Why can't startups develop self-driving cars?

As of March 17, startups have always been innovative, advanced and free from inherent thinking, but in the future, a key technology-self-driving cars, startups are not doing much. On the contrary, technological advances in this area rely largely on technology giants such as Google. So why is it hard for startups to build on the development of self-driving cars? Science and Technology website TechCrunch recently wrote the analysis of this issue:

Over the next 10-20 years, large-scale technological innovations will revolve around machine intelligence, robots and sensors. These areas require strong financial support and a lot of patience, and the combination of these two elements is incompetence even by the most progressive venture capital companies.

As Google has shown us by self-driving cars, the combination of machine intelligence, robots and sensors in complex situations-like driving cars-is already better than humans, and for most people it was unthinkable 10 years ago.

There is no doubt that Tesla, after many trials and tribulations, has made an electric car that attracts worldwide attention. But GM has already launched a mass-production electric car in a year, and Tesla has taken advantage of the general dilemma that traditional carmakers are largely unable to embrace all-electric cars and integrate into modern electronic components.

The Tesla roadmap includes "automatic Navigation (autopilot)" and the ultimate "autopilot (autonomous)" feature. They may indeed be able to launch these features slightly ahead of traditional carmakers, such as Mercedes and Lexus, which are also actively developing similar technologies, but the winner of the competition will be Google, which has been technologically advanced for several years and has the ability to pay huge licensing fees.

The pure computational performance necessary for cognitive computing is huge, and the integration of next-generation sensors (such as LiDAR) is extremely complex, and the regulatory environment for introducing smart machines is unpredictable. The experiments and unpredictable schedules needed to develop this technology are clearly part of research rather than development, and VCs have always hated research-they like to develop.

Even the DeepMind, the biggest VC-backed start-up in the field of machine intelligence (with 50 million of the investment and 75 top researchers), has recently been bought by Google at $500 million. Google's recent acquisitions have been so frequent that they seem eager to make a brick for the future through acquisitions. In addition to DeepMind, they have recently acquired AI research company Dnnresearch and recruited key figures such as Ray Kurzweil, who many consider to be the godfather of business cognitive computing. Google is also one of D-wave's clients, a highly criticized quantum computing company.

Companies such as Google, IBM and Microsoft have built machine learning teams that can take advantage of the company's computer networks built around the world. According to Kurzweil, up to 100 billion transistors are needed to match the number of nerve cells in the human brain, and a single chip cannot have so many numbers by 2025. But as the transistors in the microchip are getting smaller, the process may be accelerated.

It will take a while for the market to understand machine intelligence. IBM has proved that computers can win over humans in chess and jeopardy, and they are also transitioning their cognitive computing to medical and other fields. Even for companies of this size, shareholders are complaining about the cost of the transition, and the director of Watson's supercomputer program has recently changed. If IBM's Watson is a start-up, its shareholders would have forced them to sell them in order to put their money into more efficient short-and medium-term investments.

Other companies with large computer networks are also joining the game. Facebook recently set up an artificial intelligence laboratory and hired Yann LeCun from New York University, while also acquiring speech recognition company Mobile Technologies.ebay to recruit Hassan Sawaf from the international Company for Scientific Applications (SAIC) As a vanguard in its research on machine intelligence. Yahoo and Carnegie Mellon University have reached a deal to visit researchers in the latter. Apple, which has always focused on computer clients, is also trying to keep its messaging system scalable and not lag behind in Google's relentless research on machine intelligence.

The benefits of this trend are immense. Every aspect of the economy will soon have its own version of self-driving cars, even the frontiers of medicine. Anyone with undiagnosed or confirmed illness can tell you that the guesswork in the diagnosis and the excessive professionalism of medical experts is maddening. A computer can analyze the disease as a whole, quickly narrowing the range of possibilities and repeating the elimination test. Given the need for fundamental change, it may take a long time for the medical field to transition to such a situation.

In addition to repeated fears, startups are sure to take on more challenges than pictures and chat apps. But can they compete with industry giants in areas such as machine intelligence? If Google, IBM and Amazon bring in a so-called "cognition-as-a-service", this may usher in a new wave of startups, It's like Amazon's infrastructure, Service (Infrastructure-as-a-service) 's contribution to the Web.

IBM's Watson based company prepared a 100 million dollar investment and already made its first investment. By acquiring new cognitive services, such as Wise.io, Expect Labs and BIGML, Amazon can start a cognitive service and make it a sizable one. These can bring a new generation of "smart" startups. (Eskimo)

(Responsible editor: Lu Guang)