Jason Antic

2/ I'm a very open person. So I don't mind talking about money in public. The free MIT licensed version of DeOldify on GitHub actually has come at a great cost- to me and my wife. First, let's start with opportunity cost: I went part time in May 2018 at my day job to

3/ pursue deep learning. I started at 16 hours/week; then went to 8 hours/week in Nov 2018. That's well over $100K in salary lost right there! Then there's the "monster computer" to run all these experiments. That cost nearly $10K and the electric bill from it has exceeded $4K

4/ believe it or not! So that's money. Then there's the time sunk in: This has been an obsession for 1.5 years now. I've worked pretty much non-stop, both negatively impacting my mental health and my relationships. I'm only now getting that back on track.

5/ That all being said, I'm not asking for sympathy. I'm getting a lot out of this, too, and now the work can continue sustainably because of this frankly wonderful deal we have with MyHeritage. It's great in many ways. We can focus on what we do best (deep learning),

6/ while they take care of difficult things like marketing and infrastructure and a whole bunch of other things that are simply not feasible for us to deal with right now. Us as in @danasday and I. It's two of us and we intend on keeping it that way.

7/ What's the longer term vision? If we can put work out that won't wind up shooting ourselves in the foot with creating our own competition that'll needlessly jeopardize our own business, then we'll release it as open source. That's the idea.

8/ But to be frank, we have to keep inventing to achieve that, so we won't be able to do that right away. And there's no guarantees that this will happen any time soon.

9/ If things didn't cost money and we had all of our needs met already (we're not looking to be rich), then yes, we've have a different approach! But in order to operate in this world and do what we love, we have to be self-sustaining. We are not apologetic about that.