top of page

Tianhao Wu, CTO and Co-founder AYR


Dr. Wu possesses more than 20 years of Machine Learning experience. With a Ph.D. in the NLP field, he developed search engine systems at Ask.com. Dr. Wu architected machine learning platforms and led Data Scientist and Engineer teams to deploy solutions for customers. Holding multiple patents, he has successfully worked out over two hundred real-world AI use cases. He is now CTO and co-founder of AI company AYR.


What's your industry?

Artificial Intelligence, Machine Learning, and Intelligent Automation


How did you end up sitting where you are today?

Since I was a Ph.D. student, I have worked on natural language processing. And after my graduation, I worked in the industry on text mining/understanding for search engines (ask.com) and RPA (Workfusion). I also deployed hundreds of real use cases for customers such as Citi Group, PNC Bank, Bank of America, Humana, Intel, Walmart, etc. During my time working on all these real-world applications, one big issue that I noticed is that AI is tough to apply in a real-world setting. The reason is that real-world data is dynamic and changes very often. The existing AI approach heavily relies on the pre-existing training data and cannot evolve with real-world data. In short, the current AI solutions are a one-time pre-built model; it is not "real" intelligence that can learn from real-world changes.


After considering how to solve this issue, we started Singularity Systems Inc., now named AYR, after a recent rebranding with our patented technology (real-time AI) to help enterprise customers harvest their AI fruits easily. Real-time AI can learn from data changes very quickly with few data points. Working with customers like the Wonderful company, Bank of New York Mellon, Tangoe, and other third parties, AYR can achieve over 80% STP (straight-through processing). In comparison, the current solutions on the market can only achieve up to 30-40% STP. With data volume increase, AYR's model can learn more quickly and achieve over 95% STP for most use cases in a week or two. The STP means the AI model correctly recognizes everything on a document (no errors).


To make AI more accessible and easier to use, we founded AYR. Our goal is that everyone can create their own AI models and use them.


What kind of work does your daily role involve?

There are three types of work that come together to make AI available to ordinary people. One is AI modeling. The second is engineering work to make the AI model user-friendly, and the third is to have a systematic way to manipulate data. These are the three major things I work on daily.


We design/implement/train/embed all the AI models used for enterprise customers (all AI models can be deployed locally on-premise). In addition, we automated the entire modeling training pipeline into a no-code system for the customer so that the customers do not need to go through complicated AI modeling details to train/use/improve AI models. For the customer, it is just a simple click of one button, then everything about AI modeling will be automatically handled by AYR's system.

The engineering work will provide a no-code platform and 'out of box' products for customers to use AI models seamlessly with high efficiency (speed) and high accuracy (the result). We also provide the customer the UI to customize pipeline, pre-processing, post-processing, human-in-the-loop, auto report generation with any coding, and with the flexibility to meet complicated business requirements. AYR.ai can also connect to UiPath/Automation Anywhere/Boomi/AWS/Azure and other third-party platforms without code. AYR can handle over 100K pages/hour with just one GPU server. And the volume can be increased in a linear ratio by adding hardware.


AYR, Inc has also developed many data enhancement tools, such as IDS (Intelligent Data Simulator), to help customers label/clean/generate data. Also, AYR uses a massive amount of data (billions of documents) to train backbone models and has learned 90% of the knowledge. The customer only needs to fine-tune (customize) their model with their only data/use case to finish up the remaining 10% of the modeling. AYR's backbone model is like ChatGPT. A pre-trained large generic model.


What gets you excited about your industry?

The AI industry is the key to the digital age, which is coming and will change humanity's life and impact all industries. AI will dramatically increase human efficiency and change how people work and live.

However, AI could potentially bring extraordinary danger to humanity. That is why we need to pay attention to how to regulate AI to limit (or eliminate) the AI's damage.


My interest is that everyone is excited about the ability of AI after ChatGPT, but I can also see the road blockers of AI in front of us. AI can understand humans pretty well now. But do we (humans) understand AI?

This is where the dangers of AI lie because nobody (even the ChatGPT's creators do not know how the ChatGPT model works, as the logic behind these answers is not fully understood by humans).


What's the best advice anyone ever gave you on your journey in business?

Advice that I've heard the most is, "the best technology does not win the business." It is valid for a particular customer or use-case for a short period. In short, the best technology will win all businesses. But only the best technology will win the entire industry.


What's the most challenging project or situation you've overcome to date?

The biggest challenge is how to educate customers. It's like we are a car dealer trying to sell the best cars to customers, but 99% of customers need to learn how to drive a car. So we have to be a driving school and a car dealer at the same time.

Customers still need to learn about AI's ability (and limitations), so they are hesitant to use AI. However, the efforts of the AI industry and tools, such as ChatGPT, are helping people understand AI more and see that it can be easy to use.


Are you using any AI tools right now to help grow your business, or, if not, do you plan to use any this year?

Yes. At AYR, we are building a no-code AI platform, and it is an enterprise-level toolset to be able to handle millions of documents and billions of pages. With this technology, businesses can save millions of dollars and give thousands of hours back to their people.


Overall, do you see AI as a good thing for business?

YES. AI is a must-have for business in the near future. A company without AI ability cannot compete with a company with AI. Soon (2 or 3 years), a business that functions without some AI will be rare. The reason is simple: one person with AI can finish >= 100 people's work without AI.


Overall, do you see AI as a threat or an opportunity in business?

It is both an opportunity and a threat to business and humanity. For businesses, adopting AI can increase efficiency dramatically and can dominate the market. A business must adopt AI to be able to compete in the market. And this pace is (will be) very fast. Faster than people can imagine. So business decision-makers need to have urgency.


Technology makes it possible (and easy), and human society and politics are in potentially significant danger, too. For humanity, AI can be used by governments, organizations, and even a small number of individuals to control most people. From the technological point of view, in less than ten years, one man (or woman) can use AI to control all people on this planet. Imagine Adolf Hitler had an AI system that knew everyone's personal information (where you are, what you talk about, who you connect with, and even what is in your thoughts). AI Technology can make all of these possible very easily (for the one who controls the AI system, whether it is a human, a company, a government, or even another AI).


If you had one wish for the future of your industry, what would it be?

I would wish for the AI industry to benefit all (or at least most) humans, not become a destroyer or only work for a small number of "key" people.


Before we finish, where should people follow you to find out more about your work?


コメント


bottom of page