Most startups these days suffer from a severe bout of FOMO. Whether you’re the new guy at your startup workplace or you are at the startup meet held every week, the talk eventually drifts towards Machine Learning or Artificial Intelligence. The life of a startup is pretty slippery, and good decisions help launch the startup gungho but it could also crash and burn. So when everyone who’s everyone talks about ML and AI, isn’t that how we must turn our heads too? Well, not necessarily.
It’s important to be aware of the buzzwords but it’s more important to know what they mean. But before getting into that, let’s talk how AI differs from ML. When you write a program to recognise a particular animal, say a cat. You will feed it information and images of how a cat looks like. Two ears, four legs, a tail, evil-looking eyes. You train your program to recognise a cat by feeding it hundreds of images of a cat. Training your program to understand or learn on its own what a cat is and then guess is ML, showing your program what a cat is, is AI.
Still a little hazy? Well, video games are one of the top examples of AI, your computer opponent constantly tries to engage or destroy your playable character. And the harder he/she is to beat, the more complex the set of rules were programmed in it. That is AI. A specific set of rules programmed by the programmer for the program to give a tough fight for the character. It knows when to shoot and it probably even knows the best way to hit you is by shooting at your feet. But it will only know this, if you ‘tell’ it i.e. program it or set rules for it.
ML however would learn how to beat you, there are rules, yes but the rules change and it gets smarter and smarter knowing your every move eventually beating you. Yes it will beat you. The AI needs to be told what to do while ML well, the machine learns and it can’t get more straightforward than that. It sounds scary and exciting but it can’t do much away from the main game.
Consider the same ML video game character tries to predict the next possible occurrence of a hailstorm and it will fail. It doesn’t know the rules nor can it learn from it. The program that was used to recognise the dog can’t do squat when it has to defeat a character in a game. The rules are different, the learning is different. It’s like taking a cricketer and expecting him to make a touch down amid all those players in a Super Bowl finale.
Now back to startups. Even though your company uses AI, it doesn’t necessarily mean they’re an AI startup focusing on NLP and deep learning. Advertising yourself as an AI startup without the proper knowledge or without the right resources is spelling out doom for your company. Only when you as a company can build a system with algorithms that are self-learning and can make decisions, it can be called a true AI/ML company.
The vast difference in understanding AI among startups and the direction they’re heading towards, is what makes a startup make or break. But most importantly the lack of science-focused problem solvers who enjoy a lot of math is a gaping problem. It is difficult to find people of the right talent to scale-up but hiring the wrong kind is just going to make things far worse. Competition from AI startups is another problem but if you have scientific experts and employees who are tolerant enough to watch your company grow, then you can withstand the test of time.
Secondly, implementing AI or machine learning doesn’t come cheap. Even the simplest program could cost you a lot due to the data needed, data to be mined, and then finally feeding your program so it can begin learning from it. When you invest in a startup or are a startup yourself, always consider if there is enough investment. Though going with the flock is the easier way into startups, it doesn’t always work.
Startups are usually on life-support for most years before they can finally throw them away, grow their wings and fly into the setting sun. But, yes there’s always a but, the life-support itself may not be enough if you make all the wrong decisions like being part of the flock for example. Machine learning is a lot of work, a lot of cloud server technology and that translates to a lot of money. The lack of money simply means bad resources or lack of good resources itself. Patience is key but so is being smart. Save up before investing in the best AI/ ML project because being short on cash is just being haste.