Artificial intelligence from DeepMind better than a weatherman

Artificial intelligence from DeepMind better than the weatherman?

A British startup is working with a team of meteorologists to create an algorithm that predicts the weather. The motivation here is the desire for more accurate predictions than we have provided by currently existing systems. Work on weather AI is already yielding early results. The results promise well for the future!

To cite a silhouette of the company DeepMind, Wikipedia indicates that Since 2014, it has been owned by Google. What’s more, the initiative already has its first major achievements behind it. The DeepMind team is credited with creating a fully automated neural network. Its property of being able to learn quickly and being able to remember, which helps it play in a similar way to humans. Artificial intelligence from this company can also function on the principle of short-term memory, leased by us, the people.
In this way, it learns similarly to the human brain, but only in the short-term memory phase. In the brain, hippocampus is responsible for it, in AI algorithms.

Is artificial intelligence smarter than humans?

The issue, while complex, seems clear – no! However, considering its capabilities and the lack of limitations of the human being, in a way it can be considered so. In fact, like everything in life, it has its good and bad sides, but just like there are no people worse and better, so there are no computers smarter than artificial intelligence, nor people who can act so efficiently
in the area of computing power as it. These two aspects – humans and technology – are complementary, that is, neither of them prevails, but complement each other.

Artificial neural networks – DeepMind’s sleight of hand

As indicated by the portal specializing in issues related to AI the term neural networks speaks for itself. ( These are networks of electrical impulses similar to those emitted by our brain neurons. Such a system of artificial neurons is able to perceive stimuli and interpret them in an appropriate way. Like humans, it collects information, turning it into data. On the basis of such collections AI is able to perform pre-programmed operations, which is similar to decision-making, as a result of their analysis.

Deep machine learning technology

Article from MIT Technology Review, which we are using here also raises the issue of deep machine learning. Back to information from the Artificial Intelligence deep learning is another way of saying test. A test within which a machine slowly clarifies the data picture based on its verification. It verifies on information already acquired, which allows it to classify individual data segments as specific knowledge. It is aided in this process by prioritization, determining the rank and usefulness of the acquired sets. “Deep learning” is used in Google’s algorithms, as well as on Facebook’s news feed for predicting earthquakes or heart disease.

Weatherman from DeepMind

An English company decided to go one step further. The AI team decided to set it problems in the sciences. It has partnered with the Met Office, the UK’s national weather service, to do so. The fruit of this collaboration is the tool Deep Generative Model of Rain. It is used to determine the probability of rainfall, within the next hour and a half. As TR points out, this is the most difficult thing to do when predicting weather, which greatly reduces the effectiveness of the forecast.

DGMR beats opponents!

An incomparable performance against other meteorological tools has been demonstrated. DGMR has proven to be best at determining the location of precipitation, its extent, direction of movement, as well as the intensity and strength with which it will strike a given territory. The result is astonishing, because in blind tests the solution from DeepMind turned out to be better in as much as 89 percent. instances of tools being tested.

Application of a new discovery

In the text we have an apparent praise for current tools that are able to forecast weather in a long-term way. However, there is actually a problem with predicting sudden events, especially rains. A new initiative from DeepMind could help the events industry, aviation, or even medical and other emergency services (water, mountain).

About the project

A team of scientists has put its AI through training. This was done on data available from radar measurements provided by public weather stations. Many countries emit the so-called “rain” every day. snapshots, or daytime weather measurements that track cloud formation and movement. In England, it’s every five minutes. Measuring so frequently creates a time-lapse video of cloud layers, from which the AI determines patterns in how they form and move. The researchers fed the data into a deep generative network, a type of artificial intelligence that is trained to generate new data samples that are very similar to the actual data it was trained on.

How “meteo-intelligence” works?

It works in such a way that after generating false radar images, it follows the actual measurements, overlaying them and creating possible sequences of atmospheric events. Researchers compare it to watching a few frames of a movie and guessing what will happen based on them.

56 Met Office weather forecasters not involved in the project were invited to test. Their lack of involvement was meant to demonstrate the authenticity of their assessment, as this one did not involve personal input into the research effort. The DGMR tool was blindly benchmarked against others and won the comparison uncontested, losing by only 11 percent. cases.

Artificial intelligence can be useful!

DeepMind boasts of new invention, reminding us of practical applications for artificial intelligence, which is not the case in too many branches of its operation, which the company emphatically regrets.

The editors of Mobile Trends are also interested in the further development of such solutions, while they encourage you to read other articles
about artificial intelligence.