6 ways people are making money with machine learning

source: www.graymeta.com

Machine learning is definitely VERY cool, much like virtual reality or a touch bar on your keyboard. But there is a big difference between cool and useful. For me, something is useful if it solves a problem, saves me time, or saves me money. Usually, those three things are connected, and relate to a grander idea; Return on Investment.

There has been some astonishing leaps forward in artificial intelligence and machine learning, but none of it is going to matter if it doesn’t offer a return on your investment. So how do you make machine learning useful? Here are some real life examples of how machine learning is saving companies time and money:

  1. Find stuff

I’m sure you’ve spent time looking for a picture or an e-mail. If you added it all up, how much time was it? How much money do you get paid per hour? Companies have this problem as well. We are all totally swamped in digital content. We have files and folders everywhere, and they’re filled to the brim with stuff. To make things worse, we’re not tracking it very well either. Platforms similar to what my company GrayMeta makes are being used to scan everything businesses have, and run things like object recognition, text analysis, speech to text, face recognition etc. to create nice, searchable databases. There’s a serious reduction in time people now spend on searching for and finding stuff. That savings is much greater than the cost of the platform. Tadaaa! Thats ROI baby.

2. Target your audience

One of the biggest problems advertisers have today is people ignoring their product. I admit I find 99% of ads annoying and irrelevant. I go out of my way to not click on or look at ads. The problem is that ads are still too broad, and usually don’t reflect my personal interests. Platforms that advertise want to fix that with machine learning.


Companies that provide content to viewers are now using computer vision and speech to text to understand their own content at a far more granular level than before. This information is then dynamically used to drive what ads you see during or alongside the content. Are you watching a movie about dogs? Don’t be surprised to see an ad about dog food. More relevant ads mean more engagements, more engagements mean more money.

3. Be more efficient with storage

Did you know that most cloud storage services have different pricing based on how quickly you want your content? Stuff stored in a place that is instantly accessible costs you about $0.023 per GB. But stuff you don’t mind waiting for costs you about $0.004 per GB. Thats 5x cheaper. News organizations have a lot of interviews, b-roll, and other important footage that they’re moving to the cloud. Lets say they have 100TBs of content. To access that quickly (because news happens fast) they keep 100% of that content on the more expensive tier. That costs them $2300 per month, or $27,600 per year.

Now, they’re using using machine learning to decide what content should be stored on the more expensive tier. Trending keywords on social media initiate a query in a database that has granular metadata for every video (thanks to machine learning). Positive matches to that query initiate a transfer of that video to the more expensive storage. The company can now store the 100TBs on the cheaper storage, saving them $22,800 per year.

4. Be even more efficient with storage

It also costs money to use that 100TBs the company above is storing in the cloud. Let’s assume, that by the end of the year, 100% of that content will have needed to have been downloaded, edited, and used for news production. That will cost $84,000. If you don’t know what is in your cloud storage, you have to download it to find out, and that costs you money. Do you have a folder labeled b-roll with lots of video files that you can’t identify just by the file name alone? Thanks to machine learning, people can know what is in every single video without having to download it. They can pull down the exact file they want, instead of entire folders or projects, saving tens of thousands of dollars per year in egress charges.

5. Analyze stuff

Most of machine learning is about predicting things. A popular VOD company takes the list of all the things you’ve watched, when you watched them, what was trending right before you watched, and trains a machine learning model to try and predict what you’re going to watch next. They use this prediction to make sure that that content is already available on the closest server to your location. For you, that means the movie plays quickly and at the highest quality. For the VOD company, that means they don’t have to store everything they own on every server in the world. They only move video content to servers when they think you’ll watch it. The amount of money this saves is extraordinary.

6. Avoid fines and save face

The FCC and other governmental bodies can fine broadcasters for indecent or obscene things like nudity, sexual content, or graphic language. Other distribution partners may just have strict rules about what they can or cannot play. You’d think that it would be easy to spot questionable content before you send it to distribution, but it turns out that studios spend upwards of 120 person-hours just to check stuff before it goes out the door! If you pay these people $20 an hour, thats $2400 per movie, per distribution channel! If you consider that every single country is at least 1 channel, then you have things like inflight entertainment, day time television, prime time television, on demand… PER COUNTRY.. it gets insane. Fortunately, machine learning is saving these companies tremendous amounts of time and money by flagging content automatically. Humans are still needed to review and approve, but the amount of time they spend doing this is reduced from weeks to minutes. This is one of the most significant returns on investment in machine learning that I’ve personally seen.

Machine learning can be a very useful tool in the delivery of your goals as an individual or a massive company. Figuring out how to glue together some cool tech and real problems isn’t easy. Thats why it is always important to consider the usefulness of ideas and the return on your investments.