Image source: Generated by Unbounded AI
Author|Lightspeed
Compiled | Shiina
Edit | Zuri
Everyone wants to have their own generative AI tool. However, only a few people in the world can actually build large-scale language models, new AI application infrastructure and supporting technologies.
This requires not only a new and special set of technical capabilities, but also deep knowledge and creativity.
The engineers who are pushing the limits of today's generative AI technology are experts who have spent years honing their skills at the world's largest technology companies.
What’s interesting is that these days, these big companies are largely playing catch-up because their top talent has been poached.
With the flow of these top talents, a group of more flexible and innovative generative AI companies have emerged, which are called "AI Native" companies.
Recently, the American venture capital institution Lightspeed (which has invested in AI for seven years and has tracked the development of this field for a long time), based on publicly available employee information, carpet scanned 3,500 companies involved in the field of generative AI. Information from multiple companies identifies which AI companies are winning the battle for the best technical talent, and analyzes the sources of these talents, including academic qualifications, advanced degrees, previous companies, years of experience, patents, publications, etc. This article also sorts out and refines the essential information of this talent trend report.
01
Top eight, why them?
The following eight companies have a crushing advantage in having top technical talent. Lightspeed calls them "the Generative Eight" (hereinafter referred to as "the top eight").
No more selling out. These eight companies are:
Tome — AI that helps generate presentations and visual narratives
Character AI — Have conversations with various AI-generated characters
Anthropic — an AI security and research company building chatbots for the workplace
OpenAI — Releases ChatGPT, demonstrating the power of large language models, and DALL-E for image generation
Hugging Face — provides a set of tools for building models
Jasper — a model for creating marketing and brand-level creative content
Stability AI — Created Stable Diffusion, a powerful image generation model
Midjourney — Released a powerful image generation tool that runs entirely on Discord
Perhaps you have noticed that these eight companies are not purely companies that make generative AI applications. As mentioned at the beginning of this article, generative AI requires a series of supporting tools or ecological environment.
Therefore, Lightspeed clarified the scope of its survey as "screening companies with AI as the core, that is, building an AI ecosystem, rather than companies that apply AI to existing applications."
Under this premise, Lightspeed also selected a group of companies that have made the most significant progress in financing. The final rankings were generated by standardizing and weighting large amounts of publicly available data.
Image source: Lightspeed does not explain the sorting
These eight startups, which can be said to have the highest level of engineers and AI experts among their peers, have all achieved financing success and found product-market fit to some extent.
To a certain extent, they are also the most promising company in the field of generative AI to become a FAANG-level company.
In this "Top 8" list, OpenAI, which was founded the earliest, is already a household name. It has an originating position in the field of generative AI. It has a team of 400 people and its valuation has reached US$28 billion, the highest on the list.
There are three companies in the 2021 batch that were established late. Because they caught up with this wave at the right time, they have also achieved eye-catching financing results:
Tome, a PPT production technology service provider, received US$43 million in Series B financing in February this year, with a valuation of US$300 million; online AI dialogue company Character AI, a company founded less than two years ago, launched this year with zero revenue In March, it received a new round of investment from A16Z, with a valuation of US$850 million. At the end of May, less than a week after its new product was launched on the mobile terminal, the number of downloads reached 1.7 million, directly crushing ChatGPT;
Stability AI is a representative in the field of image generation. Its key technology, Stable Diffusion, is an AI technology model that generates images based on text. It only takes a few seconds to generate high-resolution, high-definition images without losing reality. Sexual and artistic pictures, in October last year, Stability announced that it had received US$101 million in investment from Coatue Management, Lightspeed Venture Partners and O'Shaughnessy Ventures LLC, with a post-investment valuation of approximately US$1 billion at the time.
However, the one with the most gimmicks and the most popularity is Midjourney, which has more than 17 million users worldwide.
According to public information, this company has annual revenue of hundreds of millions of dollars, but there are only 11 people in the entire company. What is even more shocking is that half of the eight R&D personnel are undergraduates who have not yet graduated. And there is currently no way for the outside world to value it, because it is growing too fast and the number of users continues to increase. Some overseas technology media have called out a valuation of US$30 billion, because they believe Midjourney's net profit in 2024 may exceed US$1 billion.
Interestingly, many of the companies in the top 8 were founded by OpenAI and veteran employees who resigned from Google.
Especially the latter, Google has invested heavily in AI research for years, but until recently had few visible applications or consumer-facing products or services.
02
Strong AI schools in the United States continue to provide talents
The new innovation landmark is here
After sorting out the talent backgrounds of the "top eight" companies, Lightspeed also found that the biggest thing they have in common is that the educational background of their technical talents mainly comes from many well-known universities (as expected), as shown in the figure below:
It can be seen that the strong AI schools in the United States have exported the largest number of top AI talents. Among the top 10 universities, there are only two non-American universities: the Artificial Intelligence Research Institute of the University of Waterloo in Canada ranked fourth, and the Artificial Intelligence Research Institute of the University of Leuven in Belgium. The institute ranks eighth, which also reflects its international character.
In addition to universities, the report also mentioned a special "coordinate" - in Silicon Valley, young developers and start-up companies in the field of generative AI are flocking to the Hayes Valley community in downtown San Francisco. This place has been dubbed the new "Cerebral Valley" because it is becoming a new gathering place for AI. In this vibrant and innovative community, the future of AI is being written. Young entrepreneurs gather here, with their passion and creativity for generative AI technology, and continue to promote the advancement of technology.
While it's impossible to track whether employees in a specific geographic area are remote workers (except for the 33 employees explicitly identified in the data), it's safe to assume that the top eight employees are using remote conferencing tools heavily.
03
11 “rising stars” in the application layer:
Talents with Tsinghua background emerge
While the "Top Eight" attract the vast majority of the public's attention, Lightspeed's research also uncovered another pole of the magnet that is attracting top talent.
Lightspeed calls them “The Next Gen,” the rising stars of generative AI.
Rising Stars places more emphasis on focusing on problems in a specific domain, or focusing on specific use cases tailored to real customers.
Although many companies in the "Top Eight" are expected to continue to maintain their leading position in AI, the "Rising Stars" may be those companies that truly provide practical applications and AI tools, making AI available on everyone's mobile phones and computer terminals. Companies, whether in industries like healthcare or entertainment, they will be the broadest "nerve endings."
Taking this as a starting point, Lightspeed screened companies that had at least 10 employees and had obtained some external financing. In the end, there were 11 names worth paying attention to as the "rising stars" that promote the application of AI technology in practical scenarios. They are:
Pictor Labs — Accelerating disease clinical research through digital pathology (AI-assisted histopathology)
Endpoint Health — Provides AI-assisted precision treatment for immune-related diseases
Otter.ai — AI-powered audio transcription with new chatbot meeting assistant
Bria — Commercial-ready image and video content control
Wellsaid Labs — text-to-speech and dubbing
Subtle Medical — AI-driven image processing for radiology
Copy.ai — Marketing copywriting via chatbot
Lexion — AI contract assistant for sales, procurement, HR, legal, and more
YOU — a search engine with a chatbot interface
OctoML — Provides cloud-based tuning and running of LLM models
Runway — video, images, and other creative tools
The talent competition among these 11 companies so far is as follows:
Different from the "Top Eight", the talent backgrounds of the 11 "Stars of Tomorrow" are more diverse. Among them, the University of Washington has become the most prominent talent training place, and the domestic Tsinghua University has also contributed 5 talents to these 11 companies. Top talent.
Silicon Rabbit checked other information and found that the depth of the University of Washington's use of generative AI technology has reached the next level. This is mainly reflected in the use of generative AI to create new proteins. In 2022, the famous protein research scholar and professor of the University of Washington David Baker led his team to release RFdiffusion, using diffusion models to design a variety of new single-chain proteins and complexes that are completely different from natural proteins.
Continuing to track the previous employers of technical staff at 11 companies, the University of Washington appears on the list again, as shown in the chart above.
In terms of geographical coordinates, Lightspeed specifically mentioned that the “rising star” companies Wellsaid, You and Runway are all headquartered in Seattle. Perhaps, Seattle will be the new landmark for the next generation of generative AI companies.
The report ends with a key statistic: According to Pitchbook, investment and financing in the AI field in the first quarter of 2023 was approximately US$1.7 billion (excluding Microsoft's US$10 billion investment in Open AI announced in January).
Another Pitchbook data for reference is that in 2022, there will be 78 investments in the generative AI track, with a total financing amount of more than US$1.37 billion.
Although this report is more concerned with the analysis and presentation of the current situation rather than the prediction of future technological development trends, it is not difficult to see from this data reference that the field of generative AI is in its golden period, and the "top eight" still need to Further breakthroughs in the training methods and algorithm optimization of generative AI, and the "rising stars" extending generative AI to more fields, have just begun.
Babbitt Park is open for cooperation!
Chinese Twitter: https://twitter.com/8BTC_OFFICIAL
English Twitter: https://twitter.com/btcinchina
Discord community: https://discord.gg/defidao
Telegram channel: https://t.me/Mute_8btc
Telegram community: https://t.me/news_8btc