Deep Learning and Advanced Natural Language Processing is the latest development of Artificial Intelligence (AI) -In conversation with Mr Tom Sonoda, CEO of UTAGOE Japan

Speaking to us about Artificial Intelligence, Mr. Sonoda gave an insights on how the concept of Artificial Intelligence started and its advancement in today’s life. He also talked about his company and the recent projects they are working on.

He gave us glimpse of his latest research materials on deep learning during his talk and told that deep learning is a main concept in Artificial Intelligence.

He also gave us lecture and demonstration on how image-recognition process works with AI.

When the engineers feed the network an image, the first layer process info. This layer then ‘talks’ to the next layer, which then go at processing the image. When process is repeated over number of times, with each layer identifying key features and isolating them until it has figured out what the image is.

He explained how machine visually identify human voice parts from acoustic signals, how machine detect different objects in combination of factors at pixel level and be able to match with similar objects. The applications of such technology is limitless, for example, machine can auto detect simple objects in the advertisement and remove unwanted objects from its image. Also  machine can produce semi-realistic imagery from simple road map by guessing landscapes from the simple map and auto generate texture mappings. But most amazing part is how AI started to dream learn!!  it means that AI processing of imagery is getting ever closer to human neuron networks they just keep learning skills without being exhausted.

Another remarkable advancement he is excited about is Natural Language Processing (NLP) like ELMo, BERT and GPT-2 that has been developed recently in the year 2018 and 2019, he mentioned GPT-2 as the the most powerful technology and strongest concept of NLP which has been announced just two months ago. He said, “Using GPT-2 model, a natural conversation is very much possible, fake news can easily be generated and it’s harder to detect who is having the conversation with who.”

While asked to Mr. Sonoda about how he see Human world vs AI world after few years, then he smiled and said, “Computer would obey human commands for couple of years; but after that computer will interact with computer and learn from each others.”

Waseda University Incubation Centre

I asked additional questions to Mr. Sonoda.   

What area of AI development, your company focus on?

Mr. Sonoda: Deep Learning, NLP, ML and Vision.

How many students in Waseda University are working on AI?

Mr. Sonoda: There are a total of 800 students working on AI in the university and around 15 students work in the Waseda University Incubation Centre.

What are some of the challenges you interface while working on AI in Japan?

Mr. Sonoda: Its mainly human resource problem and not having library of their own.

Is there any particular community for AI in Japan he belongs to?

Mr. Sonoda: No idea of any such AI community in Japan but he follows Open AI community, an NPO started by Mr. Elon Musk.

How do you like collaborating with overseas AI developers  for example in India, China, US, Canada and other countries?

Mr. Sonoda: They are welcome. Would love to work with qualified engineers.

Which AI company do you follow and respect the most and reason behind it?

Mr. Sonoda: Google.ai and the reason he gave was the nice environment to work in.

I gained a lot of knowledge and got to know some amazing development being done in the field of AI after talking to Mr. Sonoda. I even enjoyed playing the deep learning reverse game that they have developed.

“We need to constantly be open to new ideas and approaches, such as Artificial intelligence (AI) and, be willing to challenge assumptions.”

This article is prepared by Mr. Awesh Kumar, an Associate Innovation Researcher at Seekers Base, Japan.