"heavy" artificial intelligence, lee and innovation works in the direction of the next few years.However, he faces an important problem: a startup now AI, AI scientists is the core, and "wen can pen, Ann wu can mount qiankun" AI scientists rare, in his words "of the entrepreneurial venture".
at this point, the industry in the face of the move.That is: how to turn an ordinary AI scientists to "entrepreneurial hero" .
as 30 years ago began to research in artificial intelligence, lee, feel rocks "" technology innovation works has the ability to push this move, and in this step is to obtain the stable strategic advantage.
lee told lei feng network,
AI entrepreneurship is now scientists all over the world, followed by mathematicians in the world, is the ordinary people of the world in the future.
this is lee in a white paper on the strategic innovation works artificial intelligence sharing conference behind closed doors, lei feng network reorganize its becoming "ten li: AI venture truth", presented to the reader.
AI scientists are super otaku
main institution of the innovation works itself is investment and investment institutions, after is, of course, we see the project, the founder, they have the idea, direction, we will use the funds to invest in it.
the past Internet startup mode, are already very classic "lean startup" description:
a few kids do a product up casually, can finance financing, cannot pull down financing.How to reach users, iterative products, after the liquidation, become a classic pattern.
this business model, its dividend era is over.Of course there will be later, but not as much as he used to be.Entrepreneurship threshold greatly improved because of artificial intelligence is the next batch of entrepreneurial orientation, and entrepreneurship in artificial intelligence is the core of characters is AI scientists actually, no AI AI scientist was impossible.
AI scientists tend to be super otaku, oneself curtilage inside the room, all day doing the experiment, suddenly you to throw him to a brutal terrible world, his own entrepreneurial success rate is not high.
a lot of AI scientists generally've never thought about entrepreneurship, now suddenly want to start a business, and then find themselves exceptionally long long board, board special short:
he perhaps technology very cow, but may perform enough;
maybe his product demonstration is very good, but it is a Bug;
may also he products do well, but don't know the market;
or understand market but don't know how to sell.
especially AI itself is a business of ToB, so not so easy to save a bureau.So the AI scientists need to understand business, understand ToB, he needs an engineer.
AI entrepreneurship "not good"
we will speak the AI business very good at ordinary times, today I will tell you about the AI is not good.
the first is: AI scientists have short board.
this've said just now, what do we want to help "otaku" make up a short board.
the second is: AI entrepreneurship is very expensive.
just to say "lean startup" is very cheap, because a few children don't get paid, with the zero element can be pushed out the first App.
we've just invest in a company, in a month after the money is used up.I said you don't just how money ran out of eight people, gave you millions.But, they say, buy machine in three million.
the third is: AI need data.
picture recognition, at least need hundreds of thousands of pieces of sample data, or even hundreds of millions.Who gave you get data?
so do artificial intelligence investment has a very big head of places: the top is cast out.
we will sweep around the past two years, from the most powerful team of unmanned company cast in two, not two.Then I can't find a team, because so many qualified people.We do
Internet finance, probably after Saul threw three, then there may be one or two missed opportunities, missing one or two, and then not.
because AI scientists have so many, to be able to get things make there is only so much to a point.
the present situation of the AI is "not enough".Everyone to grab the few trees, has taken the tree arch to the sky-high.I think AI the land needs, at the same time, instead of those very few crops.
so we set up the "academy" artificial intelligence. we may spend tens of millions of the machine, and then help 10 20 venture company;We get the data from various channels, AI scientists can do test;We try to get more potential AI scientists, can consider to start this road, to help them reduce the possible failure rate of 95% to 40%, so we are able to produce their own value.
of course, those who take refuge in innovation works, we can help you to solve all problems, also require their return.Could have five million account for 10% of the shares, may now give us 15%, we think it will be enough.Later if you can make a unicorn, we are have a lot of returns.
this academy in get money, at least have to spend two hundred million yuan.But if we fertilizing, presumably than those "crop" would like to us.
two or three years later, the AI will spread like Android
in the long run, is really forever by AI scientists to entrepreneurship?Not really.
any technology has a development path, a good example is the Android.When we do with the CSDN noviceblue mobile developers conference.Meeting for the first time, I asked the audience, how many people are optimistic about Android?There are about five hands.I ask how many people watch the Symbian?Five hundred hands.
but we firmly believe that Android is the way of the future.Just because the platform is not enough.Now the inside of the university of Android, iOS, training course is very popular.If you are a computer student, your self-study, to do training courses, a few months you can start to make the Android.
AI is also such a state.How long will it take to
?Let's bold hypothesis two or three years.Two or three years, the our academy incubate scientists would be a very unique and valuable method.Three years later platform out, a lot of intelligent students can self-study.More and more platforms, tools, the AI will become more and more easy to use.
after the young people to start, I feel more than a startup now scientists may be able to succeed.Because business requires motivation, have sex Wolf, willing to hard.Have to his reputation, worth all gambling.
qualified people can be AI engineer six months, a qualified person refers to: mathematical genius
an old professor, with thirty years of work force out a new algorithm is proposed.This is possible.
but can true power is still young.Many young people do not only suffer from a platform.
I tell you a secret.
if you are an eligible young man, we need only 6 months can take your training to become a AI engineer. twenty years, is not you imagine thirty years.It's not like a materials scientist, rocket experts, the expert is really need the work force of 30 years.
so, what is qualified?
unfortunately, not all of us. "eligible" is simply: mathematical genius.
of course, it also covers statistics, automation, computer.So much of China's population, light is a mathematical genius we should be to produce a hundreds of thousands of year.
imagine thousands of maths genius, there in the face of AI may be interested in fifty thousand.(because the Chinese students are particularly willing to go after one of the hottest thing, what is the most popular definition? Cool, can make a lot of money.)
there are twenty thousand exposed to some training platform, it took six months to do, in this twenty thousand people may have two thousand is suitable for leaders.For example, he is in the field of AI lei jun, Fu Cheng, etc. These people.
the two thousand people in the end is the best investment.Our job is to make these people showed up.
so short that we are holding the scientists to, another three or four years, we are going to put these young people are trained.Let their cognitive this is the best time. so the secret is: we are going to dig all the maths genius in China, and then guide them into AI entrepreneurship.
AI take over human?Our problem is look much science fiction
how should we think about the AI?
someone see afar dog defeated lee se-dol, lenovo to the AI to take over in an instant by human.In fact, the worse.
one of the most difficult problems in the AI, is across the field of natural language understanding.To do this, it is necessary to the understanding of the context, requires interdisciplinary knowledge, also need the human "Common Sense".
like I suddenly and you say: "good noon also didn't eat hamburgers, McDonald's doesn't taste good."This sentence everyone understand what meaning, but the machine is difficult to read. it can put every word recognition is correct, but still unable to "understand".
again, for example: iron open not to touch, to touch the hand can't touch water electricity.These things don't speak, we all know.But how the computer know these things?
how do you go to teach a computer interdisciplinary knowledge?How do you teach it desires?How do you teach it what is beauty?What is love?What is religion?What is faith?These things are still very far away.
speculation may happen with sure that will happen, the two is to distinguish clearly.AI can't do anything just told, we all can't fathom how long will be breakthrough.Some say five years, some people say that fifty years, others say never.
I think we should really be discussing how to create value with AI, how to let the human to have no hunger and cold, let every man to live with dignity.
in the future, for example, many blue-collar and white-collar jobs will be replaced, also includes a reporter.Of course some in-depth articles machines may also wrote out for fifty years.But if you get some information from the Internet, for example, hkust fly earnings release, products for more than 30%, analysts said the stock how, the future of artificial intelligence is favored to what of, this kind of machine have been written.
when machine can replace the simple work, when after five seconds to think of things people don't have to do it, when so many people will be unemployed, the unemployed should be how to do?How do we go to training them?What is the child's education?How to let people keep looking for the right thing to do?Maybe the creator is don't want us to do the boring work, let's do something meaningful, hence the machine to replace us.
just tell these things will happen in 10 years.
to feed the world's future may also be AI, of course, we may all be AI pet, wearing VR headsets to play games at home.Machine will have a sense of self, will replace people, will not become a species, although not impossible, but these are unknown.
unfortunately, we see much science fiction.
"new species" AI "to replace" slavery ", "these can be imagined, of course, but there were more interesting question inevitably, more worthy of our thinking.
AI "low-hanging fruit" hasn't picked the
there are a lot of school of artificial intelligence.Symbol school, connecting school and so on.But in addition to the deep learning method, after years of being validated, it is not too much development.
simulation analysis method, want to turn it into a rule and expert system, over the past 50 years have proved that this idea is no good.Maybe one day, of course, there will be a breakthrough, but until that day should be no good.
in terms of my own background.In 1988, I began to do the voice recognition.The first set of system is made of complete machine learning methods to speaker-independent speech recognition.
now it seems this is a very small way: there is a person in the world can read voice from a paper, my tutor will put this method into a set of expert system.
as I really firmly believe: the structure of the machine with the human brain, with people's way of thinking is not the same. we stubbornly put A to B is difficult, as we can't force yourself to become A depth of learners, and to analyze things - our brain thinking, it is not like that, it is not natural.
if the scientific methods produce artificial intelligence, is an unknown area.Unknown things has its charm, research should be unknown, you need to do you want to have a breakthrough innovation.Everything you do in the academic field measure is: I want to do something that others never done before.We can assume that the brain AI is related with the future, we can go to prove that it is or not.But from an investment perspective, bet the risk is too big.
the deep learning is also because of the lack of data, met some bottlenecks.But in recent years we have seen several particularly big change:
the first is the special began to produce large amounts of data in some areas, and I think we now have not been finished.
the second is the use of GPU allows us to more efficiently, do deep learning very quickly.
now I think, the depth of the so-called learning is far from being picked the fruit. for the application of artificial intelligence flowers, one by one big fruit in front of you.In this case, you should go to the flower, why?
we sweep the GPU and huge amounts of data in the world again, should be enough for our VC industry for five years, so it is very clear from the Angle of investment.
go down again, I don't think we AI can be only deep learning.For example there are enhance learning method, also in the exploration.AlphaGo not only exist inside a method.So I think academics should begin to help and to explore the possibility of more actually, when we ate the food in these two years after may have a better chance.
I don't have AI religion
of course AI or may not have further breakthroughs in the future.
if not, then the golden age of AI in the past.Here is the Internet of things or other something.As investment institutions, religious belief, we don't have an AI we will meet the flexibility.
as the mobile Internet era, we should be most of the mobile Internet VC in the industry.But then we adjusted according to the situation.
if the academia and industry have a reasonable division of labor, I am very optimistic about the future five years the investment community and create value, for the so-called AI bubble I think will not happen.Individual cases will be a bubble, of course, but I think I could eat food is too much.
it academic and industry division of something like this:
the division of Labour on the one hand, is a very natural organic;
on the other hand is a little bit envy envy hate inside.
in general academic is looked down upon by industry, but at a certain moment suddenly a mature technology with industry, academia on this technology is can't do the achievement of the industry.So the academia will be compelled to do something new.For example: face recognition to do again now, academia have been a dozen however industry.So in the field of artificial intelligence, rarely see a old professor only research a proposition.
AlphaGo itself has no business value
AI will bring us what value?
I want to say first AlphaGo.AlphaGo so compelling, largely because our experts speak it too suspension.
before I think weiqi is more difficult than chess at least ten years or fifteen years, but then I turned out to be too pessimistic.There are many reasons I too pessimistic.I thought go than chess is a astronomical figures, but the astronomical Numbers also.
the best before AlphaGo ai player reached the spare five sections.So AlphaGo new Master and professional nine section of the gap between roughly equivalent to the gap between professional and amateur nine nine.This is indeed a great leap.
why is there such a phenomenon?That is to say, why go under artificial intelligence improvement so big?
actually has a very realistic reason, is the man who want to earn money not to go. you see AlphaGo expert team also didn't so great, is twenty great expert in machine learning.In Google may have two thousand people, there are one thousand people like this in the Microsoft.The reason is that Microsoft and Google in the past did not want to take the power of two thousand experts defeat go player, more of their time doing voice recognition, face recognition these valuable things.
on this worthless things, can use 20 experts is good.
finance, health care is a commercial AI
AI of commercial value, the influence is huge.
AI most easily applied in the field of a large quantity of data. the data is accurate, the best automation annotation.
AI the most easy to be applied in the field of frictionless. a domain if there is a manufacturing, testing, logistics, such as friction, that'd be in trouble.What is frictionless field?Health care is no friction, finance is no friction.
AI in the field of to earn the most easy to be applied. there is no doubt that the most money and finance.
and financial will undoubtedly is fastest AI field of conquest.Because your algorithm can quickly turn into money.
medical is a particularly large areas.And health relative to traditional, can produce a great chance of value-added.And it is not based on big data.What is the best doctor, is he himself is a deep learning machine, based on his experience to do a lot of good times.
suppose he judges the five thousand patients, about a lot, found some wrong, the following his judgment will be very accurate.But a good doctor may judge at most five thousand patients, but our data is the patient's level of fifty million.So health beyond the doctor should is a very necessary, a global trend.
but AI medical need to break through some privacy, there may be some challenges.
robot world, want to rely on intelligent driving to tap
in addition to the big data applications, as well as the application of the science fiction type.Including robots, unmanned driving such areas.
now look very clear, and global reach consensus is unmanned.Sometimes you have to do a science fiction things, need is ready, right place, right time and to push.But once you start moving it.As our mobile Internet transformed the whole industrial chain, the previous SP, nokia and so on.This industry change, basic old enterprise will all die, with a new batch of.
travel will be the next industry.We are very lucky, currently has a Shared economy, and electric vehicles.These two areas has been promoted, can promote the process of met with some resistance.
now, unmanned can change the world economy.I believe that 10% of the world economy is associated with travel and transportation.Although the real unmanned arrival may be ten years, but some other things can be much faster.
such as scenic tourist car, such as delivery trucks.
you might ask, if automated driving technology is not mature, the truck down the highway?No problem, we have stopped all the warehouse beside the highway is not.
one thousand trucks wrong way?Then we can to build roads, on the way to put a lot of signs and sensors, this also is not very difficult.
so we can play our future three to five years a lot of patches, let unmanned can be used under the environment of many limited, so don't think autopilot has ten years, now has nothing to do with us.
we seldom see a industry from beginning to end all "surrender".
a car company also dare not say unmanned?Each are desperately trying to go to solve, the entire industry forces came in.
the power of the capital in the world are investing in unmanned company.
new entrepreneurs, many business in the field of unmanned.
this is an irreversible trend, to do the layout of the new industries.
, for example, all drivers do?No car to stop, the parking lot?After the car the what?Road to provide sensor?Where is the fastest to make the most money?
these we actually don't have to worry too much, because the most business sense and ability of science and technology are already in gauging the this thing every day.They, or we will find a solution.
when a driverless car can run on the road, the car can dialogue.The traffic accident in front of, for example, my car to give way.My master was in a hurry to work today, you gave me out of the way, I give you two cents line not line?
in this case the robot becomes feasible.The robots at home and look forward to playing with the kids the way evolution, it is better to look forward to driverless cars to promote the evolution of the robot.
more artificial intelligence related content, please focus on lei feng network.