Well I’m getting super close to super stable super awesome elephant in your hand. Lots of cool progress today. My main problem now is that sometimes it decides to get smaller and smaller and smaller for no reason. It’s like Alice.
I was down last night. My birthday, first without my mom, tired, it all got to me. She never liked me owning my own company, she liked it when I was part of something stable, but she loved me anyway. One of my first memories of her is of the Apollo moon landings, she woke me up at about 3 or 4 am in the UK to come watch James Burke and Sir Patrick Moore narrate the journey to the surface. We watched it on our 14” PIE TV set which had been in pieces two nights before as my dad was changing our all the vacuum tubes. I remember them all on the floor on a table cloth all laid out neatly. But last night we watched the pilot to “For all mankind”, which is wonderful by the way, and the porch light kept flickering off and on and I couldn’t help but think it was my mom reminding me of this story. Because every chance she had to remind of how she woke me up for the moon landings, she was there for that.
The strange thing about the elephant in AR video above is that it’s actually 2D. We place and track the elephant and guess the scale based on the size of the tracked object (your hand) so the bigger your hand gets, the bigger the elephant etc. The problem with that is, well matching it to the video. The machine learning output is basically “here, at this X and Y and this scale” and you have to try and work out what that is in X, Y and Z. Let’s just say the blueprint is very pretty (picture attached).
Machine learning is all just “here’s some 2D data from the 2D data you gave me” and then you make it look like you know more.
I fought and lost today with AWS on getting more virtual CPUs, my first conversation, which I thought was with a human, was actually with an AI. And then I talked to a human who said he needed to review it. I actually think the second “human” is actually just a second AI with more virtual CPUs attached to it and that by that AI telling me the first contact was an AI I’m now lulled into thinking I’m talking to someone real.
I also fought and lost today on setting up a single URL that automatically expands instances. Mostly because the documentation released from Epic (just recently) says, and here it is in it’s entirety:
"This solution supports launching of one instance of a UE4 Pixel Streamer. Once you have a successful instance you can create an AMI and setup an Auto Scaling group following AWS best practices. You would need to track the launched instances and IP addresses that are available to suit your use case.”
Super super useful.