Your Billion-dollar AI Strategy
How to win in the new AI world
“The world’s most valuable resource is no longer oil, but data” — The Economist
I’ve got some good news and bad news about your data. First … the good news … yes, your data is your most valuable asset.
The bad news? Unless you have a clear strategy for turning it into an AI asset you will lose to someone who does.
Increasing ROI from more data with deep learning
As I have explained previously, deep learning is a big deal because performance continues to improve as data is added.
Feeding the AI algorithms more data is your only sustainable, competitive advantage.
Only proprietary data provides an ongoing competitive advantage
Better AI performance is achieved through:
- faster computers,
- better algorithms, and
- more data.
Unfortunately your competitors have access to the same assets.
Faster computers are commodities
Deep learning algorithms rely on matrix multiplication, the same calculation used to render video games. AI Engineers are achieving performance breakthroughs by running their algorithms on the same Graphical Processing Units (GPUs) originally designed for games.
As GPUs improve your AI engineers will build larger, more complex algorithms faster. But these improvements won’t create a competitive advantage because everyone has access to the same hardware.
Everyone has access to the newest AI algorithms
Google, Baidu, Facebook — even secretive companies Apple — are all publishing their research results. Even these tech behemoths can’t keep up with the torrent of AI breakthroughs released on sites like arXiv.org every day.
These algorithms quickly become available on Github for your AI engineers. Unfortunately your competitors have access to the same libraries.
Every product team can download public data
Training data is expensive, so most AI researchers and startups rely on publicly available data like ImageNet or Wikipedia. But since anyone can download public data it won’t provide a competitive advantage.
What’s left? Your proprietary data
Your only source of ongoing competitive advantage is training your AI algorithms with better proprietary data than your competitors.
Your organization must become a data-generating machine
Ever wonder why companies like Baidu and Google are giving away so many free products?
Why doesn’t Google put ads on Google Translate?
They give away these products because the training data generated by user feedback is more valuable than the incremental ad revenue. Google and Baidu are competing for the best translation engine by driving more data through their AI algorithms.
Their entire AI strategy is based on generating more training data.
The AI performance loop drives your data strategy
Data is your oil.
You need a strategy for generating more training data through partnerships, new products, or research.
What data? The data which will make your AI products better.
You’ll start with an AI product release, ideally something simple which leverages transfer learning techniques. You then do performance evaluation with analysts and customers. Your AI engineers will look for a few quick wins by tuning algorithm parameters and hardware.
While you optimize your current platform your data scientists will perform data gap analysis to identify data necessary to achieve the next breakthrough. Your product manager will then need to create a data acquisition strategy for buying or generating it.
Finally you will acquire the data and your engineers will retrain the modelsand the whole cycle starts again.
The billion-dollar AI prize goes to those who can iterate through this loop the fastest.
Get a free early copy of my new book
Wondering “what do I need to know about AI?” but don’t have time for courses? Get simple, practical AI advice for product & corporate innovation teams. Get your free copy here.