changhu are also constrained by the needs of the business in large enterprises, becoming assistants to business growth.
"Being a technology expert gives a natural advantage in entrepreneurship in the technology era," a senior executive told "Titanium Media Venture Family." The pace of AI development is very fast, and although large enterprises want to keep up, the limited ability to collaborate and solve problems between various teams actually hinders the display of AI innovation capabilities.
"If I could start over, I would make the product more innovative and better adapted to the market," Jia Yangqing publicly stated.
On the other side of the ocean, a similar scene is unfolding.
On March 22, Intel's chief architect Raja Koduri left to start a startup focused on generative AI for gaming, media, and entertainment. He stated on social media, "Generative AI reignited my passion for entering the software field, and I hope to use non-CUDA hardware to handle this workload, which was my original intention for starting a business."
Having participated in the development of LLaMA's Lacroix and Lample, he left to found Mistral AI, a company specializing in large language models and generative AI, and received $113 million in seed funding within a month.
At this turning point in time, their departures seem to reveal unresolved dreams.
Dreams and Reality
The bridge between dreams and reality is money.
"The most frightening thing for technical personnel to start a business is to be trapped in technical obsession and neglect commercial indicators. Capital is the lifeline of a company, whether it's financing or self-sufficiency," said an industry investor.
There are many commonalities between technical expert entrepreneurship and scientist entrepreneurship. For example, having solid professional capabilities in a certain technology, having a halo effect from past industry resources, and having deep insights into the prospects of technology and industry changes are all loved by VC investors.
Similarly, the challenges faced by scientist entrepreneurs are also reflected in technical experts. For example, the comprehensive ability to think commercially, manage, and market needs to be supplemented. While technology focuses on doing deep and refined work in a specific area, entrepreneurship is about monetizing technology, bringing commercial value, facing market competition, and having the ability to innovate profitably.
Top technical experts "embrace the thighs." The four major players in Silicon Valley's large models all have allied partners.
OpenAI is backed by Microsoft. Over the past three years, Microsoft has cumulatively invested $13 billion in OpenAI, and its valuation has reached $29 billion.
Following closely is Anthropic, which has partnered with Google. Google has invested $300 million in exchange for a 10% stake in the company, and Anthropic has chosen Google Cloud as its preferred cloud service provider. Founded by Dario Amodei, former vice president of security and policy at OpenAI, Anthropic's valuation has exceeded $4 billion in two years and has accumulated strength to compete with the flourishing OpenAI.
Next is Inflection AI, which has partnered with NVIDIA to build the world's largest AI cluster. Lastly, Cohere has reached a cloud computing AI agreement with Oracle, which plans to sell access to Cohere's large language models to its cloud customers. Founded by Gome, who previously worked as an AI researcher at Google and was one of the authors of the "Attention Is All You Need" paper, Cohere's post-financing valuation has reached $2.2 billion.
Second-tier technical experts compete in speed, with the speed of market insight, product development, and monetization, also known as the first-mover advantage.
A serial entrepreneur said that before a new technology is introduced, academic papers come out, followed by apps or demos made by technology enthusiasts on technical forums. This is the time to act. The premise is long-term cultivation in a specific technology field, which allows judgment of which technological iterations bring new business opportunities and which cannot currently be commercialized.
"Get rid of technical superstition and look at technology from a business perspective. Technology is a tool to help companies monetize and acquire users," said the serial entrepreneur.
Third-tier technical experts ultimately become CTOs. "Everyone has their own strengths and weaknesses. For founders with a technical background, they have a great advantage in the early stages of entrepreneurship, being able to quickly develop products. In the expansion phase, they need to capture the market. If their capabilities do not match, we will help them find a CEO partner. This is not to deny the ability of technical entrepreneurs, but from an investment perspective, it is a strategy to help the invested company generate cash flow and develop healthily as soon as possible," said an industry investor to "Titanium Media Venture Family."
Fourth-tier technical experts "wander in the community" and await acquisition. "In the rapidly developing AI landscape, we acquire innovative AI projects in the 'community.' There are many excellent engineers in open-source and technical exchange communities who develop applications that can help large enterprises quickly pass the cold start phase when trying new businesses. Internal project initiation, development, and cost require approval and evaluation, which all take time. When the technology trend arrives, we don't want to miss out, so we explore good application projects in the community and communicate with those willing to sell their projects," said an investment professional in a corporate venture capital department.
The classification of first-tier, second-tier, third-tier, and fourth-tier technical experts does not specifically refer to differences in the abilities of technical experts, but rather presents the diverse survival states of technical experts between dreams and commercial realization.
For technical expert entrepreneurs, the line between technology and making money seems less distinct. Technology may no longer be a sufficient competitive condition, as when technological concepts have been validated, the remaining competition will be more about money and application scenarios.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
