Artificial Intelligence — Examples Of Some Pretty Cool Stuff For Your Business.
This isn’t a techy post, if you’ve wondered what Artificial Intelligence, Predictive Intelligence and Machine Learning is and what it can do for your business, read on.
I’ve just come back from Dreamforce in San Francisco. Salesforce’s big announcement was their huge investment in an intelligence platform called Einstein. Einstein is a collection of products and companies they have acquired such as PredictionIO, BeyondCore and MetaMind.
Artificial Intelligence has been around forever but now thanks to extremely powerful cloud infrastructure, open source models, pay-per-use API’s and the massive amount of data we’ve been collecting, we can actually do some shit with it.
That’s about as techy as this post will be.
What can Artificial Intelligence/Machine Learning (for the purpose of this post assume these are the same thing) do for your business?
Image Recognition
If you’ve got Google Photos on your phone you know you can search for “dog” and it’ll show you every picture of a dog in your collection. The same goes for Facebook facial recognition, it knows what you look like. How did it do this? It learnt. How did it learn? Similar to how we do… Someone showed them the way first by grabbing a bunch of photos and pointing out what objects were in each photo, then the machine started finding variants of those things and assuming they were basically the same. If it made a mistake, someone would tell it, and it would try not to do it again. Every time you tag someone you are validating or correcting the learning process.
One way image recognition is being used right now is in hospitals. If someone comes in with a massive headache at 3am they get an MRI done but if there is no neurologist available the scan is sent to a 3rd party who reviews it to see if there is anything serious.
They have 20 minutes from receiving the scan to diagnose if there is a serious issue but if you have a brain hemorrhage you may only have minutes to get onto the operating table. This is in use today, when the scan is sent it goes through image recognition to detect a potential hemorrhage, if found, the scan is put to the top of the list and given 5 minutes to respond, instead of 20. If a hemorrhage is validated, it’s fed back into the machine and it keeps learning, getting more precise, learning what to look for.
Another way is for field service technicians. If they find an old part that needs replacing for a machine that is 20 years old, they can do a quick scan with their phone and it’ll search every image for that part and return what the part is.
Predictive Sales
- This is already in use everyday. When you go to Amazon or eBay, it will recommend products to you based on your past buying behaviors. The more you feed into the machine and the more you train it through interacting with it, the more refined the recommendations will become. Some companies are taking this further by having the text and tone of the pages change according to your behaviour. If some advertisements are loud and appeal to you, it will recognise that you are spending longer on that page and dynamically change the rest of the site to tailor your experience.
Think of engagement for predictive sales, if a customer is opening emails, going to your page, liking your Facebook page, maybe it’s time for an automated offer or a phone call. The difference between the traditional world and the new is that the machine will be trained to your specific behavior. If you are someone who opens up 1 email and buys, you’ll get an offer early in your engagement.
If you’re a fence sitter that takes a lot more, after an email open, you might get a targeted Facebook advertisement and if you engage, you’ll get an offer. If that doesn’t work it’ll keep refining and learning. Imagine the price finding capability too. The key is that the machine will test and learn.
Predictive Maintenance and fault finding
- This is where big data and Internet of Things will make a huge difference. Internet of Things is anything that’s connected… This could range from security camera’s through to connected bins that notify when they are full. Imagine a connected city where a bin notify’s whenever it’s full. The machine will learn and optimize disposal truck routes based on the input.
This could also then feed into complementary models such as the public transport system. Maybe there is a relationship between garbage and the amount of buses required for transport that day? The great thing is that I don’t have to figure that out, the machine will find the correlation, test and learn… It may obviously also just be coincidence which is where we will always have to play a guiding role.
If you have 100’s of connected printers across dozens of buildings, the machine will learn when a printer is likely to fail based on their activity. Again the key difference is that the machine will learn, not us. We don’t have to say “yeah printers start failing at 10,000 pages so we should service them at 7,000”, the machine will do this automatically… and then learn based on model, age, etc. Now throw financial information at it and it will learn when it’s time to replace a model rather than service.
Language Translation
- Google Translate can now translate as well as humans… It’s only for 2 languages at the moment but they’ve trained the machine to first understand what is trying to be said, the context, the meaning, the message. Then it translates using traditional methods but can do so in context. We’re actually all part of the training… have you noticed how Facebook asks you to rate a translation?
So why can my business do this right now?
Isn’t this the realm of data scientists, silicon valley, and other companies that you hear of doing this stuff but yours can’t.
- Cloud. Okay so what does that mean? Cloud is basically someone else’s computer. That’s all that it is, but that person’s computer is actually a lot of computers and a lot more powerful than yours. Think of opening an excel spreadsheet with 1 million rows… it really struggles. With cloud you have access to super computers and you only pay for what you use, you don’t have to own the computer, you just rent your time, kind of like going to an internet cafe to use a powerful computer for an hour to process your 1 million row spreadsheet.
- API’s. Okay so what does that mean? An API is basically a phone call, and you just pay for the call. Google for instance have a vision API — you don’t know what’s in a photo, you send it to Google, it has a look and tells you what it is. You can build this into anywhere in your business and just pay for the phone call. So Google already has the machine trained and you just pay to use it. Many products are making it trivial to access their features purely by API.
- Features rolled into existing products. Salesforce is a great example of a company that is making these capabilities for everyone. They are starting with sentiment analysis in their Service Cloud and Marketing Cloud, so it will detect if the customer is happy or upset based on the text. It will also classify the type of issue based on the text, and if you end up manually changing it it will learn the new classification. Think of me sending Telstra an email with a brief summary of the issue and it gets routed to the exact person who can fix it and also tells them how urgent it is based on the tone of the message. Companies are making this accessible to you without the need for a data scientist and programmer. If you want to make custom capability, they are also allowing you to build out capability through drag and drop interfaces, the race is on to make this easy for us and it’s going to benefit everyone.
- All of your Data. Think of all the data you’ve accumulated in different systems, about different things. We can bring this all into the Cloud and have standard models mine through this data to find relationships and predict the future, this is easy and available right now. You just need a champion.
- Collective Data. Think of all the data Facebook or Salesforce have. Facebook can offer targeted advertising because they know a lot about you, as well as everyone else. The same with Salesforce, they can provide predictive lead scoring based on your data as well as everyone’s. The more data the system has the more it will learn.
I’m absolutely super excited by what we can do! Have a think about all of your data and what you could do better, automate and augment. It doesn’t have to replace people, it just makes them do their jobs better… and best of all, we can do this all right now.
