The Question To Ask Yourself When Viewing An ASR Demo

One last point on ASR and its enthusiastic marketing. I’ve hit this topic a lot in the past few weeks. I hit it a few times last year too. There is one very good way to tell that proprietary speech recognition companies haven’t hit the big leagues, and that is looking at the way they’re marketing. It’s behind closed-door demos to people who aren’t likely to know very much about technology. We know for a fact that when you have a product to sell, the giants buy in. Microsoft bought Minecraft off Notch for $2.5 billion. Have you looked at Minecraft? It’s not exactly an example of the latest and greatest technological wonder, but there’s money there, so now it belongs to Microsoft. They didn’t just buy it, they rewrote it from its Java edition to create its Bedrock edition. Another giant buy? Hopstop was a company that figured out how to give really good mass transit directions, and Apple allegedly bought them for $1 billion. That tech ended up in Apple Maps before they quietly killed Hopstop.

“Thanks for the history lesson, Chris. What’s the point?” The point is when you are being told that something is wonderful, new, and that you should buy in, you have to ask yourself why it hasn’t been sold off to someone way bigger than you. With these far less impressive feats of technology being sold off for billions of dollars and tech giants willing to spend real money on promising technology, there’s a solid question as to why a company hasn’t cashed out. Alphabet’s subsidiary Deep Mind has shown a willingness to burn through half a billion dollars a year on AI research. Youtube’s automated captions fall apart whenever there’s a bit of an accent or some music in a video. Ask yourself, if you were running a company, and you had about $100 million in investor money, would you not sell out for $1 billion or $2 billion? Somebody would buy good ASR. Just ask your wife:

To be frank, Youtube’s automated captions dominate proprietary software from what I’ve seen. Voice recognition is open source. Anybody can get their hands on it. That doesn’t mean the claims behind it are true. Take, for example, this blog post, the seven best spots for open source voice recognition software. They boldly claim that it is “more cost effective as the software performs the task of speech recognition and transcription faster and more accurately than a human.” I’m sorry, but if software that is “more cost effective,” “faster and [more accurate] than a human” is available free on the internet, I’m quite sure that Google, and by extension Youtube, would have figured it out by now. As I said over a year ago, there is an indeterminate amount of time and money needed to get this tech from where it is to where people are saying it is. That won’t stop sellers from selling it to consumers or investors anyway. Similarly, we cannot stop being advocates for consumer awareness.

(P.S., you can buy the Stenonymous Blog for about $6 million.)

Leave a Reply