Shulou(Shulou.com)06/03 Report--

AI Product Manager growth path

The following are some summaries of my usual knowledge, just some personal humble opinions, the companies, books, videos and websites that appear below are all seen and experienced by themselves, not advertisements for them, not advertisements! Not an advertisement! Not an advertisement! Just leave your comments in the comment area with different opinions.

I. the rise of AI and the Internet

1. The dividend of the Internet disappears.

1) the number of users on PC and mobile is fixed.

At present, the annual shipments of PC and mobile are about the same value, mobile is about 400 million units a year, PC is even less and is declining every year.

2) large traffic entrances are carved up by giants

From the point of view of most people's mobile phones, there are only a few apps that people download.

Chat (QQ, Wechat), information (hot headlines, Zhihu, Weibo), takeout (ele.me, Meituan) and other software, basically mobile traffic has been taken away by these giants.

3) the cost of obtaining customers has been greatly increased.

This is no longer a time when a few people can work together to develop an APP. Basically, investors will ask, how do you get customers, that is, can you find the right people? How to get people to use it.

During the taxi-hailing war in 2015, Capital invested a lot of money, and both Didi and Kuaidi spent 100 million a year to get passengers, as did shared bikes in 2016. With the saturation of APP, even if there is a good idea, you need a lot of money to burn.

two。 The transformation of traditional industries by the Internet is limited.

1) Medical essence

The essence of medical treatment is that doctors treat people. However, there is one problem that cannot be solved by the Internet, that is, the number of doctors across the country has not increased because of the Internet. The essence of the Internet is to solve the problem of information asymmetry, it connects doctors and patients together, but in essence, a doctor can only see one patient, but the efficiency has been slightly improved, and the essential problem has not been solved.

AI can bring new possibilities. In the future, she can take the place of a doctor to see a doctor, prescribe medicine and so on. In essence, it improves the efficiency of seeing a doctor.

2) the essence of logistics

In essence, the problem of logistics is the same as the medical problem, which is also a driver driving a car, which does not fundamentally solve the efficiency pain point of the industry.

3) the nature of manufacturing

The production efficiency of the manufacturing industry has not improved greatly since the second industrial revolution. After the first industrial revolution, the manufacturing industry upgraded in an all-round way, accompanied by the unemployment of a large number of handicraftsmen, followed by the innovation of the industrial chain brought by steam power, while the second industrial revolution was the popularization of internal combustion engine and electric power, which brought new power. The manufacturing industry has entered an era of rapid development, thus bringing about the progress of productive forces to promote the development of society. The three revolutions brought about by the Internet are the information revolution, from which the manufacturing industry does not benefit much, while AI can bring new possibilities to the manufacturing industry and greatly increase productivity.

4) other similar industries include the taxi-hailing industry, the catering industry and the education industry, which have not been fundamentally changed by the Internet.

3.AI has infinite reverie.

1) A whole new world

A) New ways of interaction: voice interaction, video interaction, gesture interaction

B) New shopping methods: virtual fitting

C) New game experience: a new generation of immersive games brought by VR/AR

2) there are many opportunities

A) AI opportunities brought about by mass entrepreneurship and innovation

The mass entrepreneurship and innovation plan put forward by the country in recent years encourages the public to innovate and start businesses, and the country has recently included AI in the national development plan, which shows that AI will bring many opportunities to young people in the future.

B) AI can bring new business models

Looking at the 20-year development history of the Internet, we can see that every small innovation of the Internet will bring new business opportunities and business models. Godfather Jack Ma's creation of Taobao makes it possible for merchants and users to trade directly; the Tencent empire founded by Xiao Ma has brought a breakthrough in social style and made me understand the true meaning of "wool comes from pigs". In recent years, the sharing economic model and live broadcast economic model is an innovative business model, and China's shared bike is called one of the "four new inventions" by foreigners.

3) A large number of jobs

A) AI trainers:

This profession can already be seen on some recruitment websites, and the salary is not low.

B) Robotics consultant:

I think this is a position for robots in the future.

C) Virtual counsel:

I think that after the virtual industry rises in the future, there will be laws and regulations for virtual things (robots, intelligent products, virtual world NPC, etc.), and this profession will emerge as the times require. Similar, I guess there should be machine comforters and so on.

D) AI PM:

This should now be accepted by most technology enterprises, and it is also accompanied by AI.

4.AI will increase productivity on a large scale

1) the efficiency of seeing a doctor has been greatly improved.

AI start-up enterprise Yitu Technology is already cooperating with some three hospitals, and uses the AI system platform to replace doctors to analyze images such as tumors. In the future, it will assist doctors in more areas to analyze the disease and change the efficiency of a doctor per unit time in the past.

2) Unmanned vehicles to solve the pressure of traffic logistics

Nowadays, all the major first-tier cities have caused traffic jams because of the popularity of private cars, which has caused great pressure on the traffic and transportation of big cities. The development of e-commerce has caused the transportation pressure of logistics.

3) Intelligent robots promote manufacturing upgrading.

The high-precision repetitive operation of intelligent robot not only reduces the labor cost of the enterprise, but also reduces the risk of personal safety, on the contrary, it can greatly improve the productivity of the enterprise.

4) the development of economy drives the progress of civilization

The development of productive forces brought about by AI, which is the driving force of the progress of human civilization, will bring new opportunities and breakthroughs for human beings to explore the unknown world (starry sky, deep sea, origin of life, etc.).

5) Robots solve problems such as companionship, nursing, customer service, etc.

5.AI can create personalized service experience for users.

1) Taobao can match clothes size according to personal information.

It is well known that buying clothes on Taobao is to ask about the size and other information of customer service clothes. Later, you can use machine learning to establish the size information and style preferences of each user. Next time, you don't need customer service to answer the user's size information. You can directly recommend the appropriate style of the user.

2) Love and marriage websites create a circle of friends of interest according to users' social attributes.

Dating sites can use users' data to train machines to build models for user portraits, and then recommend matching dating objects for users.

3) peer-to-peer education promotes the dissemination of personalized knowledge.

On the basis of meeting the requirements of general education, people increasingly advocate individualized teaching programs in accordance with their aptitude, and only use AI to create personalized educational content for each user.

4) Baidu can realize accurate search according to user portraits.

This Baidu is already in use, so there is no need to say more.

II. Necessity of AI PM

1. Open Source of algorithms and data Resource Warfare

1) Free of Google and BAT algorithm framework

Future algorithms and frameworks are certainly free, which is the routine of big companies, "wool comes from pigs", and big companies make money from other businesses such as cloud computing.

2) narrow profit of technical services

3) Commercial withdrawal of data value

two。 Technical personnel are not suitable for the needs of business development

1) it is not the creation of technology but the realization of technology that promotes social progress.

A) Watt invented the steam engine and entered the steam age to promote the industrial revolution

B) Bell invented the era of telephone opening communication

C) Edison invented electric lights to free mankind from darkness

D) it is not the originators of the Internet, but the CEO of the major technology companies that promote the development of the information age

2) the thinking of AI technical personnel is not necessarily suitable for the discussion of business model.

3) Business landing requires a sense of innovation.

3. The landing of products needs professionals to discuss.

1) Baidu introduced Lu Qi

Baidu AI technology is in the forefront of China, leading with the other two BAT companies. However, Robin still invited Lu Qi to develop the landing business of AI for it in Silicon Valley, followed by the departure of a large number of technology cattle (Wu Enda, Yu Kai, etc.). The side proves the importance of AI product people in the landing application of future scenarios.

2) the landing of intelligent speakers needs more product thinking.

Among all the high-tech enterprises in the United States, Google is good at technology, and there are a large number of technical talents in AI, but as Lu Qi said, Amazon is the most successful in the exploration of commercialization of AI in Silicon Valley, because Amazon's smart speaker Echo is at least popular, selling the concept of AI, and letting users know that there is such a high-tech thing.

4. Future products involve all aspects (philosophy, psychology, emotion)

1) Smart Speaker (emotion)

The dialogue scenario in the intelligent speaker involves the emotional analysis between people, which cannot be handled by technical personnel. The so-called skill industry has specialization, and the exploration of these scenarios still requires product people who understand users to open up the market.

2) Virtual shopping (understand users)

5. The vertical application of subdivided industries needs the development of compound talents.

1) companion robots need emotional communication.

2) Intelligent investment requires financial knowledge.

3) Intelligent medical treatment requires medical knowledge.

4) unmanned vehicle business needs to understand humanities and law.

III. Thinking about future products

1. Future tools

1) Interactive tools

People use voice, machine vision presentation, watches, helmets, glasses, etc.

2) vehicles

It could be mobile focus platform, Starbucks, study.

3) Social tools

Virtual community, theme park, game world, etc.

2. AI+ industry

1) definition: there was no industry before the development of AI technology.

2) Features:

A) low barriers to the industry, starting at the same starting line as the giants

B) too few opportunities

C) Technical requirements are high, and entrepreneurial teams have to be high-end.

D) High requirements for innovative thinking

3) Application:

A) self-driving cars: before the development of AI technology, who dared to talk about self-driving without greed.

B) Intelligent speakers: before ASR and NLP broke through, no one dared to mention the concept of intelligent speakers.

C) Urban brain: breakthroughs in computer vision and machine learning make image analysis possible and help govern cities.

D) face recognition: this direction is purely the product of the development of computer vision and machine vision.

E) accompanying robot: one of the necessary steps for the development of robot in the future.

3. Industry + AI

1) definition: for an industry that has always existed, AI only brings about industrial upgrading

2) Features:

A) there are deep industry barriers and the giants have no advantage

B) be more friendly to startups

C) Industry talents who understand AI are more important than AI talents

D) more than 70% of the products landed on the industry + AI

3) Application:

A) AI medical images

B) unmanned logistics freight

C) AI data subscription

D) unmanned retail

E) AI security

4. Consideration of the way

1) Control mode

In the future, products will be controlled more naturally, from a small number of geeks to highly educated students to the elderly and children.

2) perceptual richness

A) input: multi-perceptual input, voice, gesture

B) output: language, image, behavior

3) orientation: from functional orientation to human-centered.

4) content

A) tagging and refinement of content

B) content personalization, user profile and precision

C) form: take voice and video as the main body (no longer actively looking for it)

5. Product form

1) Integrated cloud:

A) end: voice, vision, action

B) Cloud: data, algorithms, services

2) the form is convenient and natural-oriented.

6. Application category

1) critical applications

A) users have very high requirements and very low fault tolerance rates

99% accuracy means one accident every 100 times, and 99.9% means one accident every 1000 times, so it requires a lot of technology, so it is not very friendly to most startups.

B) Landing application

i. Surgical robot

ii. Medical image analysis

iii. Unmanned flight

iv. Intelligent dispensing

v. Self-driving

C) the technical requirements are very high, and it is impossible to handle without being high-end.

D) the project has a long cycle, distant commercialization and remote profitability.

2) non-critical applications

A) Landing application:

i. Intelligent security

ii. Face recognition

iii. Sweeping robot

iv. Accompany robot

v. Food delivery robot

B) the technical requirements are not high, and the general technology can be achieved.

C) the expectation of user tolerance is not high

D) the project cycle is short, the profitability can be judged in a short time, and it is more friendly and more opportunities for entrepreneurs who want to look for opportunities on AI.

Fourth, the product person skill tree

The history of 1.AI development

1) AI gestation period (1943-1955): the introduction of computing machines and intelligence

A) Minsky and his classmates built the first neural network computer

B) Alan. Turing proposed Turing test

2) the birth of AI (1956): several scientists at Dartmouth Conference (McCarthy, Minsky, Shannon, etc.) put forward the term artificial energy, and formally had the concept.

3) enthusiasm and expectation (1956-1973)

A) Simon proposed the physical symbol system

B) Samuel wrote a checkers program

C) algorithm invention

i. The proposal of Behrman's Formula: the embryonic form of reinforcement Learning

ii. The proposal of Perceptron: the embryonic form of Deep Learning Model

D) Establishment of artificial intelligence laboratories in colleges and universities (MIT, Stanford)

E) widely used in mathematics and NLP to solve algebra, geometric proof and English problems

4) the first cold winter (1974-1980)

A) logical provers, perceptrons, and reinforcement learning can only do simple tasks

B) the mathematical model was found to be defective

C) the interruption of government cooperation and the transfer of funds, the pressure of public opinion

5) rise of AI (1980)

A) the presentation of expert systems

B) the proposal of BP algorithm

6) the second cold winter (1987)

A) the performance of desktops produced by Apple and IBM exceeds the performance of expert systems

B) the United States Government Project Agency rejects AI as the next wave

7) Modern AI (around the beginning of the 21st century)

A) IBM Deep Blue defeated the chess champion in 1997

B) the blue brain project of the Ross Federal Institute of Technology in 2009 successfully simulated part of the mouse brain

C) big data led to the rise of deep learning

D) IBM Watson challenged the quiz show "the Edge of danger" to win the championship in 2011

E) AlphaGo defeated the human go champion in 2016

F) AI in 2017 was included in the strategic development plans of major countries

2.AI general understanding

1) basic computing power layer: cloud computing, GPU and other hardware acceleration, neural network chip

2) Technical framework layer: TensorFlow, Caffe, Theano, Torch, DMTK, DTPAR, ROS and other frameworks or operating systems

3) algorithm layer (machine learning)

A) supervised learning

i. Definition: the marked data is the teacher, the machine draws the model, and then outputs the prediction data results.

ii. Solve the problem

① regression problem

② classification problem

iii. Algorithm model

① linear regression model

② K-nearest neighbor algorithm

③ decision tree

④ naive Bayes

⑤ logical regression

B) semi-supervised learning

i. Definition: general knowledge uses unlabeled and labeled data training models for pattern recognition

ii. Solve the problem

① spam filtering

Analysis of ② Video website

iii. Algorithm model

① semi-supervised SVM (support vector machine)

② Gaussian model

③ KNN model

④ Self-trainning

⑤ Co-trainning

iv. Advantages

Compared with supervised learning, ① can save manpower cost and improve the ratio of input to output.

Compared with unsupervised learning, ② can get a model with higher allocation accuracy.

C) unsupervised learning

i. Definition: do not provide marked data to the machine, let the machine process the data and output the results

ii. Solve the problem

① association

② clustering

③ dimensionality reduction

iii. Algorithm model

① K-means algorithm

② self-coding

③ principal component analysis

④ random forest

D) reinforcement learning

i. Definition: a reward that will be fed back to the machine when the positive state of the machine-aware environment is transferred, which makes the machine learning towards the positive signal trend, thus maximizing the cumulative reward value.

ii. Solve the problem

① automatic helicopter

② robot control

③ mobile network routing

④ market decision

⑤ industrial control

⑥ efficient web page indexing

iii. Algorithm model

① K-rocker betting machine (single step reinforcement learning task)

1. ε-greedy algorithm

2. Softmax algorithm

② has model learning (multi-step reinforcement learning task)

1. Strategy Evaluation algorithm based on T-step Cumulative reward

two。 Strategy iteration algorithm based on T-step Cumulative reward

③ exempt model learning

1. Monte Carlo reinforcement learning

A) same strategy

B) different strategies

two。 Time series score checking learning

A) Q-learning algorithm

B) Sarsa algorithm

④ imitation learning

E) transfer learning

i. Definition: refers to the transfer of learning results from one area to another

ii. Solve the problem

① lifelong learning

② knowledge transfer

③ inductive transfer

④ multitask learning

Consolidation of ⑤ knowledge

⑥ context-sensitive learning

⑦ meta-learning

⑧ incremental learning

iii. Algorithm model: TrAdBoost algorithm

F) Deep learning

i. Definition: multilayer neural network

ii. Solve the problem

① predictive learning

② speech recognition

③ image recognition

iii. Algorithm model: RNN, DNN, CNN

iv. Advantages

① detects complex interactions from features

② learns low-level features from raw data that are almost unprocessed.

③ deals with high cardinality class members

④ handles untagged data

4) General technology layer

A) speech recognition (ASR)

i. Concept

① principle: input-encode-decode-output

② recognition mode

1. Traditional recognition: generally use Hidden Markov Model HMM

two。 End-to-end recognition: depth neural network DNN is generally used.

ii. Far-field recognition

① voice activation detection VAD: high signal-to-noise ratio (SNR) of far-field recognition

② Voice Wake-up: smart devices need voice wake-up words to make them work

Difficulties in ③

1. Wake-up time: the time it takes for the user to send a voice to the device to respond to the user (still a little longer at present)

two。 Power consumption: the current power consumption is not low

3. Awakening words: generally 3-4 words

4. Awakening result

A) fail to report: he should not call him (it is easy to fail to report if there are too many words to wake up)

B) false positives: did not call him he should (the number of awakening words is too small, it is easy to make false positives)

iii. Microphone array

① background: in the complex background, there are often a variety of noise, echo, reverberation to interfere with the recognition scene. At this time, microphone array is needed to deal with noise.

② action

1. Speech enhancement

two。 Sound source localization

3. De-reverberation

4. Extraction and separation of sound source signals

③ classification

1. Linear: one-dimensional (180 degrees)

two。 Ring: 2D (360 degrees)

3. Spheres: 3D SPAC

Number of ④

1. Commonly used are 2, 4, 6 wheat

two。 There is a big difference in pickup effect among single wheat, double wheat and multi-wheat in noisy environment.

3. 5 wheat and 8 wheat have the same effect in quiet environment.

iv. Full duplex

① simplex: a talks with B, B can only listen to A.

② half duplex: a (middle miss, attention down, over) B (down copy, over)

③ full-duplex: multiple rounds of conversation between two people, interruptable and interruptable

v. Error correction: error correction for identified statements

B) Natural speech processing (NLP)

i. Process

① NLU (Natural language understanding)

② NLG (Natural language Generation)

ii. Difficult point

① language ambiguity: meaning (what does it mean, which the machine cannot understand)

Robustness of ② language: many words in sentences, few words, wrong words, grammatical errors (this person often makes mistakes, and the machine can't fix it yet)

③ knowledge dependence: Apple (does this mean "fruit" or "mobile phone")

④ context: contextual analysis of context (she's gone-which one is she?)

iii. The solution (this is too much to elaborate on, in-depth students can check the data)

① rule method

② statistical method

③ deep learning

④ association method

iv. Application

French and semantic Analysis of ① sentence

② information extraction

③ text mining

④ machine translation

⑤ information retrieval

⑥ question answering system

⑦ dialogue system

C) speech synthesis (TTS)

i. Realization method

① splicing method:

1. Definition: from a large number of pre-recorded sounds, the basic units (syllables, phonemes) are selected and spliced together, and two-tone (the center of a factor falling down) is often used as a unit for coherence.

two。 Advantages: high voice quality

3. Disadvantages: the database is large, which generally takes dozens of hours of finished product corpus, and the cost of 50, 000 sentences for enterprise-level commercial users is in the tens of millions.

② parameter method:

1. Definition: according to the statistical module to generate every moment of the speech parameters, and then convert the parameters into waveforms, mainly divided into three modules: front-end processing, modeling and vocoder.

A) the tone, rhythm, prosodic boundary, stress, emotion of this sentence

B) both splicing method and parametric method have front-end processing, and the difference lies in the back-end acoustic modeling method.

two。 Advantages: personalized TTS mostly uses parameter method to save time cost.

3. Disadvantages: the quality is worse than the splicing method, because it is subject to the occurrence of the algorithm, there is a loss.

ii. Evaluation criteria (judging the quality of TTS system)

① subjective test: artificial evaluation (artificial listening)

② objective Test: system Evaluation (Machine Evaluation)

iii. Bottlenecks and opportunities

Lack of ① data (available voice data)

Lack of ② talents: TTS talents are too few compared with NLP and CV talents in AI.

It is difficult to turn ③ into product.

1. Users expect the scenario to be complex

two。 There are still many difficulties in the technology.

3. The detailed design still needs more consideration.

Pressure on ④ commercialization

1. The project cycle is long (this requires a long time of data and technology accumulation and precipitation)

two。 The cut-in in the subdivided scene is still in the early stage, and the cost of trial and error is high.

D) computer Vision (CV)

i. Development stage (four stages)

① Marr Computational Vision Phase

1. Computational theory

two。 Expression and algorithm

3. Algorithm realization

Active and target vision phase of ②

③ multi-view geometry and layered 3D reconstruction phase

1. Multi-view geometry

two。 Layered 3D reconstruction

3. Camera self-calibration

Learning-based visual phase of ④

1. Manifold learning

two。 Deep learning

ii. The process of CV application

① imaging

1. Definition: simulate the principle of the camera (how to improve the quality of photos)

two。 Factors affecting the picture

A) influence of light

i. Control from a product point of view: you can change the user's usage scene through reminders, and enhance the product experience by upgrading or changing hardware facilities.

ii. Control from the point of view of algorithm: use the algorithm to process the picture to improve the quality of the picture.

B) blur

i. Motion blur: caused by the movement of human body, vehicle and camera

ii. Blurred focus: caused by the distance, quality and weather of the camera

iii. Low resolution difference blur: caused by devices such as small image magnification and camera hardware

iv. Mixed blur: multiple blur exists

C) noise, resolution

Early vision of ②

1. Definition: the processing process of pictures

two。 Image segmentation

3. Edge extraction

4. Motion and depth estimation

5. Image stitching

6. Current problems

A) the results are not accurate

B) it takes a long time to precipitate knowledge

③ recognition and understanding

1. Definition: map a picture to a text, a photo, or a label

two。 Label

A) the more accurate it is for the model, but the less data

B) influence of subjective factors

C) subdivision label

3. Data optimization

iii. Research content (this part has not been summarized yet, those who are interested can explore for themselves)

① spatial vision

② object vision

iv. Typical object representation theory

Three-dimensional object representation of ① Marr

② Image object representation based on two-dimensional

Expression of ③ inverse generation model

v. Application development trend

① face recognition

② image search

③ personalized advertising

Real-time location and Map Construction of ④

3.AI product understanding (this section has not been dabbled in)

1) everyone is a product manager (the AI era should remain the same)

2) Product understanding

A) NLP class

i. Dialogue Robot (BabyQ of Turing, Microsoft Xiaoice)

ii. Voice search (Baidu, Google)

iii. Intelligent speech input method (iFLYTEK, Sogou)

iv. Smart speakers (Xiaoya stereo of Himalayan and Orion Star, Echo of Amazon)

B) CV class

i. Unmanned aerial vehicle (DJI)

ii. Medical image analysis system (chest CT intelligent assistant diagnosis system based on photo technology)

iii. Self-driving (Power Technology, Baidu, Google)

iv. Security and defense

C) Machine learning

5. Product people get on the car (they usually dabble in it)

1. Realize the importance of AI ideologically: realize that the AI era has indeed come, and mainly broaden your horizons.

1) Books

A) the singularity is approaching

B) A brief history of the future

C) the age of intelligence

D) artificial intelligence era

E) intellectualism

F) the extreme of science-- talking about artificial intelligence

2) Video (film and television)

A) artificial intelligence

B) I robot

C) Western World

D) Terminator

E) Matrix

two。 Theoretical knowledge

1) Books

A) Machine Learning (Zhou Zhihua)

B) Machine learning practice

C) the beauty of mathematics

D) Statistical learning methods

E) artificial intelligence-a modern approach

F) computer vision-algorithms and applications

2) Video

A) Ng Machine Learning course (NetEase Open course)

B) Yang Lan's AI talk show

C) the open course of artificial intelligence at Peking University (NetEase Yun classroom)

3) website

A) Zhihu (all AI Q & An and Zhihu Live)

B) brief book (all AI articles)

C) everyone is a product manager (the best community for product managers to learn)

D) 36 krypton (AI industry research report and AI news)

E) CSDN (you can follow the AI blogger to learn the AI knowledge of the system)

F) AI paper download network (some of which cost money, but small amounts for future investments)

i. VIP.

ii. Wan Fang

iii. China knowledge Network

iv. Google academic

4) Information

A) 36 krypton

B) Tiger smell

C) Geek Park

D) Business week

E) Zhongguancun online

5) official account of Wechat

A) Rice ball AI product manager base (this is the earliest AI PM community in the industry, where many AI PM trailblazers share practical information, but it costs a little money)

B) Intelligence Club (the ultimate of science-talking about artificial intelligence, this book is published by this club)

C) Quantum bit

D) expertise (with a lot of AI expertise and bosses' opinions)

E) AI technology stronghold

3. Organize output: be sure to output your own content after reading a book or video.

1) A brief book (output what you get and think)

2) Zhihu (output what you get and think)

3) CSDN (output what you want)

4. Research industry

1) Policy

A) National AI planning

B) Talent support policy

C) Entrepreneurial capital support policy

2) Market and financing

3) comparison of cities (north, Shanghai, Guangzhou, Shenzhen, Hangzhou and Chengwu)

A) comparison of talents

B) complete comparison of industries (rudimentary industrial chain)

C) Industry concentration and company distribution

D) AI atmosphere comparison

4) Segmentation of industry

A) Medical treatment

B) unmanned vehicles

C) Security

D) VR/AR

E) Robot

F) Finance

5) Company selection

A) chips: Cambrian, Horizon, Deep Science and Technology

B) NLP category: Turing Robot, Eitman, Spitzer, Yunzhisheng, iFLYTEK

C) CV category: Shangtang technology, far-sighted technology, Yitu technology, Yuncong technology, Malong technology, polar perspective technology

D) Robots: must-have, Turing Robot, Rokid

E) platform category (giant): Baidu, Ali, Tencent, JD.com, Xiaomi

F) Applied category: learning education, wisdom tooth technology, going out and asking questions

G) driving category: power Technology, Tucson Future, Singularity car

H) subclassification: Huiyi Huiying (Medical), fourth Paradigm (Finance), Quantification School (Finance), carbon Cloud Intelligence (Medical)

6) get in the car

A) selection of companies

i. Find the relevant list of AI start-ups and summarize the number of times the company has been on the list.

ii. IT Orange to find the company status of the relevant companies.

iii. Shangzhihu and retractor find employee comments and introductions of related companies.

iv. The working staff of the company related to Shangmai hook up know the details of the company.

v. Experience the company's products on the company's official website and summarize the output

B) Research related companies

i. Company positioning and main products

ii. Founder and team

iii. Strategic financing situation

iv. Company partner

v. Core technology of the company

C) get in the car

i. Send in your resume

ii. Take the research report and talk to Hr.

iii. Talk to someone with the output on Zhihu, Jianshu and CSDN.

iv. Get to know people in the industry and find people to push inside (AI product manager base, product community)

VI. Self-reflection

The way of interaction in the 1.AI era

1) interaction is more natural and convenient.

2) people use voice (gesture) to interact with machines, and machines use images to interact with people.

3) easier to carry (watches, glasses and other hardware products as carriers)

two。 The way machines exist.

1) tools

2) Pet

3) Friends

4) accompanying relatives

5) lovers

3. A new position

1) Machine trainer

2) Robot insurance / consultant

3) unmanned vehicle administrator

4) Robot 4S store

5) Robot dispute Resolver

4. Future consumption

1) unmanned retail

2) Virtual consumption

A) Virtual games

B) Virtual social interaction

C) Virtual travel

Author: release the Flying Man Night

Link: https://www.jianshu.com/p/4b98314ad3c0

Source: brief Book

The copyright of the brief book belongs to the author. For any form of reprint, please contact the author for authorization and indicate the source.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.