Computers and so-called artificial intelligence can become increasingly powerful. But they certainly will not replace humans. This is what the head of technology at Amazon said at the DLD innovation conference.
Werner Vogels, Amazon’s chief technology officer, assumes that learning computers will need human help for a long time to come. This applies to both speech recognition and machine learning as a whole, Vogels told the German press service on the sidelines of the DLD innovation conference in Munich. So there is a lot of knowledge about different processes in industrial production that only humans have. “You won’t be able to live without people for a long time.”
To learn computer systems, people would first have to describe 20 or 30 percent of the data for them – and then the machines analyzed the rest based on this. “This relationship can change over time. The goal is, of course, as much of this as possible. Automate process.” People should retain many skills: “If the car battery breaks down and the robot car software turns off, you should be able to do it yourself to drive.”
Some tasks can only be performed by machines
At Amazon, some businesses as an online store would also have a dimension where they could no longer be managed by people alone. An example is to check with every order whether there is a fraudster behind it. The machine therefore assesses the risk based on various factors and, with high probability, engages the person for testing.
Vogels defended Amazon’s facial recognition software Rekognition, which the group makes available to the US authorities, among others, against criticism that it works mainly for people with white skin. Bad results were not due to the learning model, but to the choice of data to train it, he said. In fact, Amazon is working with regulators in several countries to achieve responsible use of the technology.
Quantum computers still need a lot of development time
Researcher Joy Boulamwini spoke at the DLD conference, the study of which showed that facial recognition systems work poorly in black women. Unlike some other providers, Amazon tried to discredit the results of their research, she said.
Vogels sees no technology in quantum computers for the foreseeable future, which is currently receiving a lot of attention. “The promising tools are not there yet. It will take a while before we run complex applications on quantum computers.” But since software development would then be radically different, you should start thinking about what and how you want to do it.
Amazon therefore has a quantum program with selected customers. The group’s goal is to provide everyone with access to quantum computers, as is the case with current cloud services. A drastic performance improvement is expected from quantum computers, but they are only at an early stage of development with many unresolved issues.
Vogels is considered the architect of Amazon’s AWS cloud service, which provides many startups and established companies with computing power and storage.