Editor's note: Machine learning has brought potential problems when IBM's Watson and Google's DeepMind outperform humans in the quiz show Jeopardy: If Fitbit can save your life, a Nike+FuelBand won't save you. Life, which product do you buy? Author Jacques Touillon analyzes the Internet of Things era, how to judge "smart" devices.
In 1996, the emergency room at CookCounty Hospital in Chicago, USA, used algorithms to determine when a patient with chest pain was at risk of a heart attack and needed a hospital bed that was scarce. The algorithm uses a flow chart-based system basic test method that is fast, efficient, and accurate: it classifies 70% of patients into low-risk categories and can detect 95% of heart attacks, while human doctors can only find 75-89 % of the incidence. More importantly, this calculation has not used any depth calculations.
Today, there are about 6.4 billion IoT devices in use, that is, almost one in the world. Even if only 1% of devices can analyze people's health by collecting data on pulse, diet or sleep, it means that the coverage of “doctors†in the world will be expanded fivefold.
But the real magic comes from machine learning. In addition to applying singular algorithms in more places, such data sizes are not available to human doctors even after decades of work experience. Imagine, for example, that Fitbit notices your pulse fluctuations, monitors your heart problems, and sends you to hospital for treatment. Machine learning means solving seemingly impossible problems with home devices.
The true value of "smart"
There is no doubt that "machine learning" makes "smart" gadgets ahead of other objects. Take Nest as an example, Nest is a typical smart device. People buy this item not because they can adjust the indoor temperature through the mobile phone, but to purchase the energy-saving function of the product, automatically adjust the indoor temperature according to people's existence and needs, and solve the problem that could not be realized in an intelligent way.
However, most manufacturers are just chasing convenience. For example, the Philip HUE lamp, although beautiful, is labeled "smart" because it can be controlled by a mobile phone. In fact, this is not a problem that needs to be solved. You won't say that this person is smart because a person will turn on the light. So why are such products also given "smart" labels?
The lack of true “smart†features in the consumer Internet of Things is also an aspect that has hindered its development. Remote access to the door lock, or the radio that automatically turns on when you go home is just a luxury, packed with fine dining or cruises – only the upper class can afford such a product.
The goal of machine learning should be to turn what you want into a must-have: a thermostat that keeps room temperature while saving you money; a wearable device that can give you personalized tips for sleep or fitness; or a source of pollution An environmental monitor that your family can detect before they cause damage.
Machine learning will permanently define who is the real winner
Products with machine learning capabilities look even more cool on the shelf. But the essence of machine learning means that among all the competing products, those products that are on the right path of machine learning are better able to maintain their own advantages for a long time.
Thanks to cloud technology, it is not a design problem to connect machine learning to the device (in fact, it is a connection problem), nor is it a hardware problem (the heavy processing can be done remotely). To some extent, this is a problem for talent, because capable engineers are rare, but this can always be solved with sufficient funds. More importantly, this is a data issue.
In order for the computer to conduct model research more reliably, the data required is massive. It needs to consider a number of factors, from user preferences to use cases, environments, and more. However, many or most of these factors are time-dependent: frequency of use, frequency of behavior, conditional frequency, changes in user behavior over time, seasonal variations with the environment, data accuracy affected by sensor lifetime, and so on.
No matter how many products, the company will not move faster. Competitors have been ahead of the market for six months, and even more users or funds can't make up the gap. Only by radically outperforming the competition, whether it's the accuracy of readings or reliable early support, as long as you stay active, you'll be a leader that competitors can't surpass.
Not just big company games
Although IBM and Google are currently developing rapidly in related fields, it seems that machine learning is too expensive for startups. But maybe it is not the case. The trick is that you can do heavy work on other people's computers. Because of cloud technology, all this will be possible. Startups can pay on an hourly basis to get some of the most complex machines. With a few lines of program code, you can even arrange many batches in succession to maintain efficiency.
More importantly, because the device itself requires only a small amount of hardware to implement machine learning, design and front-end tools still play an important role in the release of the first product.
Even though Nest wasn't so smart at first, it's just a phone-controlled thermostat that roughly predicts the time it takes to raise the room temperature with a simple algorithm. Because at the beginning it is not very familiar with its users. However, to upgrade the definition of the user's home, the company only needs to continuously send data packets. You don't have to pay more to get a better experience through machine learning (just do this step before the competitor).
Professional knowledge democratic sharing
Perhaps a startup with machine learning as a key business segment may seem a bit scary, but there are more reasons to keep us optimistic rather than scared. Machine learning adds more value than people think. It arranges doctors for each fitness area, arranges detectives for each smart lock, arranges a health inspector for each environmental monitor, and houses a housekeeper for each luxury device.
Machine learning is the way to really make smart devices stop and provide a powerful role. We have seen that early devices such as Nest and Echo have added infinite value to our lives as we continue to improve. When hundreds of technology companies join the ranks, the world will make a huge difference.
Pure Water Wipes,Natural Cleaning Wipes,Pure Water Wet Wipes,Pure Water Wipes For Babies
Shandong Tianzige International Trade Co., Ltd , https://www.tianzigedipers.com