Win10 系统的小伙伴 all know, there is a called “Cortana” assistant, is our usual search document click that small circle. When it landed in China at the end of July 2014, it was affectionately called Xiaona, which is a voice AI like Xiaomi’s Xiaai and Apple’s Siri.
However, while Xiao Ai was still developing the market for Lei, Xiao Na was “abandoned” by Microsoft at the end of 2019. Why Siri has learned cantonese today, but it is difficult to insist on na down?
1
Lack of data
Compared with PC, mobile terminal obviously has greater demand for voice interaction, and most of the voice interaction scenarios occur on smart phones. Microsoft, which is tiny in the smartphone market, has a distinct disadvantage.
Lack of data is obviously a fatal flaw in AI.
2
Competition is fierce
Amazon’s Alexa, Google’s own Assistant, and Assistant software from phone makers across China are all serious competitors.
Fierce competition led to a decline in market share, which in turn affected the allocation of funds within Microsoft, resulting in less competitiveness.
This vicious cycle continued until Xiao Na left.
3
Talent shortage
The job requirements of AI engineers are notoriously high, with extensive and complex knowledge, not only on mathematics, programming fundamentals, machine learning algorithms, but also on the development ability of engineers. \
Moreover, the recent hot CV and NLP have taken away a large number of excellent AI engineers, and the talent gap is very large for speech recognition projects.
A number of AI unicorns have struggled recently, and even companies like Microsoft have failed to protect their voice AI. So some time ago, SOMEONE asked me, will AI still be around in 2020?
In fact, it is not advisable to generalize. Today, as CV and NLP industries mature, AI is showing great strength in many fields such as speech recognition and voice print authentication.
Back when visual recognition was in the ascendancy, a lot of people were more interested in voice AI. Compared with facial recognition which requires camera and sensor, speech recognition which only requires microphone seems to have more development and application potential.
In the 2019 and 2020 technology reports, Dharma Institute has raised high expectations for voice AI and cognitive intelligence. Undoubtedly, voice AI will be another hot topic of artificial intelligence after CV and NLP in the future.
Recently, a lot of policies conducive to speech recognition have been introduced, so that AI in the field of speech has been expanded to finance, vehicle and other fields, and the visible prospect is broader.
Based on its strong accumulation in Youdao Dictionary, netease specially built the module of intelligent speech recognition in the four major projects of AI engineers. The project fully demonstrates the speech recognition technology of multi-language recognition and multi-scene application, which can not only get a more complete project experience, but also make people have a clearer understanding of the future work direction.
Here I recommend the free artificial Intelligence Courses launched by netease Cloud Classroom. This course is designed by artificial intelligence experts from well-known enterprises, combining the application experience of the industry to design the learning path, and there are lecturers and teaching assistants to answer questions for you for free, really help you from entry to mastery!
There are 2 free live classes per week. There are also replays if you miss the live time. Scan the qr code at the end of the free booking course!
The following benefits are free for a limited time:
PART 1
Free live classes
▼
Live theme
Development Plan of Annual salary of 50W for Deep Learning Engineers
20:00, Tuesday, March 17
Live broadcast of the outline
1. Current Situation and prospect: Deep learning and essential competitiveness for employment
2. Field of study: Computer vision and natural language processing
3. Extend: The importance of academic papers and benchmarks
Live theme
A Must-see introduction to Artificial Intelligence — Using capsule Network to Make picture classification
20:00, Wednesday, March 18
Live broadcast of the outline
1. Basic concept: capsule network
2. Comparison of advantages: capsule network and convolutional network
3. Practical case: Realize MINST data set classification
Scan code to add wechat, booking free live ↓
PART 2
Class video
▼
“In this video, we’ll take you through the 1×1 convolution kernel in convolutional neural networks“
1. Principle: Convolutional neural network for deep learning
2. Code presentation: Implement convolution kernel based on Python
3. Case Study: Inception Network from Inception
Machine Learning Case Study
1. Problem: data plan to solve unbalanced samples
2. Methods: Model evaluation and standardization analysis
3. Analysis: Application cases of logistic regression model
4. Operation: : strategy comparison scheme actual combat
PART 3
Artificial intelligence learning materials package \
▼
AI Industry Report
▼
Class assignments
▼
Get the way
Scan the code to book a live broadcast
Get free course videos