Back in 2015, the CEO of JP Morgan mention that Silicon Valley will go after Wall Street. Fintech startups just started rising. In Asia, around April 2014, CEO of DBS Group, Piyush Gupta decides to hire Neal Cross as the first Chief Innovation Officer. Back then when I was facilitating with the developer teams at the DBS Hackaton in 2015, he once mentioned that banks can be disrupted by tech giants. [On the side note, in hindsight, it was a great decision for Piyush Gupta to hire Neal and empower him and that lead DBS to win the best digital bank in 2016 awarded by Euromoney.]
Alibaba won the battle against Ebay. This is possible because Alibaba becomes a shadow banking system which buyers and sellers transact and put their money for a long time before taking out of the system. Now we are seeing the same for the bike sharing space that mobike and ofo are becoming online depositors. Along the same timeline in 2015, the cryptocurrency begins to realize its possibilities when bitcoins were beginning to be accepted at 160,000 merchants. The idea of blockchain started its ignition and fintech continues it rise. Yet at the same time in 2014, Google acquired Deepmind, one of the startups in using deep learning to solve ground breaking computational and thinking challenge to win Go world class players. In 2015, Andrew Ng was at Baidu as the Head of AI Research from Google and he mentioned that computer wouldn't take over the world just yet. Yet fast forwarding to 2017, according to Business Insider, banks are beginning to discuss issues on digital transformation. Banks in China are using fintech to drive change. One of the clear use case in China is to use deep learning and computer vision to enable merchants to accept payment using face instead of signatures. The other use case is to apply deep learning to determine the possibilities of fraud in applications for funds or securities. In the latest of Vlerick Business School’s FinTech Futures series in Brussels, some of the world’s leading AI and Fintech experts highlighted how deep learning is the key to extracting true value from big data.
“AI and deep learning are about understanding what is needed to train the machine, what it takes to actually make things work. We have language translation, event prediction, handwriting recognition, logo detection, pedestrian detection...."
In 2016, CloudTweak mentioned that, Fintech is on the cusp of a truly revolutionary moment, the integration of AI and deep learning into financial services. This combination really has the potential to revolutionise money and the global financial landscape in ways we never could have imagined 20 years ago. This lead us to consider why fintech can effectively use Deep Learning. [Note that there can be exception to this observation for fintech.]
Financial information such as interest rates, price, debt, equity, creditors, debtors are easily understood and transparent. Anyone can access the internet to any stock exchange or central bank to retrieve public information and this set of public information is transparent and can reconcile with the standard of other stock exchanges and central bank. In each country, there are clear standard reporting rules for listed companies and clear standard regulatory definitions on transaction payments for ecommerce transaction or any online or cross border transactions. This enables clear and understood process in applying deep learning:
- Selection of parameters / features/ variables
- Understanding of clear specific context for analytical studies
In addition, all the financial institutions and regulatory bodies come together to form consensus on international settlement of financial capital and its associated information. However we don't see that for HR.
At the moment, if you go to any job board and any job database by each country, you will notice that there isn't a clear and universally accepted definition on the attributes of a job. The current attributes like years of experience , age, attitude and knowledge are very difficult to achieve standardization across countries because of the following reasons:
- Businesses do not know how to determine a clear community accepted attributes for a specific job.
- Existing big HR consulting firms like Hay Group, Mercer, Tower Watson and Aon do not open their job family database to enable a clear global accepted standard.
- Attributes are valued differently by different organizations and countries but they are openly communicated like prices and characteristics of financial assets.
- Government bodies do not come together to universally agree on a common standard and definition on the attributes of jobs and HR processes as they have done that for settlement of financial transactions.
How does the openness and standardization of information determine the effectiveness of deep learning?
According to IEEE journal on Article: Big Data Deep Learning: Challenges and Perspectives, one needs some contextual understanding to train the machine (known as supervised learning) using label data (data that has meaning and clear understanding). If applying deep learning is like training an individual to recognises patterns, one has to understand the context and inference on defining common parameters to be able to see patterns and infer patterns before even applying that patterns to do a specific action.
For example , a company HR manager seeks 2 headhunters to find the Head of Data science. Yet, often times, the HR manager doesn't know the initial purpose of that position given he or she doesn't have prior knowledge and there is a new role (in other words there is not precedent role to reference from). In some organization, the need to define a data driven culture and a set of data strategies is deemed as key for laying the foundation for the company. In other organization, the need to solve a specific problem using a specific tool on a specific architecture sets the purpose for that position. In addition there are different skillsets in addition to programming skills that the Head of Data science should possess. Yet without a clear open definition of this job in the community, the HR manager is unable to define the prerequisites to the job. As such most HR resort to relying heavily on using culture as a determinant and relying on using interview to predict culture fit using their brain and emotion (both of which are heuristically determined). There is risks of wrong hire due to potential bias at the subconscious level:
- Preference for a certain ethics or race or skin color
- Preference for a specific schools
- Preference for a specific personality traits
As such new HR tech startups try to solve these problems and claiming to do it well on the basis of personality, cultural and psychometric traits. On the other hand there are other HR tech startups that focus on using cognitive tests to assess the talent potential. Like any standardized tests like GMAT, SAT, it is a useful indicator because cognitive tests measure attributes of the human natural brain's processing abilities of information. Yet we all know like any GMAT and SAT test, it is often a snapshot of the individual talent and it doesn't explain how it fits well with the job. The talent potential to job position fit remains only a one sided equation at the moment.
So what is the next step?
The next step for HR is to have a open clear taxonomy like Finance. Clear taxonomy can work if there is international standard to defining HR and sharing HR information just like financial information. After that, the next step is real time and transparent HR information to enable the quantum leap in HR tech revolution.When there is clear, standard and transparent HR information, supervised learning and deep learning will begin to make explainable sense to advancing mankind.
I'm optimistic about the future of HR. This day will come. :)