Blog: How Small Companies Can Overcome Talent Shortages in AI
By Devashish Shrestha
While the demand for AI talent rises, its supply doesn’t (at the same pace). Silicon Valley tech giants attract talent in abundance, providing desirable salaries, computing resources, and vast quantities of data. Big companies aren’t the only ones in need of AI talent. In fact, smaller ones may need it more. The imbalance of supply and demand fuels competition making it difficult for smaller companies to nail down the top notch AI talent they need. While smaller companies may not have the means to do everything bigger companies can, there are powerful tools at their disposal well worth investing in. Smaller businesses and start ups can compete by going after things that bigger companies may overlook (or completely miss).
Education & Exposure
There may be no other more effective solution to the AI talent shortage than education; however, our education systems struggle to keep up with rapid advancements in technology. The need for more robust AI education is widely acknowledged but doesn’t have a simple fix. The Forbes article “The AI Skills Crisis And How To Close The Gap” states: “One of the ways to address the AI skills gap is to increase resources for digital, math and technical education.” Many people think tech companies seek data scientists exclusively, and while many AI positions require PhDs, companies should consider hiring people with expertise in physics and math as well. Education doesn’t only take place in a classroom or training session: those looking to become professionals in AI also benefit from hands on experience. In the future, there will be more and more ways to learn about AI on your own. According to Forbes, 63% of companies provide training on data analytics. This number is expected to increase as demand continues to increase.
Nurture Existing Talent
Instead of turning outwards, companies may find it more feasible and lucrative to train existing employees in AI. Fortunately, there are many outlets that those looking to become fluent in AI can take advantage of, such as Massive Open Online Courses (MOOCs), computer scientist Andrew Ng’s Deeplearning.ai, EdX, and Udacity. These resources make it possible for people to skip the classroom completely and teach themselves AI skills from home.
It’s important to keep in mind that companies launching successful AI initiatives require more than technical skills. The emerging world of AI can seem inaccessible to those without the vernacular to talk about it. It’s important to have team members who are skilled at communication and can break down the lingo so people from all professional backgrounds can partake in the conversation. Though the specifics depend on the company, it’s important to have people on board who understand every part of the business process, from problem solving to deployment.
The AI talent shortage is due to many things, and CEO of AI4All Tess Posner reveals that a partial culprit may be the lack of diversity. AI teams at large tech companies are nearly dominated by men. According to the Guardian, 85% of AI researchers at Facebook are male — 90% at Google. AI is not only male dominated but almost entirely white. In 2015, the National Science Board reported that 2.5% of Google employees were black. At Facebook and Microsoft the percentage was 4%. The lack of diversity not only creates and perpetuates an exclusive culture but is also detrimental to the integrity of the industry.
Major failings in many AI systems — such as object recognition and classification and chatbots — are due to a lack of diversity. Machines, like humans, only learn what they are exposed to. It’s not a fluke facial recognition systems failed to recognize faces of people of color while having far less trouble doing the same for white faces. This disaster in computer vision is not random: the machines were not trained to recognize non-white faces.
There should be proactive measures in the short and long run to change the statistics above. There is clearly much to be done, and large companies ought to lead the way to remedy the crisis. One way they may combat the diversity problem by being transparent about what data their models are trained on.
Articles from the piece: