Last time we have discussed the reasons that Deep Learning is receiving a lot of attention. Today we will take a look at the current state of AI development in global companies and how Korean companies are dealing with the changes in the industry.
- Not Like a Human, but Like a Human. Artificial Intelligence Computing (Part1):
The Global AI Craze
These days most global IT companies are involved in AI development. This makes sense considering the endless possibilities for creating value in the AI industry. Especially since the advent of the Internet of Things (IoT), there is a need for objects to be able to detect their surroundings. Deep Learning, of course, is receiving attention as the core technology in this field. According to the consulting firm, Mckinsey, AI is expected to spark a USD 6 trillion ‘intelligent labor automation’ industry by the year 2015.
Accordingly, companies are taking actions for mergers and acquisitions and to acquire talent in order to be competitive in the market. The company with the biggest foothold in the business is Google. Google is sucking up as many AI companies and specialist as it can. Google is in the process of a merger under the name ‘AI Manhattan Project’ worth hundreds of millions of dollars and in June 10s of companies joined the project.
Google acquired the smart temperature control firm, Nest, for USD 3.2 billion in January. This was seen as a move to managing data centers through Deep Learning systems. Google also recently acquired the Deep Learning venture firm, DeepMind, for USD 650 million. Google has also been purchasing robot technology related companies and it appears that they have plans for something more than just an AI robot fusion project.
Other leading companies besides Google are also performing active research in AI. Microsoft implemented a voice recognition function in the personal assistant product ‘Cortana’ that they released this year. This voice recognition function is said to be one step beyond Apple’s Siri, which set the standard. Also, Microsoft announced a Deep Learning technology that allows a computer to distinguish the breeds of dogs at the MS Research Science Council in July. This project, called Adam, analyzes approximately 14 million pictures of dogs. Adam boasts analysis speeds that are approximately 50 times faster than Google DNN technology.
Facebook announced the release of a facial recognition program called DeepFace that was developed through an AI research group in August of last year. DeepFace is a service that recognizes the face of users no matter angle the user is viewed from. DeepFace is regarded to be second only to the eyes in accuracy. IBM has developed a genuine AI ecosystem using their Watson super computer/ The company known as the ‘Chinese Google’, Baidu, has established a Deep Learning research center in Silicon Valley and recruited Andrew Ng who is regarded as 1 of the top 3 people in the AI field.
The Korean AI Market Begins!
Machine learning and AI technology are at the core of the voice recognition, image recognition and related fields. As much as people are showing interest in these technologies, leading global IT companies are also scrambling to get into the AI market. But how are Korean companies responding to the great efforts being made in AI around the world.
Naver, one of the top Korean portal sites is playing a pivotal role in Deep Learning research in Korea. Naver provides voice recognition services and are in the testing stages of implementing Deep Learning for news summary and image analysis applications. Naver has shown a 25% voice recognition error rate with their Deep Learning algorithm. Naver is also developing and implementing an algorithm to summarize new article with and without headlines.
Aside from Naver, ETRI (Electronics and Telecommunications Research Institute) have developed a terminal embedded automatic interpretation service called GenieTalk that is capable of interpreting 4 languages. When the user speaks Korean into a microphone, it is translated simultaneously into Chinese and can even emit spoken language. GenieTalk was evaluated to be 15% more accurate than Google Translator.
In June, the top Korean mobile messenger service Kakoatalk’s chairman, Bumsuk Kim, invested in an image recognition technology development startup called Cldi. Several Korean companies are making attempts at Deep Learning algorithm startups but at this point they are all still struggling. They are struggling because it is not get investments in technology and the Korean software infrastructure is relatively weak. A lot more time and effort is required to promote growth in a new company in order to overcome these obstacles.
We have now taken a look at Deep Learning. It is clear that AI cannot be left out of this next generation industry. A diverse set of elemental technologies must be fused together in Korea as well to create a high value service. Ray Kurzweil, the godfather of AI, refers to the era after the year 2029 as ‘singularity’. Singularity refers to a time when AI capabilities will bypass those of the human brain. I wonder if Deep Learning will be responsible for the construction of a new culture for mankind.
Written by Jongho Choi, University reporter for LG CNS