Now You Can Rent Your Own!
Beware the coming of AI! Or so say luminaries like Bill Gates, Stephen Hawking and Elon Musk. They predict a dangerous journey ahead as these machines continue to develop in their capabilities and power. First, they’ll seem like beneficial servants, then they will take your job and finally, they will enslave and surpass you and all of us! Pretty scary – eh? How about some perspective? I’ll bet you did not know you could rent one right now – for no money up front and pay only for what you use.
That’s right! Once again, the cloud enables what used to seem exotic and expensive into ubiquitous and affordable. This time it is Machine Learning as a Service (MLaaS). Machine Learning is the kind of AI where the machine learns from the data itself rather than be programmed to execute an algorithm.
A good example close to home is learning to recognize distinct recurring patterns in your photographs by examining a large collection of them, then the machine can organize your photos for you. It sounds simple but just a few years ago it was a fiendishly difficult task for a computer.
BTW, real world checkpoint here: if you are an Amazon Prime customer – the service is called Prime Photos and it is free! It uses an Amazon Web Services (AWS) MLaaS called Rekognition to power the offering. Prime Photos is a consumer product but you can access the underlying Rekognition MLaaS for whatever objective you may want. Oh, and for those custom uses the price starts at One Dollar per thousand images processed and goes down with volume.
Any patterns can be discerned with a sufficiently large data set and the right software. Besides image analysis, voice recognition and text analysis are very doable. You can also meet such mundane business needs as customer analytics to predict churn, score leads and segment your market. Pretty good value for almost no money, right? As you can guess the market response is starting to explode. From almost nil a few years ago, we are now at a little over $1 Billion today and forecasts are we’ll hit just short of $20 Billion, annually by 2025. That’s a remarkable 38% compound growth rate.
Ok, where do you get “rent-a-brain”? Well, as you might guess all the big cloud guys have an offering. We have already mentioned AWS. They have AWS Machine Learning as well as prepackaged applications like Rekognition and Lex (for voice recognition); Microsoft has its Azure ML and Cortana, its personal assistant AI. Last but not least, IBM offers the iconic Watson of Jeopardy fame. While these guys command about 73% of the current market, there is a host of new entrants targeting many small and medium size firms and you do not need extensive expertise to use their tool. If you are interested, take a look at these: BigML, and Tenpoint7.
Another development that expands the availability of MLaaS to medium and smaller players is the expansion of open data sets. Remember, the software needs a large enough data set to learn from. Sure, behemoths like Google and Facebook have plenty of data upon which to train their AI’s. Likewise, for the Walmart’s and GE’s of the world, but what about the 190,000+ other firms in the U.S. that are not large enterprises? Open data sets provide a cornucopia of data – for free. And there is a bewildering array of them covering almost any topic you can think of.
Just recently an enormous, publicly (read: free) available database was announced: The Allen Cell Explorer. It is an online catalog of over six thousand 3D stem cell images. Ironically, the process to create these images even used machine learning. Now researchers can utilize this database to predict variations in cell layouts that may foreshadow cancer and other diseases. Who knows what tests and treatments might emerge. Guess what many of them will probably use to enable their work? That’s right – MLaaS. Healthcare and the life sciences are one of its biggest users.
If you think this all sounds like sci-fi (and you have got to admit, it is shockingly close) remember the words of a great sci-fi author, William Gibson: “The future is here. It’s just not widely distributed yet”.
By John Pientka