What if I told you to tie your shoes, but you had no laces? Or to cook dinner, but you had no pots or pans.
There are certain tools we need to succeed, which we often don’t have access to or are held back by a gatekeeper.
Dozens of AI / Machine Learning startups experience this same problem because they don’t have enough data to properly train their AI algorithm. Startups that aim to eliminate the error involved in judging cancerous tumors. Startups that aim to improve personalized medicine and create a healthier future.
But, no different than unprivileged schools which lack the books and quality teachers to educate their students, many startups lack the resources (data) to educate their AI and advance society.
How do startups currently amass the data they need?
They can collect data themselves. However, as Auren Hoffman points out in “Where Should Machines Go To Learn”, even companies which raise $800 million have very little data. For instance, Uber had to invest $500 million in map data, just to create a better mapping system.
Perhaps they can broker a major data-licensing deal with Facebook, Google, Amazon, or Uber. Yeah, right. To broker a deal with any one of these companies would take time, rare connections, and mutual benefit (what can a small startup possibly provide Google?).
So, if the only two current options are to collect your own data or go to the big guys, which takes too much time, money, and is virtually impossible, then it sounds like a new option needs to be created.
What’s a better solution?
The logical solution to this problem is for Facebook, Google, Amazon, and Uber to allow companies to rent their data. Instead of brokering one-off data deals, they could provide data to the masses at once. No different than how Amazon rents Cloud storage, they could rent out specific datasets which they’ve already used, so startups can gain traction on teaching their AI algorithms.
Obviously, since a lot of this data is defined as PII (Personally Identifiable Information), it needs to be handled with the utmost security. Nonetheless, I’m confident there are enough data scientists and security engineers to build the proper infrastructure, allowing “Big Data on Demand”.
Why should Big Data care?
We don’t neglect underprivileged children at poorly resourced schools. No, we work together to improve their circumstances so they can have a fair shot at learning and succeeding. It just so happens, in this case, there are a few “people” that have all the books and are holding out on these children.
In my opinion, it is the duty of those more fortunate data companies to distribute the necessary data for other companies to succeed. I’d love to hear what you thoughts are on “Big Data on Demand” in the comments below.
We all have skills, resources, and privileges unique to us that can greatly benefit others when shared. That’s why I created Quick Theories–a brief, weekly newsletter where I share creative insights to help inspire your next idea. If you could use more creativity in your life, sign-up here: quicktheories.com
Making healthcare a little bit better via Redox