HashThySelf Happenings

What exactly is the value of a technology? Sometimes, when I talk about HashThySelf to people who are likely to benefit from it, the questions that tend to pop up are along the following lines:

• Well, what is the accuracy?
• If I upload a pic of Brad Pitt, will it tell me it is in fact Pitt?”
• Why can’t you make it better?”
• Aren’t there face matching technologies out there that are far more accurate?

We do, of course, have internal metrics by which we gauge the accuracy of our face matching procedure. But it turns out, as it does with almost every technology today, that you can pick 95% of the low-hanging fruit with a fairly low investment in time and effort and a few clever trade secrets. The remaining 5% takes a tremendously greater investment, often by an order of magnitude.

What this means, in practical terms, is that every technology has its limitations. Yes, even Google :-). A creative and resourceful product person knows exactly how to exploit what is there, rather than complain about what is not. That’s exactly what we decided to do with HashThySelf.

Instead of going the whole hog to get the technology to CSI-Miami grade and then pricing the licenses way higher than we currently do, we decided to offer a simpler solution that is more than sufficient for the vast majority of users, at a price that is orders of magnitude lower! And, by the way, CSI-Miami-grade face-matching is simply not out there for public use today no matter how much you want to pay, or what other companies claim.

We started off this whole enterprise as a low-budget effort, intending to keep it that way. We developed the technology to the stage that the accuracy was good enough for most uses and made an executive decision to stop right there. This is not to say that we won’t develop it any further; just that we defer that decision to a point when we are convinced of a real customer need for that level of accuracy, and have a customer who is willing to pay for it. Identifying terrorists is simply not in our business plan. Instead we rely on creative people like ourselves to come up with novel uses of the technology as-is. Often we even participate in the brainstorming process with such creative people to jointly come up with compelling use cases (as has been the case with our first few happy licensees).

And so here is the bottom line:

• Sure, there are zillions of companies out there from whom you can license face-matching software
• Some of them claim accuracies and speeds that are too good to be true (and aren’t)
• In truth, only a few (very few) of them are more accurate than HashThySelf, and that too only by 1-2 percent, depending upon which test-set you choose
• You will end up paying an enormously greater license fee for the marginally increased accuracy. So you must be sure that it’s really worth it for you. After all, what’s the point in buying a sledge hammer if all you want to do is drive a nail?
• HashThySelf will work with you to analyze your use case and see how our technology can be put to use for you, to add real value

- A. Hasher