Google Brain and Courser co -founder Andrew from ‘S AI fundLaunched in 2018, it is a enterprise studio, which focuses on building corporations from scratch and investing in startups, which he had in the field of co -founder. At the starting of this month, the fund announced a second fund in the amount of $ 190 million, which incorporates a number of high -profile corporate investors.
On the heels of this announcement, we caught up with NG his experience, building a enterprise studio with generative artificial intelligence and how the approach of startup co -founders works.
“I am personally involved in every construction of the company,” said NG. “I think that many VC claims that they are operators. But we are operators to such an extent that we write the code with our portfolio company.”
The AI fund will co -show the company with general directors, actively participating in the strategy, coding, discovering clients and others. Compared to the first fund – in 2018, USD 175 million – its second includes more corporate investors who provide precious observations and access to specialized industries, equivalent to renewable energy.
NG, which is also an adjunct in the field of computer science in (*7*)Stanford UniversityHe has been working in the artificial intelligence sector for 33 years, starting with AI’s internship when he was 17 years old. He founded the Google brain syndrome in 2011 and led it for two years. Then he co -founded Courser In 2012, cooperating with universities to expand access to education and is still on the company’s management board. NG joined the Chinese search engine Baida In 2014, as the foremost scientist, building the instrumentation of AI. In 2017, he founded Landywhere he stays the chairman, focusing on visual AI solutions particularly vital for production. In the same 12 months, the co -founder Deeplearning.Ai for technical training in generative artificial intelligence, and soon after he founded the AI fund.
Most of his time now spends at AI Fund and Deeplearning.Ai, which works with many leading AI generative laboratories to build courses. In the case of the recent development of reasoning and multi -stage NG, the wider date “Agentic ai“To create a more integration understanding of” agency systems that use design patterns such as reflection, use of tools, planning and cooperation of Multiagent to get better results than Zero’s hint. “
The AI fund has 25 team members and began investing from the second fund. Until now, a co -founder of about 35 portfolio corporations, with some in secret. Investors in the recent fund include many corporate investors – AesIN HPIN MitsuiIN MitsubishiIN Qbe AND Telus Ventures – and also Venture Capital sequoia AND Nea.
During our conversation, NG spoke about several portfolio corporations, including Gaia dynamicswhich helps corporations comply with the tariff. The AI fund began cooperation with the company before the presidential election in the USA, expecting tariffs and trade, becoming more complex.
He also mentioned You have skyfirewho works on the implementation of drones and runs the program of the first responder who attributed to save lots of the life of a policeman, and Profitwhich helps retailers value products by automating comparisons with competitors to optimize prices.
Tell me about the AI fund and what is different in it.
WITH: Unlike traditional VC, our basic activity is not a transaction flow competition. Our foremost business activity is to discover promising ideas for startup, confirm the needs of the market and customer needs. Then we recruit the General Director to cooperate with us to build a company.
My team is there with CEO, writing a code, making connections with the discovery of customers. We are actively involved in debating the company’s strategy, which is to determine priorities over the next week, with which customers can contact and what is the right price. We are there to establish a company with CEO.
In the case of fund 2 you have a lot of corporate investors, while there are not so many for Fund 1. Why this modification?
WITH: It seems that the basic part of the strategy is to discover precious ideas for a startup. We really wish to cooperate with corporate partners who are interested in building corporations with us.
For example, one of our limited partnerships is Aeswhich is the largest producer of renewable energy for corporate clients. They make a lot of energy and energy buildings. AES will see applications in renewable energy sources. I do not have access to large sun and wind farms. It seems that there are many ways in which AI can assist in large -scale energy infrastructure. It is only due to cooperation with a company like AES can they take us to those large solar and wind farms and allow us to play in a sector where we have no right to play otherwise.
It seems that a significant fraction of our ideas at the startup comes from corporate partners who have noticed the market need, often in some sectors of the economy, which is very large, very vital, but completely foreign to a typical consumer or completely foreign typical AI engineer. I think it was interesting how often we play in these spaces.
We think it is extremely exciting, and no one else cares.
How is the project gathered to develop into a company?
WITH: We have more ideas at any time that we could work with a large margin. So our team will look at many of these ideas, and then determine which of them to verify, mainly on the basis of technical assessment in terms of what AI technology may and cannot do, in addition to a easy business grade.
If we determine to maneuver forward, we recruit the general director as the foremost co -founder next to us. We consider ourselves a smaller co -founder and we work very actively with CEO to build a company.
Our team is very supported by AI. My financial director, my legal advisor, my marketer – everyone knows how to cod. We are literally all builders.
I discovered that due to AI Weamnet, the band size needed to build something exciting is smaller, much smaller than it was. And the speed of iteration is also much faster than it was.
What is the success for the AI portfolio company?
WITH: Success could be peculiar business indicators-revenues and possible output-so that investing at an early stage normally perceives success.
It is completely different that co -founder of corporations. We use our resources, our technical knowledge, and our ability to recruit to startups. We introduce our resources in a more unique way in comparison with the traditional undertaking.
We run a recruitment service in a full cycle and commit ourselves to assist our corporations in recruiting their first key management. When a recent startup, which is unknown, reaches the director, it is difficult to call them back. When someone reaches out of the AI fund – because the AI fund is known – and supports the company, saying that we think it is really promising, we are capable of persuade very qualified management staff to significantly consider and often take opportunities.
This is actually one thing that we do for our startups, which we think is good when it comes to assist in recruitment. And then it helps us answer people we know that they are highly qualified management in corporations that we think are really very promising.
What are the key people you are looking for?
WITH: When building AI startups, a deep understanding of technology is very vital. It’s hard to overestimate. I think that the difference in performance between someone who really understands AI technology and individuals who have knowledge is a huge difference in performance.
It is not so in which “they are less good, but they will come up with it, it will take a little longer.” It works or does not work. It’s about not spending six months on a blind street.
We attempt to make sure that the general director and CTO are at least between them at the top of technology, because many recent possibilities create recent technologies that didn’t exist a 12 months ago and signifies that it was impossible a 12 months ago. Being near or at the latest, it is absolutely crucial in terms of the right moment of instincts about the proper architecture of technology. … Having this excellent technical judgment is really vital to avoid such errors. And you simply go much faster.
What are the skills needed for the startup that desires to take this technology and apply it to the problem?
WITH: We interview a lot of AI engineers. I imagine that two things are vital with generative artificial intelligence. One of them is the way you employ artificial intelligence in your individual work, and the second is the way the software uses AI. So man uses artificial intelligence, and your software uses artificial intelligence, and each are vital. As for the way a man uses artificial intelligence, programmers who know how to make use of AI encoding assistants are dramatically faster than those that do not do it.
I know that there is such a stereotype and there is a lot of truth. There are so many projects that might take me three months and six engineers – now I just build them myself at the weekend or one of my engineers at the weekend. This speed is really true.
I’ll share one thing more: there is a stereotype of a recent breed of native AI programmers, which is really fast and it’s true. However, I see that experienced engineers who are at the top of artificial intelligence are also much more productive than inexperienced engineers who are at the top of artificial intelligence. So I think that this recent stereotype is, say, a freshly graduate of College, which is really fast with artificial intelligence, and this stereotype is true. A freshly graduate of College, at the top of artificial intelligence, is much faster than a traditional engineer with 10 years of experience who is not at the top of artificial intelligence. But this fresh graduate of College is also significantly exceeded by someone from 10 to twenty years of experience, which is at the top of AI.
I have the honor to work with many of them, made at a speed that in my opinion the world didn’t see two years ago. And it’s just great how small a team can do various things. To be honest, we build something in the morning, we get user reviews in the evening, repair them at night. The next day we are in the next cycle. The speed is simply unbelievable. I adore it.
Any final thoughts?
WITH: AI technology is still improving. We see many possibilities in visual artificial intelligence, in building a voice stack and many efforts of data reinchnications, enabling everyone to learn coding. It seems that artificial intelligence is not one thing; This many various things create recent possibilities.