The Internet is not + good machine intelligence + the era is coming soon

This article was written by John Henderson, former business development and operations manager of Summly, published in Medium. The authors think about machine intelligence giving machines the ability to think, and the ability of the sensor technology to give the perception (such as the power to see) to the many vertical areas of subversion possibilities, in many areas of work may be replaced by machines within 5 years.

the next 10,000 startups ' business plans are easy to predict: do X and then add AI.

-kevin Kelly

The original Alan Turing thought that computers would think by 1950. In 1997, supercomputer Blue defeated the chess world champion Kasparov. In 2011, IBM's Watson became the best player in the Jeopardy knowledge race. In 2015, a car crossed the United States on its own.

These achievements are all dependent on machine intelligence, which is the general designation of artificial intelligence, machine learning, deep learning and other related fields. The computational power of geometric growth, larger data, and stronger algorithms, this wave of perfect storms has driven these areas to rapid growth in recent years. Thousands of machine intelligence startups are emerging (see the Machine Intelligence Company Overview ), many of which are ready to change the world in the next 5-10 years.

The most interesting of the teams and companies I have seen so far is the combination of machine intelligence expertise and those technologies like computer vision, natural language, or speech recognition. If a computer can think, and then give it the ability to read, read, or write, there is a surprising chance of being there. Suddenly, the computer can do all kinds of things that can only be handled by people before. There are two types of tasks worth considering: one is "communication task" and the other is "visual task".

Communication Tasks

that is, the computer tries to learn to interpret and produce written and spoken content. Summly daily uses natural language processing (NLP) to summarize about 200,000 news articles, while also providing users with personalized content streams. Siri and Google Now combine NLP with speech recognition technology, and users can ask questions verbally and receive verbal answers. The technology being developed, such as Viv, is further: what they have to do is give the computer a conversational memory and allow for questioning and clarification. In other words, such a dialogue is appropriate.

The interactive paradigm that this science fiction has been imagining is becoming a commercial reality. This innovation is fundamental because it allows us to interact with machines that do not have a natural visual UI, such as automobiles, wearables, robots, or even your house.

A closer (1-5-year) opportunity is the possibility of automating all types of communication tasks currently performed by humans. How can you identify which task is being automated and what conditions are ripe? And where is the value created? To answer these questions, you have to look at the interactions that follow the patterns and then look for the time-consuming tasks. After analysis, the following conditions are met:

Medical Diagnosis: interactions with doctors generally include a series of symptom descriptions, and then infer the most probable disease. This is an ideal case for the intelligent computer of mastering language skills. It is likely that IBM Watson will soon become the best doctor in the world. It has come into contact with the latest medical knowledge, accurate and consistent, and can theoretically serve everyone for 24 hours.

Meeting schedule: Traditionally, this has been the role of secretaries and administrative assistants. My personal assistant is called Amy Ingram. "She" is the fruit of x.ai, and few people realize that talking to them is not human. Similar services can be obtained from claralabs and so on.

language learning: want to learn Chinese? If you have a conversational partner who speaks fluent Mandarin, it would be better if he would never mind correcting your mistakes and never getting bored. How about an all-weather real-time translator who knows all the languages?

news report: narrative science is training computers to write news stories. Of course, it is still not enough to write an opinion like the Atlantic weekly, but the robot writes the results of the event and the financial reports are already at its fingertips.

Staff Recruitment: Recruiters are largely an intermediary between candidates and employers who are based on a defined set of features (location, industry, skills, experience). Companies looking to use machine intelligence and NLP to automatically identify and recommend candidates appear early. Of course, the first interview must also have a computer host.

Online travel agency: flightcentre.com does the "real-time assistant" above help you find a hotel in Prague a man or a robot? If it was in 2017, I already have a definite answer.

Legal tasks: The forensic process requires a lot of junior lawyers to comb through the mountains of documents to find clues. Organizations such as Equivio have combined machine intelligence with NLP to automate this function, as a result of the fact that the deputy, who drank 8 cups of coffee, had already been worked on by the computer until 2 o'clock in the morning, and that the result was not only more precise, but also that it did not have to drink coffee or sleep.

Call Center: The call center guys are trained, and their answers are scripted and respond differently to different situations. But not everyone can remember, or some do not follow. You should have experienced these situations more or less. It would be nice to have intelligent computer use NLP and speech recognition technology instead.    If done well, the size of the market is huge.

Visual Tasks

in the February 2015, computer Vision made a breakthrough: machine eyesight was higher than man's. To be more precise, it means that computers are more accurate than people to look up a series of images and classify them. With this level of computer vision (the ability to see) combined with machine intelligence (thinking ability), there is a fascinating possibility to emerge.

A lot of work requires people to look at the images and then evaluate the content. For example, airport security is a clear example, security to scan the box and then see if there are firearms and other contraband, such as radiologists, after more than 5 years of training to accurately through the X-ray diagnosis of disease.

I believe that it will not take long for computers to perform such tasks to become reality, and the results are faster, more accurate and less costly. The following possibilities will subvert the industry:

Car: self-driving cars are the most advanced case study combining computer vision and machine learning.

Agriculture: instead of paying someone to drive around to see the crop, why not allow computers to monitor crops based on satellite imagery?

Military: cockpit monitors and drones detect objects and use machine intelligence to automatically identify the other person as a threat (I bet the CIA's VCs have made a pile of undeclared investments in the field ...). )

basic Facilities: Google's Street View is a street image for most parts of the world.   The city department would certainly like the idea of using algorithms to automatically identify roads and buildings that need to be maintained, rather than letting inspectors drive around.

Medical: Check for skin cancer Why do you have to see a doctor? Send a picture of freckles in the past, let the computer analyze it.   Similarly, diagnostics such as NMR and X-rays should be possible.

Building maintenance: Security checks need to check oil pipelines, refineries, etc. where there is no oil spill. Is it possible to have a camera drone patrol and then let the computer analyze the footage in real time?

There are tens of thousands of vertical fields that can be automated, many of which are quite obscure and difficult to master. KeyMe seeks to automate a $6 billion locksmith market through photos. Tractable is helping the plumber to verify that the pipe is welded.

so please pay attention to the traditional industry, the internet is not + good, the era of machine intelligence + is coming soon.

article from Devstore News, Welcome to read more Internet News (www.devstore.cn)

The Internet is not + good machine intelligence + the era is coming soon