Signing off for now
Hi everyone. Well, this has been a heck of a week. As I'm sure you've heard by now, Donald Trump has been elected president of the United States. In the face of this danger, we must all hang together, or most assuredly we shall all hang separately.
On a personal level, I've been turned down by Y Combinator. Of course, you should always expect most investors to reject you. However, I think this wasn't just random bad luck, but part of a general investor mindset that I hadn't appreciated earlier. The last generation of big software companies were almost all "lightweight" - Facebook, Twitter, LinkedIn, Snapchat, Pinterest, Slack, Evernote, DocuSign, SurveyMonkey. Even Uber and AirBnB, which sell "physical" products, delegate all of those aspects to someone else. Getting version 1 of these companies working, on a technical level, requires nothing but good web or mobile programming chops. If you have a site like that, and you don't have any real users yet, it's probably because a) you're lazy, b) you really can't do marketing, or c) people don't want the product, and in all three cases the company is most likely doomed. It's not because of hard AI problems that various people have been hacking away at since 1987. (To YC's credit, they recognized that strong machine learning was the main barrier here, but still said they thought we were "too early-stage".)
This, of course, leaves plan B: get money from customers, so you aren't dependent on investors. Unfortunately, I'm not in a position to do this right now. Therefore, I will have to put the project on hold for a few months while I take care of finding another way to pay rent, moving to a more convenient apartment, fixing my furniture, bicycle, and laptops, and hopefully finally getting a dog. Though this is obviously disappointing, it's really not that long in the scheme of things - Konrad Zuse took seven years to make his machine reliable, during which he got drafted by Hitler, and he still got there in the end (though sadly it was later destroyed in Allied bombing raids, which hopefully won't be a factor this time around). I've shut down the CandleCRM server to save on hosting costs, since the machine learning models and database cache eat a lot of expensive RAM. The data is now held by Tarsnap, which is a secure encrypted backup service.
I'm still very enthusiastic about the ideas I've been exploring, so if you know anyone who is interested in semantic graphs, machine reading, epistemological databases, and so on, I'd be very happy to meet them. If you believe Nick Bostrom's technological completion conjecture, then someone has to get there eventually, and if worst comes to worst and I can't possibly do it myself, I'll just join up with them when they come along. It's pretty incredible that we'd all happen to, not just be around in the extremely narrow window when AI is invented, but actually be in a position to influence its outcome, and I won't let that opportunity go to waste.