by Isaac Tan
MALAYSIA – You simply cannot go anywhere these days without hearing a single mention of AI. It is everywhere, and while it is here to stay, it is also still in its infant stages.
At this point in time, AI is mostly appreciated by businesses who can reap the benefits that it boasts. For example, marketers have long enjoyed the ability to better spend their advertising budgets by speaking to the right group of people with the intelligent targeting and enhancements of popular digital marketing platforms like Google and Facebook. Meanwhile, streaming giants like Netflix and Spotify are known for their sophisticated recommendation engines that suggest the right TV shows, movies or music to keep users hooked on their products.
While consumers do ultimately benefit from the implementation of AI in the businesses that they engage with on a regular basis, there are very few instances where they get to interact and experience it for themselves.
In 2017, there were almost 450,000 artificial intelligence publications released, with 31% of the publications originating from the US alone. This is largely thanks to companies like Google, who are spearheading the AI movement through exponentially growing the number of AI papers, acquisitions (e.g. buying British AI company, DeepMind, in 2014 for $400 million) and initiatives (e.g. driverless cars through Alphanet subsidiary, Waymo).
Unfortunately, the impact of AI is barely noticed or appreciated by the general public. Although there are headlines everywhere about the implications of these technologies, people tend to react with a feeling of uncertainty and concern – as with anything new that we are unfamiliar with.
So why is the deployment of AI falling behind?
1) Talent shortages
The tech boom that began 15 years ago introduced many new job roles to companies. Suddenly there was a surge of demand for product Mmanagers, software engineers, digital marketers and UI/UX designers. Many of those initially recruited to fill these roles were self-taught, as courses or degrees to learn the necessary skills were far and few in between.
The current AI boom is generating a similar sense of déjà vu, as companies struggle to hire talent that can drive their AI-enabled strategy forward. Suddenly, everyone is looking for AI product managers, machine learning engineers, and data scientists. To complicate matters, larger companies have allocated so much budget to hire for these roles, which means that startups will find it extremely difficult to compete if they do not raise enough funds.
Based on our own hiring experience, even those claiming to be AI product managers or engineers lack sufficient knowledge to mobilize a project from theory through to implementation. A lot of time needs to be invested by the company to make sure that they hire based on the person’s potential to grow into the role, or to otherwise find someone who would fit the role down to a tee.
2) A lack of knowledge sharing
Any initiative needs a loud enough voice in order to grab the attention of others. Many successful product launches have gained the traction simply because of the sheer volume of people talking about them. Ultimately, the more support there is for a particular product or service, the more investors that will pay attention and the faster the industry will grow.
AI knowledge is still confined to a relatively niche group of people who were either fortunate enough to have learned from pioneers in the field, or work for companies spearheading the movement. This makes it difficult for everyone else to appreciate AI as they’re not exposed to it nearly enough.
More knowledge transfer needs to happen beyond the technical community behind AI, so that people can begin familiarizing themselves with the concepts and become excited, rather than fearful.
3) More AI education
Andrew Ng has long been a key figure in the field of artificial intelligence. He is part of a small handful of educators who are making AI education accessible to the rest of the world. AI is not being widely taught in educational institutions across the world just yet. So when Andrew Ng introduced machine learning courses on Coursera, the online learning platform that he founded with fellow Stanford professor Daphne Koller, it opened the door to a world of AI education for many.
Now, Coursera hosts 200+ AI and machine learning-related courses in over 20 languages across different degrees of specializations. Whether you’re looking to be a computer vision engineer or learn how deep learning can help your business progress, there is something for everyone.
In the field of AI, you can’t just rely on one general solution that engineers can build and apply across different industries. Real-life implementation of AI needs domain knowledge, and the only way we can increase the number of people with this knowledge is through education. AI engineers have the capabilities to design the algorithms needed to execute a particular set of tasks, but they need to be connected to domain experts in order to know exactly what to do. For example, an engineer can design a system that can recommend the right selling price for a house, but he or she won’t know what factors to consider without the advice of a property expert.
4) Difficulties of building a large training dataset
For a proof-of-concept, it is relatively straightforward for a team to spend a few hours a week to generate a dataset in order to validate their hypothesis. But as soon as that needs to be translated into a fully functioning product, you can’t rely on the same team for the training data anymore.
In fact, even hiring a separate in-house team to focus solely on preparing training data is extremely time-consuming. The hours needed to screen, recruit, train, manage and retain that team alone will take up to a month at best. Many companies lack the processes and capacity to handle the load of training data that is required, but they do it anyway just to get their products from ideation to launch. The fact that 80% of a data scientist’s time is spent on cleaning and pre-processing data is enough to show that there is not enough time focused on researching and developing AI applications instead.
Where one company focuses on building advanced AI programs, another company can be hired to focus solely on the preparation of the training data. The AI industry is an ecosystem of companies that need to work together in order to turn theory into reality.
AI is a collective effort
I like to say that “people don’t know what they don’t know”. Getting everyone up to speed with the endless possibilities of AI technology is a collective effort. We need more innovation and frameworks for AI, and we can’t get there with endless theorizing. Big and small businesses should work together across industry lines to start innovating and implementing new frameworks for AI in order to encourage higher adoption.
As different industries have different requirements for data, machine learning models and outputs, it’s essential that we gain the necessary knowledge in the industry to make a bigger impact in the development of artificial intelligence. In order to make that happen, we need to innovate and create an efficient and effective system for each other to acquire familiarity on different industries.
Isaac is a Product Manager and data geek. He is currently working at Supahands to bring innovative ideas to life by helping the team build the world’s most efficient workforce by combining machine and human intelligence. You can typically find him combing over mountains of data, working between design and engineering teams in delivering cutting-edge products and services for both business and users.