It’s Easier Than You Think
My friend Greg Head says: “ADD before OCD.” Greg is an advocate of narrow focus for new businesses.
Just like Amazon focused only on books when it launched, and Facebook focused only on Harvard students when it launched, Greg advocates that new businesses focus on one thing, and should be awesome at that one thing.
Greg also teaches that in order to find that focus, it may be necessary to try many different approaches before discovering the one thing to focus on…hence, “ADD before OCD.”
This approach of “ADD before OCD” is analogous to how some powerful AI search algorithms work. It may come as a surprise, but many AI search algorithms are a combination of randomness coupled with trial and error.
For example, Simulated Annealing and Random Restart Hill Climbing are classical search algorithms that enable many powerful Artificial Intelligence applications in the modern world, such as designing printed circuit boards or to search for the perfect mixture of ingredients for beer. While very powerful, these classical AI search algorithms are surprising simple to understand.
The algorithm is as follows:
- Start your search randomly, and check your results.
- Now try something else that is similar to your previous search. If it yields better results, use that one instead.
- Once in a while, randomly try something else. And keep repeating this process.
This is oversimplified, of course, but the concept is accurate.
This AI search algorithm is how we identify what careers we want to work. It identifies what exercise programs we choose to engage. It even resembles how we date and choose a partner.
In other words, during the initial phases of a career search, an exercise program search, or a significant other search, we often try many approaches until we discover an optimal goal state. The discovery process involves lots of trial and error, and often a degree of randomness in our choice of what to try next.
These classical search algorithms work the same way, and are a great way for an entrepreneur and startup to identify key components of their business from product to target market: Try many different approaches, throw in a degree of randomness, and as various approaches begin to bear fruit, hone in on what’s working.
Another class of very powerful AI search algorithms is called Genetic Algorithms. Genetic algorithms are loosely based on evolution, and are used in the code that creates advanced robotic design. Once again, what seems mysterious and complex is actually quite simple.
For example, suppose we are searching for a block of 6 numbers that contain all prime numbers. The search for prime numbers is a contrived example for the sake of simplicity. We start with 3 blocks of 6 numbers (see the first column below) and assign a “fitness score” to each block based on how many total primes numbers are in the block.
In our example, the first blocks scored 50%, and two other blocks scored 33% each. The first block “reproduces” with the second and third block by crossing it’s “genes” to create new blocks (see the new blocks in the second column). Our two new blocks now have a fitness score of 50% each. Pretty good!
Finally, the first block in the second column experiences a “random mutation,” changing the “10” to a “13,” thus improving his fitness score.
Again, this is an oversimplified example, but the concept is accurate. And this is exactly how great new products and services are created.
Let’s rewind to the turn of the century and apply the “fitness test” to a few existing products. Let’s give the Sony Walkman a 40% and let’s give digital storage a 60%. Now perform a genetic crossover…what do you get? The first-generation iPod.
Fast forward a few years and perform a few genetic mutations like online streaming and compact size, and now you have an organism at the top of the food chain. This is how genetic algorithms work. And this is how you can utilize the concept of genetic algorithms when designing a new product or service.
So, you may be saying, “Great, I get genetic algorithms and how they can be useful, but I have a constraint…I don’t have access to millions of dollars to build hardware.” Now we’re talking about a class of AI algorithms called Constraint Satisfaction.
The idea behind constraint satisfaction searching is to first establish constraints on your search. “Constraints?? I’m an entrepreneur…I can do anything I put my mind to!” Of course, you can! But sometime constraints help to identify solutions faster.
For example, many startups have launched web applications utilizing the “man behind the curtain” strategy (their web app doesn’t really work, and all requests are actually handled by humans). When the constraints of money are very real, it’s best to work within those constraints rather than pipe dreaming or waiting until the constraints are removed. In fact, constraints can spark creative ideas that lead to very successful endeavors.
I’m reminded of a story about ice cream. Up until 1904, the customary way to serve ice cream was on a plate. At the 1904 St. Louis World’s Fair, an ice cream vendor was doing great business. Business was so great, that he ran out of plates. The ice cream vendor was forced to work within a constraint: he had no plates, lots of customers, and couldn’t leave his booth.
The booth next store was selling crispy waffle-like pastries. This constraint sparked an idea in the ice cream vendor. He bought several of the crispy waffle-like pastries from the pasty vendor, rolled them up like a cone, and scooped his ice cream on top. A creative solution bound by constraints that gave birth to a whole new business.
You Will Soon Be Able to Code Your Own AI
I’ve described a few of the most common and powerful AI search algorithms. Once explained without all the math and code, you can see how conceptually easy they are to understand.
In fact, in the near future, the math and code will not be necessary. Thanks to advances in speech recognition, natural language processing, and robotics, the creation of new AI search algorithms will be accessible to anyone with a creative and novel search system. That means that an entrepreneur, an investor, or a business owner will able to create advanced search algorithms and decision heuristics simply by speaking and interacting with an advanced user interface.
It’s exciting because the technology opens AI advances to anyone around the world with a creative mind, and offers opportunity to people who may never have had the chance to learn advanced mathematics and programming.
Of course, AI is more than just creative search algorithms—just like a business model is more than just finding the right model. Once a business model is found, planning must commence in order to effectively implement a strategy.
As such, AI also encompasses intelligent and creative planning algorithms. AI planning algorithms are incorporated by airports every day to plan the thousands of flights around the world; AI planning algorithms are utilized by military to deploy troops and weapons; and AI planning algorithms help businesses strategize long-term product rollouts and resource allocation. But that’s the topic of another article.
For more on the topic of AI Planning algorithms, check back next week. If you can’t wait, contact us directly and let’s chat about how AI can support your business goals.