Latest News

  • Daniel Chan

Conversational AI barriers today

I am running a technology firm with years of experience in providing chatbot and natural language understanding applications for enterprises, over the years most of our clients are facing issues with quantity and quality of data which is almost a pre-requisite for any customisable artificial intelligence development.

Under DeepPROject acceleration program from DEEPTECK, we have formed a small team of PhDs in machine learning and natural language understanding experts who are interested to solve this real world problem with their deep domain and research knowledge.

We will start by addressing and narrowing the problem and reveal step by step logic, progress and approach of our proposed solution publicly in MEDIUM and DEEPTECK THINKTANK slack group. By doing this, we hope to engage and attract the best talent around the globe who is interested in solving this problem to join our team and gather the best thoughts from top notch domain experts to refine our thinking and execution.

About DeepPROject

DEEPTECK has started a new fast track acceleration program called DeepPROject for scientists to transform technology idea to real venture with different platforms to guide scientists on team formation and commercialisation with supports from a team more than 10+ professionals including startup founders, entrepreneurs, sales, marketing, product development, ICO and fund-raising specialist.

Problem statement

Conversational AI tools such as chatbots, sentiment analysis and auto information extraction presents huge opportunity for businesses, enterprises, and organisations to improve their efficiency in external and internal customer engagement. Despite the numerous benefits offered by AI conversational tools, its mainstream adoption is not yet prominent due to its inaccuracy and high costs in using it with the following limitations:

  1. Conversational AI requires massive amount of data (Gigabytes) for training the AI model. Today most enterprises do not have sufficient data for this.

  2. Even data is available, it is unstructured requires huge amount of time for cleansing

  3. Most data available online today is not domain specific such as hospitality, retail, travel specific. Making AI engine not accurate on domain specific topics.

  4. Privacy issue from EU General Data Protection Regulation will likely affect the collection of conversation data for any AI training activities even from very large enterprises thus further barrier for data collection on training AI engine.

Our Team

Antony Gu

  • Ph.D in NLP and deep learning

  • Expert in dialogue generation (chatbot) with deep learning

  • Have 10 years of programming experience, 4 publications in top conferences

Gautier Marti

  • PhD in Machine Learning

  • Expert in Quantitative Finance and NLP/NLU

  • NLP Quant Researcher at AXA

Gary Ng

  • CEO of Customindz |Artificial Intelligence Solution Provider

  • Expert in Startup, Fundraising, AI, Blockchain & Tokenisation

  • Founder of DEEPTECK & Cofounder of HKSSG — Hong Kong Startup Support Group

16 views0 comments

We Need Your Support Today!