Recently Browsing 0 members
No registered users viewing this page.
By Ather Fawaz
Google AI model outperforms traditional methods of weather prediction
by Ather Fawaz
A couple of weeks back, Google AI used a machine learning model to improve the screening of breast cancer. Now, the firm has used a convolutional neural network (CNN) in nowcasting precipitation.
In the paper titled "Machine Learning for Precipitation Nowcasting from Radar Images", researchers at Google AI have employed a CNN to give a short-term prediction for precipitation. And the results seem promising, and according to Google, outperform traditional methods:
Unlike traditional methods, which incorporate a priori knowledge of how the atmosphere works, the researchers used what they are calling a 'physics-free' approach that interprets the problem of weather prediction as solely an image-to-image translation problem. As such, the trained CNN by the team—a U-Net—only approximates atmospheric physics from the training examples provided to it.
For training the U-Net, multispectral satellite images were used. Data collected over the continental US from the year 2017 to 2019 was used for the initial training. Specifically, the data was split into chunks of four weeks where the last week was used as the evaluation dataset while the rest of the weeks were used for the training dataset.
In comparison to traditional, venerable nowcasting methods, which include High Resolution Rapid Refresh (HRRR) numerical forecast, an optical flow (OF) algorithm, and the persistence model, Google AI's model outperformed all three. Using precision and recall graphs, the quality of nowcasting was shown to be better on the U-Net model.
Precision and recall (PR) curves comparing our results (solid blue line) with: optical flow (OF), the persistence model, and the HRRR 1-hour prediction. Left: Predictions for light rain. Right: Predictions for moderate rain. Moreover, the model provides instantaneous predictions. This is an added advantage because the traditional methods like HRRR harbor a computational latency of 1-3 hours. This allows the machine learning model to work on fresh data. Having said that, the numerical model used in HRRR has not entirely been superseded by it.
Google envisions that it might be fruitful to combine the two methods, HRRR and the machine learning model for having accurate and quick short-term as well as long-term forecasts. According to the firm, they are also looking at applying ML directly to 3D observations in the future.
If you are interested in finding out more, you may refer to the paper published on arXiv here.
Experts baffled by mysterious creature discovered in Kiwi's kitchen
What do you guys think?
Google Earth gains the ability to measure distance and surface area
by Vishal Laul
Google Earth is (re)gaining the ability to measure distance and surface area with a new tool, the company announced today with a blog post.
The tool allows users to measure the distance between two points on the globe or the surface area of a region. It's adequately clever as well, allowing for the ability to move the borders of a selection and change it into whatever irregular shape one would want to measure.
Google says that this was a highly requested feature, but perhaps that’s because the desktop version has offered a similar tool for more than a decade. Google turned Earth Pro into the standard version for desktop in 2015, and began offering it for free. That's still available if you are looking for something more sophisticated and feature-rich.
Still, it’s good to have this feature available on the more common versions; the tool is available on the web – which only works in Chrome, for now –, and Android starting today, with an update for iOS “coming soon.”
Researchers believe aliens could send malware and destroy humanity
by Christopher White
There's no doubt that we have a lot of computer security issues on our planet right now. Ransomware seems to be a daily issue, with new variants constantly being released; hardware issues in CPUs give bad guys the ability to steal your data; and lax security in the "Internet of Things" (IoT) enables bad guys to run denial of service attacks against the biggest companies in the world. However these may be simple annoyances compared to what extraterrestrials could do to our planet's IT infrastructure.
In a paper written by Michael Hippke and John G. Learned, the researchers explain various ways an alien civilization could destroy the world, either intentionally or unintentionally, by embedding code in a message. They speculate that even simple markup languages like TeX and LaTeX could be used maliciously, and highlight the difficulty in decoding the languages manually. In addition, the paper details that an alien AI could begin a negotiation with humanity, in essence social engineering an attack.
One recommended solution is to build a "prison" on the moon, a computer that is used to decode alien messages, but is isolated from other networks and which could be remotely destroyed if necessary. However they go on to say that, "[c]urrent research indicates that even well designed boxes are useless, and a sufficiently intelligent AI will be able to persuade or trick its human keepers into releasing it." While there are no silver bullets to this problem, and the researchers note that the overall risk to humanity is low, it's a topic that can be fun to think about.
This topic is hardly new as there have been many books and movies that explore the concept of malicious invaders. For example, in the movie Species, the SETI project received a transmission with details on how to splice alien DNA with human DNA and the result was mayhem. What other interesting books, movies, and TV shows have you seen that address this topic?
Source: Cornell University via Schneier.com| Image courtesy of Evolving Science