Google’s Acquisition of DeepMind Hints to More Intelligent Web Searches
It seems that Google is venturing into sci-fi territory with its latest acquisition. Or is it? Google has shelled out $500 million to acquire the UK startup DeepMind, according to TechCrunch. DeepMind focuses on artificial intelligence and machine learning.
To date, Google’s true intentions regarding the acquisition of DeepMind have not been released. Google X Labs produces some innovative and helpful products (i.e. Google Glass and self-driving cars), so there is no shortage of opportunity to utilize artificial intelligence there. Google’s primary mission, though, is “to organize the world’s information and make it universally accessible and useful.” So it may very well implement new applications in artificial intelligence in more practical instances such as its Web search algorithm instead of trying to create self-aware machines.
Improving Search Results
Currently, Google’s internet search results rely on information and content that is properly labeled online. This is why SEO still involves a lot of care toward meta information, word content and descriptive anchor text for links. The exception to this is Google’s image search, which uses a mathematical algorithm to analyze image features to find the same or similar images that have been posted elsewhere online. But these internet search algorithms are woefully inadequate when searching for videos, audio files and interactive content like video games. Yes, these file types can be located when they are labeled correctly in their titles or descriptions, but otherwise these media types are completely missed by search engines.
Here is where artificial intelligence would be useful. By 2012 Google had already developed a neural network that made use of machine learning to recognize the content within YouTube videos. In this case, a “deep learning” model of artificial intelligence was used, where the machine’s conclusion was not based on pre-loaded knowledge and detailed sets of conditional statements but rather on a framework where the machine was able to create its own concepts. The result of this experiment: the computer was able to create the concept of, and identify, a cat, according to the Verge.
Artificial Intelligence: Separating Fact from Fiction
The fast-paced development of technology in the 21st century as well as our fascination with the possibilities presented to us through the world of science fiction has built up our belief that science makes just about anything possible. While we have in fact made extraordinary strides in technology thus far, we are still a long way from creating a machine with the intelligence to match anywhere near that of a human being.
The concepts of common sense and even understanding are still elusive in the field of artificial intelligence. Will Google Search ever be developed to the point where it can determine if Web content is humorous or not? Whether Web content is offensive? Google already dealt with this issue back in 2009 when, according to ABC, offensive images appeared as results for the search term “Michelle Obama.” The incident prompted Google to issue the following statement:
The beliefs and preferences of those who work at Google, as well as the opinions of the general public, do not determine or impact our search results.
So while Google does not ultimately feel responsible for the search results it provides, that doesn’t necessarily mean Google doesn’t want to offer users the refined search results they desire. Currently Google Safe Search for Web images censors search results by examining the content of the page hosting an image. It also analyzes images for being potentially explicit, although Chris Crum from Web Pro News found that many webmasters are finding their innocuous content blocked from search results because of such efforts. So it seems that an internet search engine that can reliably vet Web content is still an elusive goal.
Deep learning in artificial intelligence forms the foundation for automated machine learning. This can be applied to accomplish relatively simple things such as creating smart home appliances that can program themselves based on the unique activity in a home, teaching a robot how to climb stairs or, more importantly, understanding the purpose of stairs. Deep learning’s goal is to have the machine learn how to climb stairs and why they are used, instead of being programmed step-by-step on how to climb them. In this respect, the idea of machine being intelligent enough to be on par even with a seven-year-old human is still a good way into the future, although some experts say creating such a machine will be possible within the next 100 years.
What Makes DeepMind Unique
DeepMind’s moniker comes from an area of machine learning called deep learning. Deep learning tries to mimic the natural neural network in the brain by processing data by means of context, memory and positive reinforcement. Some of DeepMind’s coolest work has been training software to play video games where no information or rules about the video game were loaded onto the program before it started playing. The software learned how to play the video solely upon the positive/negative reinforcement of the game performance or score.
At one point, Facebook was also interested in acquiring DeepMind, according to Re/Code. Facebook’s interest may have lied in deep learning’s potential to make targeted online advertising truly dynamic. It turns out, however, that the world of deep learning experts is pretty small, with a good number of them still in, or fresh out of graduate school. So to say that the professionals at DeepMind and their work are in demand would be being quite modest.
In the end it seems that Google and other companies are looking towards the latest trends in artificial intelligence and determining how to use them in practical applications. While artificial intelligence will always be the mainstay of science fiction fantasies where it is a key feature of the autonomous computers and robots of the future, it’s fascinating to see how artificial intelligence is being used today to enhance our own personal daily technology uses and experiences.