[Editor’s Note: We are delighted to have Ashish Nagar help break down Artificial Intelligence for us. Ashish is the Founder and CEO of Level AI, a company that leverages AI to help empower call center agents and managers with real-time knowledge support, monitoring and performance scoring. Prior to founding Level AI, Ashish was Product Manager in Amazon Alexa's Conversational AI team. He also helped build two Silicon Valley startups Kinestral Technologies and Relcy Inc. in a leadership role in the Business and Product teams.]
AI is often misunderstood — what is something you misunderstood or got wrong when learning about AI?
One of the key misunderstandings for me as a builder was that developing and improving AI products involved mostly inventing, developing, and refining proprietary algorithms. On the internet, you read about large models launched by OpenAI Google, and others, and you are transported to a world of advanced mathematics, abstractive work and design. The reality in my view is 20% of that, and the other 80% of the work in AI is really about the data which underpins any AI project. Building and maintaining clean data sets is at the heart of building and improving any AI system. There is a lot of engineering and product thinking which goes into building those and I often see that missing in AI projects, companies. As with any system, if you put crap in, you’ll get crap back out.
Is AI a risk or an opportunity for SMBs?
It is definitely an opportunity. SMBs have a unique opportunity to create efficiencies in their business, gather insights and delight their customers with AI. Currently, SMBs are better positioned than ever to leverage AI for their businesses. Over the last 5 years the availability of off-the-shelf tools has increased. This enables SMBs with limited resources to get results with small technology teams or partners. For example, Google’s Auto ML tool allows model development, training and deployment of many AI models suited for a business in a very simple way. This was not even possible a few years ago. So access to these technologies is creating a very level playing field for SMBs.
Business owners have limited time and resources — how can they be leveraging AI / ML in their business today?
Data strategy
Companies of any size need to first have a data strategy before they have an AI product strategy. So very simply, look at all the input points of data potentially in your services and make sure that it is captured, stored in an accessible way for any future data analysis work.
Identifying opportunities for improvement
The next step after having a good system for collecting and maintaining data, in my view, is identifying which opportunities exist in your business where data-driven decision making could materially impact performance.
Hiring a small data science team
Often businesses are discouraged by the prospect of significant time and capital investment to spin up AI projects. With the current off-the-shelf tools available even a small 1-2 person team can create a large impact in an organization. Hiring data science talent also does not need to be about finding researchers in the space, as those tend to be extremely hard to hire and retain. Smart software engineers who are interested in analytics and decent training in mathematics are good candidates for initial data science team hires.
What new developments are on the horizon for AI / machine learning? What has you most excited?
Most of my product and business work in the last 5 years has been in the space of Natural Language Understanding (NLU). This is the branch of AI which deals with the problem of machines being able to understand human language. I am most excited with the prospect of advancements in this space over the last few years and what is ahead of us.
Understanding human language is one of the hardest problems in AI. Our language has humor, sarcasm, storytelling, hidden meaning built into it.
Over the last few years, the rapid increase in computing availability and large amounts of public language data have led to really large language models like BERT, GPT-3 possible. I am excited about the continuous advancement over the next few years.
I am looking forward to the time when a machine can summarize a short story by Salman Rushdie or Garcia Marquez in 300 words.
Help our readers cut through all the noise and B.S. What are the best resources you can point them to for learning more?
I have really enjoyed the Deeplearning.ai courses by Prof. Andrew Ng of Stanford. They are freely available on YouTube and Coursera. There is a course on ML for business people which is particularly good for business leaders to get started.
For beginning product managers, business executives who want to get into building and prototyping, I would look at the University of Michigan Data Science series on Coursera. It is a 2-3 month crash course to AI for dummies.
Finally, one of the books which I loved reading in this area a while ago was Life 3.0