Intro
The new hit Korean drama Start-Up on Netflix has been creating a lot of buzz on the internet. I received a lot of messages from friends and relatives about their questions regarding this show, as I was one of the few they knew who worked in the field, and decided to give the show a watch, and while it’s been frustrating at times, it’s been pretty enjoyable so far. It follows the romance of three main characters and their trials in starting up a new company. One thing in particular that caught my eye was its focus on Silicon Valley and start-up culture in America. Among the cast are three men, Nam Do San, Kim Yong San, and Lee Chul San, also known as Samsan Tech who are introduced as a group of skilled software engineers, but inexperienced entrepreneurs with dreams of going to Silicon Valley and working with companies like Amazon and Google.
Just like Start-Up mentions, Silicon Valley in the San Francisco Bay Area is where a majority of technological companies are based. Venture Capitalists (VCs) often have their offices here, making it easily accessible for anyone trying to pitch unique ideas. Incubators like the show’s Sand Box are created in order to allow early stage startup companies to gain more guidance with go-to-market, monetization, and engineering strategies as well as connect with other like-minded individuals to create a culture of innovation and creativity. With the rise of AI, there are boundless opportunities to incorporate innovation into your company whether it is used to help small business advertising or creating a novel medical imaging device.
Key Takeaway 1 — Computer Vision and Image Recognition Software (Start Up vs Actual Modern Day Developments)
Samsan Tech is heralded as a group of prodigious engineers, and early in the drama, they win an international competition for artificial intelligence with their image recognition software. Nam Do San describes the work to his parents as the ability for a machine to identify objects by their name and what they are. He later explains to Seo Dal Mi using an analogy where a computer is like Tarzan, having no knowledge of the outside world, trying to please Jane. Through repeated trial and error, Tarzan eventually starts to learn what Jane likes and dislikes, gathering data through his failures and successes. While this was a rudimentary explanation, it is indeed accurate. The artificial intelligence in development today is trained with hundreds of thousands of data points, and through repeated trial and error, it is fine tuned to accomplish what the developers want.
While the image recognition developments in the initial demonstration of Samsan Tech may seem far fetched, they in fact are very real! The field of image recognition is part of a larger field within computer science known as computer vision, which deals with how computers can understand photos and videos. The competition they enter is known as an image recognition challenge, where various competitors train algorithms to identify photos of everyday objects accurately and quickly. This competition actually happened in real life — the most famous of which was one known as ImageNet. During this contest, computer scientists from all over the world competed to see who could classify over 1000 different types of objects the fastest and the most accurately.
Today, computer vision is present in our everyday lives. While in the drama, Tarzan is again referenced as a smart car, real car companies like Tesla and Waymo rely on top tier performance from computer vision algorithms to produce safe self driving protocols. For example, self driving computer models use computer vision to differentiate a red light versus a green light, or whether there are people walking across an intersection. As you might imagine, speed and accuracy are both imperative, since you don’t want to be in a vehicle that doesn’t know how to respond to emergency situations. Luckily, these algorithms and models are nearing top-notch accuracy, and autonomous vehicles may not be too far off in the future!
Key Takeaway 2 — How does AI work?
As seen in the show, AI often can work wonders in our life. One of the first things we see in the drama is Yeong Sil, an AI voice assistant that is built to respond to simple voice commands and Nam Do San’s image recognition software correctly identifying faces through a camera. Yeong Sil isn’t too far from Siri, Google Assistant, or, dare I say, Bixby on your smartphones, and the Nam Do San’s software isn’t too far from Snapchat and Instagram’s facial recognition software for filters.
However, what is the difference between the voice assistant that Han Ji Pyeong uses vs the computer vision algorithms that Nam Do San develops? The show doesn’t spend too much time going into detail about the real science behind this kind of software, instead choosing to show massive lines of code and freely using jargon such as ANN (Artificial Neural Network) or activation functions. Under the hood, these AI models need tons and tons of data to perform well. All that is simply happening is that your computer begins to recognize hidden correlations within your training dataset, or the data that you use to create your model. Later on, since you have slightly different variations in your test dataset, or the dataset you use to evaluate the accuracy of your algorithm, you can often encounter strange edge cases you didn’t expect. The best models are able to quickly differentiate and correct against this (correctly interpreting different intonations within languages and recognizing different facial structures still as the same person). Nowadays, there are different methods to help improve model performance and accuracy which have all pushed modern computer science and AI forward.
Key Takeaway 3 — How do people scale it into a business?
You might very well now be trying to launch your own AI startup! While this is obviously harder said than done, there are a few baby steps that you can do to begin starting your own AI project which very well might one day add some tremendous value into society. Below is a brief framework to think about during your need-finding process:
What is an area I am passionate about?
- This doesn’t have to be CS related! (art, policy, healthcare, finance)
Does a pernicious problem in this industry exist?
- What is the magnitude of this problem? How much does it cost people in $/year, or how many people are affected?
What can be done in this area to increase efficiency or capture value?
- E.g. Certain consumer rights filings could be automated, or create a model to automatically detect severity of COVID-19 caused pneumonia.
Where can I gather the required data?
- Machine Learning and AI models take a lot of data! For your task, what would you need to use as your dataset?
Begin building towards a prototype.
- Don’t worry too much about performance at first, this is simply your first time around the rodeo. Focus on creating a defensible and unique model that has the potential to scale.
Closing Thoughts
Many of you watching Start-Up are probably super excited about learning more how computer science and AI works. Here at Persolv AI, we strongly believe that AI education should be made accessible to everyone, so none of our bootcamps require previous programming experience. Our core mission is to expand AI literacy, since it isn’t a technology built for just computer scientists to use, but rather a tool for everyone, ranging from artists to businessmen to medical professionals, to harness the power of AI. Persolv AI has a handful of Stanford Lecturers and Teaching Assistants dedicated to teaching you the most applicable fundamentals of Artificial Intelligence and Machine Learning through hands-on projects, and to guide you through your own personal project to help you jump-start your foray in the computer science world!
Applications for our AI + U bootcamp are currently open until 12/18! You’ll learn everything from the fundamentals of Python to creating your own AI Project. Apply at www.persolv.ai/apply.
