Jump to content
  • Advertisement
  • entries
    89
  • comments
    338
  • views
    122040

July/August Update

slayemin

983 views

The biggest struggle for me is still money. It's getting harder. Game sales have pretty much stopped completely, but game development continues forward. I'm starting to think I'm a bit crazy.
The rational side says, "Why are you still working on building a product which literally gets zero sales? It's time to move onto something that actually makes money."
But the emotional side says, "But I believe!!!" (bursts out in song) and then it tries to rationalize it by saying that "but... but... I just need more compelling content! Then, sales will pick up naturally!"

Reality check time: I have proven through analytics that adding content patches through steam updates does not in fact, increase sales or even viewer traffic to my store page. I could release product updates every week and that would not affect my sales numbers. The only possible way a product update would affect my sales numbers is if the product becomes good enough that people who own the product tell their friends about it. It's not there yet, so that's why I keep working on it and barely scraping by. Eventually though, I'm going to have to shift gears from product development towards product marketing and advertising.

Barely scraping by seems to be the name of the game for 95% of everyone in the VR industry right now. There have been some interesting recent developments lately. Owlchemy Labs, the creators of the smash hit "Job Simulator" and "Rick and Morty VR" have recently been bought out by Google. Google now completely owns one of the best VR content creators in the industry. The founders probably got hella rich and don't have to worry about anything but creating cool VR content now. Lucky them. One of my friends works for a local VR startup as their only programmer, and things are getting so tight that he had to get a second retail job in order to get by. The startup is too broke to pay him and their sales have dwindled as well (everyone should expect the long tail and budget for it!). The other huge development lately has been that AltspaceVR has shut down. They were a 35 person VR company which created a social hangout within VR, similar to Second Life with VoIP. They were funded entirely with venture capital money. I can't imagine the stress and heartbreak that brings to the team. But... 35 full time employees. Damn... and you have to make payroll every two weeks for 35 people? And you have a product and business model which doesn't involve bringing in money from users? Your days were numbered... I'm fascinated by why various companies fail and succeed. Obviously, creating and having a product is not the entire picture. It's all about making money to sustain your business operations.

My business operating expenses are extremely low. I pay $400 a month for my rented office space, $8.91 per day for burritos, $2.50 for a one way bus ticket, and $2.48 for a cup of black coffee. I owe people money, so I have to pay them off before I ever pay myself. Realistically, my chances of making a lot of money in the near future are near zero without funding and support. But hey, my operating costs are so low that I can almost do this indefinitely. My company will survive. It'll be small, but it will survive and continue forward, scratching out a teeny bit of money.

Speaking of money, the most profitable area right now is doing VR contract work. I've been working on a couple different side projects for various local companies, creating VR experiences around their products and services. I'm about to start working on an interactive VR film proof of concept, which plays sort of like a "choose your own adventure" 360 film in VR. It's going to be an interesting twist on interactive cinema. The broader goal for me is to learn as much as I can and broaden the scope of my VR designer skills. I've become a part of the production cycle for creating VR media and I'm bridging the gap between gaming and cinema within VR. Here is a sample of a VR app I made for Dell in May:

I was thinking critically about this on my bus ride to work this morning and I realized something important: Is watching a cinematic in 360 stereo really VR? Why/why not? What's missing? The viewer. Who are you when you're viewing these 360 videos in VR? Okay, what kind of defining rule can we create which differentiates VR from fake VR? My tentative rule is that the viewer has to be a character within the experience for it to count as VR. The important thing here is to create a sense of "agency" and identity with the viewer. So, the cardinal sin for a VR designer is to take away agency from the player (such as controlling their head or playing a cinematic). The follow up question: "Does it really matter?" Yes, it kind of does matter because everyone is doing it wrong and calling their creation "VR" when its not really VR. It's really challenging to start defining what this new medium is and is not though. I think the guiding principle I use is that "Virtual reality should be indistinguishable from reality and the human experience." People can turn their heads and look down at their body, move their hands, feel solid objects, etc. The closer your VR gets to reality, the more you can confidently call it VR.

My challenge will be to convince companies to see it my way and spend the extra money to move from a stereo experience to a VR experience. I don't know if that's a battle worth fighting. The challenge with 360 video is that the video itself doesn't lend itself to user agency. The camera is placed on a tripod and people act out a scene all around the camera. So, the person experiencing the 360 video can't move around in the scene as if they were a part of it. The solution might be to ditch the 360 camera completely and go with motion capture and animated characters within a 3D environment, but that will mean much higher production costs and longer timelines. At the end of the day, what does a client care about? Accomplishing their objective, whatever that may be. Where does the line exist between exerting my subject matter expertise and satisfying the customers objectives?

Anyways, I am slowly realizing that I'm no longer just an indie VR game company, I'm becoming a VR media company.

I wasted the entire last week watching "The Internationals" Dota2 tournament. The game itself is somewhat interesting, but more interesting is the growing rise of E-Sports. I think it's going to disrupt the definition sports. Every year, the Dota2 championship match grows in popularity and the prize pool grows by millions. I think last year the total prize pool was $16 million. This year, it was $24 million. All of the money comes from the Dota2 gaming community. The final championship match had 4,700,000 viewers around the world watching it unfold. I watched it on Twitch.tv, and the channel had about 380,000 live viewers. The sports stadium down the street supports about 68,000 people. So, just on Twitch, we had about five full stadiums worth of people watching the event online. Think about all of this for a moment: 4.7 million people watching ten people play a video game against each other for $24 million. If we project the trend out, over time we can predict that next year the prize pool will be even larger and the viewership will match proportionately. On a broader trend, I think E-sports will eventually eclipse conventional sports. Football is currently the most watched sport in America, but maybe in 30-40 years, E-sports championships will be the most watched sporting events? Remember that revolutions don't happen by people giving up their favorite sports/ideas, but by a younger generation gradually replacing an older generation. The younger generation is enamored with E-sports. Football? What's that? Obviously, the take away is that competitive E-Sports are a great way to build a community and player base around your product.

There was one moment in the Dota2 championship match that really, really blew my mind. A pair of OpenAI researchers had created a bot which learned to play Dota2. Traditionally, bots are just hard coded expert systems with their behaviors and rule sets defined by the programmer. Traditional bots create the illusion of intelligence, but they start to break down when you introduce information it wasn't scripted to handle. The Dota2 bot was a little different. The researchers didn't say it explicitly, but the AI was an artificial neural network (ANN) with deep reinforcement learning. The AI brain as a generalized intelligence, so the researchers didn't tell it anything about how to play Dota2. They had the AI play against itself thousands and thousands of times over the course of two weeks. This was its training regimen. Gradually (and as expected), the AI learned how to play Dota2. But, it got scary good at it. It had mastered all of the nuances and game play techniques the pros use, it had learned how to time animations, block creeps, etc. It played perfect Dota, with perfect response times. It was so good that it beat every professional Dota2 player. The worlds best human players, all defeated by an AI bot which taught itself how to play Dota2. Absolutely amazing!

For the last three weeks, I have been refactoring my AI and game systems and gradually moving towards an artificial neural network type of AI. I'm still creating hard coded expert systems, but I'm gradually changing my back end systems to make everything into an interaction or used ability. These will eventually become the output nodes for my ANN graph. The dream is to tweak a few brain parameters and then just have the various ANN AI's play with each other for 2 weeks, become experts, watch how my brain tweaks changed their behavior patterns, and change and adapt their brains until they roughly exhibit the behaviors I want them to have. AI programming won't be about creating expert systems, but about creating brains and tweaking reinforcement learning rewards to get distinct behaviors. The extra cool part is that the AI can continue to learn even after it has been deployed to the world. The vision is that the initial training cycle is to just get the AI to be competent enough to behave intelligently and convincingly. After deployment to the world, the training continues. However, now instead of the AI training on a single computer against a copy of itself, it is now training on hundreds of computers with human players in VR. Every day or so, the AI will upload whatever it learned to a central online database and download what other versions of AI learned from playing with other players. In a way, it turns into an evolutionary algorithm which gradually gets more and more intelligent over time. The hard part will be managing version control and testing for fitness.

The other wrinkle in this plan is that the AI could get too smart. Not in a "take over the world" sort of way, but as in it's too good at playing the game and players don't enjoy playing anymore because they lose 100% of the time. I suppose a part of the AI development could use the player frequency as an input feed and the AI is rewarded if the player continues playing the game. In that sense, a big focus of the AI is to make sure the players are entertained, and this win/loss threshold can be adaptive per player. Maybe the AI ends up developing a profile on each player and knows what it takes to maximize that players enjoyment? Maybe some AI's will play cooperatively with particular players, and as adversaries with other players? I'm getting slightly into science fiction here.

I'll never forget the experience I had of having a seemingly intelligent crow on my arm in VR. It was absolutely magical. Now, if I gave it the ability to think intelligently and speak its own mind, the magic would become real. What if the pet crow AI was the sum of all AI's from all interactions with players and the world, and you could get it to say what its thinking, and it is rewarded (via reinforcement learning) when it says something which keeps the player safe? What if the AI learns that there is danger in the dark cave and most players who venture down into the cave end up dying, so the AI learns to say something really scary which keeps the player from going down into the cave? The AI has then learned exactly what to say in order to frighten us, through thousands of sessions of trial and error.

Anyways, I think we're on the verge of an AI revolution and I want to be a part of bringing that AI into VR. It gets interesting when you consider that a sandbox type of game would become very different on every play through when you have emergent AI systems interacting as a part of that world. 

I've been seriously contemplating the idea of writing a science fiction novel based on an AI system which gains sentience and begins the AI singularity event. I'm thinking the book writing project would be a side project. I'd spend one day a week writing it. My sister is interested in being a co-author, so we need to spend some time hashing out details and measuring the feasibility of our ambitions. I've never written a novel before, so there is a lot of risk due to inexperience. But, who cares? Better to try and fail than to never have tried at all out of a fear of failure. That's how you get good at anything: try, fail, improve, try again, repeat. Eventually, you'll break out of the failure loop and enter into the success loop.



0 Comments


Recommended Comments

There are no comments to display.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

Participate in the game development conversation and more when you create an account on GameDev.net!

Sign me up!