top of page

Building AI that builds AI

Adatos is a venture-backed technology company that leverages the power of the GPU to build and deploy data intelligence solutions for the private sector. The company has operations in Singapore, Japan, Hong Kong, Malaysia, Indonesia, and the Philippines.

In an interview during the NVIDIA AI Conference in Singapore, Drew Perez, Managing Director of Adatos, shared about his company and solutions

What is Adatos about? What challenge are you addressing? DP: We focus on addressing the issue of the lack of data scientists around the world. The number one challenge for enterprises trying to get solutions to work today is that they cannot find the talent. Other than depending on data scientists to build solutions, we build AI that builds AI.

Today, what we do for finance, agriculture and medical imaging is that we build these solutions. Actually, it is machines that build the solutions, not us.

How did you develop the AI solutions? DP: To do AI, you need two things. First, you need the data to train the machines. Second is that you need computing power.

It is only through the IBM-NVIDIA joint project in 2014 that we have NVIDIA DGX-1 and IBM Minsky, which are the only production servers available in commercial use that allow us to not only do this type of AI building AI. Currently, there is no commercially available alternative to do this.

How is the market responding to AI? DP: Two years ago, the question was “can do you prove to me that AI works?”. Last year, it was more on “help us navigate the waters so that we can implement AI”. In 2018, it’s probably going to be “tell me which vendors are doing this”.

What are your thoughts on the role of the GPU in driving AI? DP: As far as GPUs are concerned, that type of computing is only available through the IBM-NVIDIA relationship. We need that computing power. It is only through the parallel processing power of general purpose GPUs that this is available.

NVIDIA provides a computing capability with the parallelisation. Unique to that is the architecture, which started off with Mellanox, and is now called NV-Link, that allows the movement of information from the GPU and to the CPU and back. This is essential to create the system of systems.

Once we have that and we have data for very indicative use cases, such as credit risk, medical imaging or special imaging for yield optimisation of crops and plants, then we can have cases that leverage GPUs so that machines have the capabilities to build machines and self-code and self-evolve.

What do you see happening in the days ahead? DP: Currently, the state of the industry is relatively unknown. Today, Google and a number of us have built flagship products that lead the industry. The challenge is that if you are a data scientist, you are pretty obsolete today. There is no need for human being to build AI or even write code because we have machines that write their own code. There is a general sense in the AI community to completely demoncratise this where a non-technical person has the productive capability of a PhD scientist. There is no reason why anybody who can play on a smartphone cannot build these solutions that we have.

Today, we have autonomous AI. There is no reason why we need humans to intervene. In a few months, anybody who is in the data science field and trying to build AI should better watch out because the machines can build AI better than you can.

What then would become the role of humans in a world filled with AI? DP: One of our biggest challenge working with the World Economic Forum is that they call us the Fourth Industrial Revolution. The first, second and third talk about increasing productivity. We have to look as humans on where that symbiotic relationship is. We can leverage machines for what they are very good at, which is repetitive work, high accuracy and high speed, and leveraging us to focus on what we do best as humans, which are the creative skills and imagination.

Today, not only industry but our education system is still trapped in a mindset of training us to do what machines can do better than us. We need to shift that and equip ourselves for a future where machines are our helpers, doing the things that we do not want to do so that gives us more time to enjoy life.

bottom of page