Image source: Generated by Unbounded AI
In the generative AI boom of more than half a year, all kinds of chatbots undoubtedly occupy the C position.
Letting AI generate text, pictures, and videos by talking to chatbots seems to be the most popular path for more people to get in touch with and experience generative AI technology.
But a generative AI company called Adept AI did not follow suit and chose to "roll" on this track. They believe that the language model can not only generate text content, but more advanced gameplay is to let it create and execute "actions", become a new generation of operating systems, and innovate the interaction mode between people and computers.
The founding team of Adept AI can be described as "burst". Its co-founders include two well-known scholars, Ashish Vaswani and Niki Parmar, who proposed the Google Transformer architecture and the authors of the "Attention Is All You Need" paper, and David Luan, a low-key Chinese technology tycoon.
After its establishment in January last year, Adept AI, relying on its super-strong team technical background and innovative business model, has received investment from well-known VCs such as Reid Hoffman, the founder of LinkedIn, and Andrej Karpathy, the head of autonomous driving at Tesla. In March of this year, Adept AI, which has just been established for more than a year, announced that it has received US$350 million in financing from Microsoft and Nvidia, and its valuation has exceeded US$1 billion in one fell swoop, making it one of the hottest generative AI unicorns currently.
01
founder story
As mentioned at the beginning, Adept AI has an extremely bright team of founders. Both Ashish Vaswani and Niki Parmar had worked in the Google AI research department for many years before. As the two most important proponents of Transformer, their joining almost made Adept AI explode at the beginning.
Image via arXiv
However, after staying in Adept AI for about a year, Ashish Vaswani and Niki Parmar devoted their main energy to another rather mysterious new entrepreneurial project, Stealth Startup, but this did not stop Adept AI from moving forward. After Ashish Vaswani and Niki Parmar left at the end of last year, the management burden of Adept AI fell on the shoulders of David Luan, another Chinese co-founder.
Many people may not have heard of David Luan before, but he is a real technical expert, and his personal experience is also legendary.
David Luan moved to the United States from China with his family at the age of 6. He has shown the characteristics of a "gifted child" since he was a child. He has been studying university courses since he was 8 years old, and became a member of the first Thiel Fellowship at the age of 19. Thiel Fellowship is a scholarship established by Silicon Valley billionaire and founder of PayPal Peter Thiel, which specially selects young entrepreneurial talents under the age of 20.
At that time, David Luan was interested in robotics. Since high school, he has tried to independently develop the control system of intelligent robots. While studying at Yale, he took two years off from school to create a robot application store project. Later, he focused on deep learning technology and founded the artificial intelligence company Dextro, whose main business is video. Automatically classify and segment scenes and scenes, and was later acquired by Axon, a listed police security company.
Later, David Luan continued to work on deep learning technology. He joined OpenAI in 2017 and was deeply involved in the development of GPT-2 and GPT-3. In 2019, he joined Google Research as the technical director. He has since joined Apollo's Impact Advisory Committee as an advisor. In late 2021, he co-founded Adept AI with former Google colleagues Ashish Vaswani and Niki Parmar.
Image via LinkedIn
02
Main business
The current main goal of Adept AI is a general-purpose operating tool based on generative AI technology, allowing users to give instructions by using voice or text, and the artificial intelligence will understand and help complete various operations and tasks.
To put it simply, the goal of Adept AI is to build a new operating system or platform based on generative AI, completely changing the process of people using applications or web pages for software operations in the past. At present, this product is mainly realized by the large model Action Transformer (ACT-1) developed by Adept AI.
So how can ACT-1 play a role in our daily use?
In the example given by Adept AI, ACT-1 exists in the form of a desktop dialog box. You can call it on the computer at any time, users only need to input commands in the text box, ACT-1 will automatically complete all operations for you step by step.
For example, if you want it to help you find a house in Houston, suitable for a family of 4, with a total price of less than $600,000, Adept can quickly understand and capture key information and then start to execute operations. The difference from Microsoft's Copilot is that Copilot only provides you with information, while Adept can directly and automatically perform actions on the page, including jumping, clicking, and entering numbers.
For some professional software, the operation process in the past may be very complicated, requiring the user to have professional operation knowledge to complete it correctly. But now you only need to issue natural language instructions, and Adept can automatically perform the entire wipe for you. For example, operations that used to require at least 10 clicks in Salesforce can now be done in just one sentence.
Not only for a single software, the ACT-1 model can also perform operations across software, automatically combining multiple tools according to the task. For example, if you want to search for an item on a second-hand platform, it can automatically open the mailbox to help you write and send the email after it automatically screens out the desired item for you.
In addition, Adept also has extremely high self-adaptation and self-correction capabilities. It can find optimal solutions and perform operational corrections based on user input instructions and feedback.
Since the operation of Adapt AI is driven by artificial intelligence modules rather than applications, different functions can be developed and updated independently, which greatly improves flexibility and scalability.
Adept AI believes that in the future, most of the user's interaction with the computer will be done using natural language rather than a graphical user interface (GUI). We only need to describe to the computer what we want to do, and it will do it automatically. In the age of artificial intelligence, today's user interface will soon be as obsolete as a landline phone.
Adept's products have not been officially released yet, but the application for user experience has been opened. Interested partners can apply to join the trial list on its official website. In addition, according to public data, Adept currently has only about 25 employees and is recruiting more technical personnel in Silicon Valley.
03
financing experience
After its official establishment in January 2022, Adept AI has only conducted two rounds of financing so far, and each round of financing has a large number of well-known investors competing to participate.
In April 2022, Adept AI received $65 million in Series A financing. Investors include venture capital fund Saam Motamedi, LinkedIn co-founder Reid Hoffman, Tesla Autopilot head Andrej Karpathy, Skype early developer Jaan Tallinn, and Stanford University computer scientist and Lattice Data co-founder Chris Ré, among others.
In March 2023, Adept AI announced that it had raised $350 million in Series B funding. The round was led by General Catalyst and Spark Capital, with other participants including Greylock, Atlassian Ventures, Microsoft, Nvidia, Workday Ventures, Caterina Fake, Frontiers Capital, PSP Growth, SV Angel, and more. According to Forbes, Adept AI is valued at at least $1 billion after this round of financing.
Image via Crunchbase
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
