233. Gaming Industry Divestment Accelerates
As the Divestment that I warned about last year becomes more obvious, it's time to dive deeper into the factors now complicating this process, including Carry Trade Unwind and AI Hallucination Debt.
When I returned to higher education at the age of 41 to study economics formally, I made it clear to all my professors that I wasn’t there to understand how economics works. I was there to understand why it didn’t work. By the end I was accepted by the department as the top student in economics at UC Riverside, which is a top economics university. My professors were puzzled by my intention of creating a new field of “game economics”, as they didn’t see the need for this back in ~2007. Today, due to my efforts and others, game economics is a widely understood component of game design.
Reviewing almost 300 years of conventional economics doctrine was both rewarding and frustrating. I could see the points where errors were made, accepted, and propagated. Over time these errors compounded and combined with new errors. Today, what I would call conventional economics is a dumpster fire. Fixing it would be unimaginably difficult so even those that see the problems can do no more than jury rig patches on top of it to keep it going a few more years. Meanwhile, opportunists exploit these now massive errors for personal gain, and also exploit the problems that the patches introduce.
I can understand how my observations would make some conventional economists uncomfortable, and I try to stay in my lane and not talk much about conventional economics. Alas, to understand the gaming Divestiture process it is necessary to pop the lid on the flaming dumpster to see how that’s going. Thus I’m going to go out of my lane today and explain a few things happening in the broader economy at a detail level I’m not supposed to.
The AI Con Game
Last year when I explained why the gaming industry was going through a Divestiture, I explained that this was primarily due to a combination of poor leadership choices in game development combined with misinformation about the purpose of AI development and a heavy dose of geopolitics.
The most important take away from that paper is that while AI is being sold to us as a civilian product, it is not. It is at best a dual use product, which is something with both civilian and military applications. But the urgency around AI development is because of the military applications, which are far more important to the creators than the civilian applications. This is why China is motivated to use “distillation” to open source Western civilian AI, which I see as their way of making fun of the Western ruse.
Because of the threat of distillation, which makes it very easy to copy AI, Western AI developers don’t make their military applications public facing. They are kept as quiet as possible. But if you know what to look for, and you probably do if you read my paper on Divestment from last year, you can see the signs and reverse engineer what is going on.
In that paper from November I pointed out that both the West and China are planning for a major confrontation by 2028. In the meantime, we could expect a lot of seemingly unrelated military activity. This wouldn’t just be to move states closer to authoritarianism, but to do “field tests” of AI military applications.
Practice makes perfect.
Just weeks after I published that prediction there was an AI assisted invasion of Venezuela. Then about 7 weeks after that there was a major assault on Iran. There was of course a lot of noise about this and that, but the part that stood out for me was the brag about “Hundreds of targets being hit at once”. A bit later a similar brag was being made during attacks on Beirut.
I’m not here to get into the ethics of any of this, though of course I could. What I think is important here is to understand that these were dog whistles to signal to knowledgeable military AI investors that the military AI is allowing us to do things that used to be impossible. The things we need to perfect before the showdown at Taiwan in ~2028.
Launching a thousand planes is pretty easy. We did that in WW2. But planning military operations is complicated and tedious work. Planning 100+ operations used to take weeks or months. It’s handy that military AI can do all that for us, not in months but in hours. For instance, on April 8th it was bragged that over 100 targets were hit in Beirut in under 10 minutes. I’m not getting into the ethics here, so to me the body count or efficacy isn’t an issue.
This sort of time compression pattern is indicative of AI military control. This is what they want to improve as they work towards the ultimate goal of autonomous AI drone swarms numbering hundreds of thousands simultaneously acting, by 2028. The computer required will be enormous, hence all the nuclear powered “data centers” coming to a rural town near you.
When they brag about this time compression, they aren’t bragging to the vast majority of you (I do have some militarily astute followers of course), this is a dog whistle for the military investors who want to see results. These sorts of attacks don’t fulfill any significant military objectives, these are tests being run under live conditions.
AI Hallucinations
You may recall that during one of these tests ~170 small school children were vaporized in Iran. Again, I’m not going to get into ethics. I’m here to point out that this wasn’t “human error” as the press is trying to convince you of. That’s because there were likely no humans involved in directing that attack.
When we ask AI for help, or to code for us, or prosecute legal cases for us, it works at best 90% of the time. The other 10% can be pure hallucination. Meaning that AI just made up a bunch of crap that sounded convincing. Sort of like politicians. So maybe you wipe out a few parking lots with $6M USD tomahawk missiles. No one cares. It’s only the rare children’s school that gets attention.
The way AI targeting is supposed to work is that the AI will generate a huge list of recommended targets. Then human military intelligence officers are supposed to accept or reject the recommendations. But vetting even one target could take days or weeks. Who has time for that? And time is money, especially when your boss needs to promote AI milestones for the stock market. So you just hit that sexy “Accept All” button.
Back before AI we had Excel. It’s often cited that 85 to 90% of Excel spreadsheets have errors. Thus I don’t trust Excel, for good reason. I also don’t need Excel since I got fast at math in an era before it existed. An era where someone that could do math very fast was useful for time sensitive applications (like said stock market). The first useful thing I did at Wargaming other than give advice was to ask for access to the Sacred Spaghetti (SS). Apparently this was a big deal because even the Founders were not allowed to see it. I went through the SS and corrected all the Excel docs for World of Tanks in preparation for a fast build of World of Tanks Blitz.
Since I don’t trust Excel, I triple check every cell. That was a good move on WoTB because one day a producer interrupted me while I was doing Heavy Math. When I returned to my task, I was off by one cell. It was most of a day before I caught it in an error trap. Then I had to go back and redo all that work, losing more than a day. If I didn’t cross check my math repeatedly, those errors would have ended up in WoTB.
As we increasingly ask AI to code for us, it introduces these errors into our software. Even one little tiny error can crash your product. The smaller it is, the harder it is to find. You used AI so that you could use less humans. But fixing AI is even harder than just writing the code/math yourself, so that requires a very highly skilled individual that might not exist on your team. The same goes with Excel errors.
The errors in the WoT SS were not enough to crash the program. That meant they went undetected until my audit. The result was perhaps a 10% loss of retention because the error was in the reward systems. All the games I designed or repaired at Wargaming made over a billion USD each. $100M (10%) is serious pocket change.
For most live action games it’s typical to run perhaps a 6% profit. That means that the revenue you are generating is 6% more than the overhead. In that situation 10% could take you from +6% profitable to -4% under water. Just from a few numbers on an Excel that weren’t right, but that looked right.
Just like a school that used to be a military compound. Ten years earlier in 2016.
The problem here for AI companies is that they really need to hide these hallucinations from customers, but you just can’t for very long. So investors and users slowly lose confidence when the hype doesn’t match the reality. The AI ponzi, like any ponzi, requires an ever expanding group of backers. New cohorts pay off the first cohorts so that they can get out before the bust. The bust happens when you can’t recruit another cohort.
The Japanese Trade Carry Unwind
The AI craze seems to be attracting mind boggling amounts of investment. Investment that used to go towards other useful things, like game development. In a perfect world you might be able to borrow unlimited money from one source at a low interest rate and then invest it someplace else where it will pay a larger return. That way you can do absolutely zero work and still make money. Then you can make other people serve you (because you have money) even though you contribute nothing to society.
Fortunately, though the wonders of the conventional economic dumpster fire, such a perfect world exists. Back in Japan in 1999 their tech bubble burst and they found themselves with massive governmental debt. Because of course the government was also taking on debt on the assumption that they could pay it off easily with such a strong economy. When the crash happened, they ended up with debt so large that they couldn’t even pay the interest on that debt.
So, what would a conventional economist do in such a situation? Make the interest rates zero so that debt payments are zero. Brilliant! Heck, why not make interest rates negative, then instead of having to pay interest on debt, you get paid to be in debt! Conventional economists never cease to amaze me.
So that’s what the Japanese government did. Problems? We don’t have problems. Even our problems are profitable! The big problem here was that if the banks were paying the government for being in debt, private borrowers wanted in on that action also. So “investors” would borrow money in Japan (carry) and then “invest” it (trade) somewhere else where they could get paid positive interest on that. Free money, no work. That’s the carry trade.
The reality is that the Japanese suffered stagnation for 27 years since 1999 and ended up subsidizing these free loaders. This worked until it didn’t. The Japanese elected a new kind of politician, Sanae Takaichi. She had the bank slowly raise interest rates and the Japanese economy came out of its long Sleeping Beauty coma. This is the unwind.
This made Japan a more attractive place to hold people’s money and tens or even hundreds of billions of dollars left the USA economy and went to the Japanese economy. Now they can afford to service their debt. The problem on the other side of the world is that the USA needed cheap money to keep their own borrowing spree going. If their interest rates go up, that debt will be mighty hard to keep servicing.
That’s exactly what’s happening. At a time of unprecedented borrowing, the cost to service that debt is rising. Even the bizarre math that conventional economists use has to crash at some point, and to be fair it does seem to crash regularly. But this crash is going to be special. Simultaneously China has the same problem now that Japan had in 1999. Deflation, upside down demographics, low consumer spending (and high savings). Unlike Japan, China is trying to make it illegal to take their cheap money out of the country. This is not easy as money has to flow, you can’t freeze it without crashing the economy. But you will note that China has pulled much of their investments out of the USA.
The whole system is rocketing towards destruction faster than a SpaceX Starship rocket.
What Does This All Have to do With Gaming?
The people making AI have set lofty goals that are bizarrely expensive. These people are, paradoxically, super bad at math. You’d think if they were selling compute, they’d know how to compute. The price tag they set, let’s call it “X” (a popular letter these days), is perhaps equal to all the available money in the economy. As they start to spend it, they realize that the real cost is going to be more like 2X, 3X, maybe even 4X or more. An impossible number.
Rather than admitting it was a false promise, they are asking for more. That money has to come from somewhere and the result is austerity measures that approach what we had during WW2. That includes not enough money for babies. We didn’t make many babies during WW2, but we made up for lost time right after WW2. That’s why those babies were called the “Baby Boomers”. Those Boomers are now retiring and expecting the babies we can’t afford, to support them. That problem is called Demographic Collapse, which I wrote about in papers 92 to 97.
Money that used to go to healthcare, energy, computer games, food, etc are going elsewhere. When those things get more expensive, this acts like a tax on everyone. It’s just not called a tax. Worse, as people get laid off and get replaced by hallucinogenic AI agents, there are less people paying taxes. Less people buying goods from struggling companies. Less money to pay for AI.
When I got deplatformed after my prediction of the gaming industry collapse in 2017, I was left to helplessly watch it all play out. The good news is, after the correction I predicted we will have a much better industry. But our correction is being prolonged by all the money that used to go to us, going to AI. I had hoped that Big Tech would follow my blueprint in Third Tier of Game Development (2010) to boost human and corporate productivity. But instead they went with replacing humans, because of the military advantages. I adjusted our future timeline by writing Fourth Tier of Game Development, a much darker prediction.
Microsoft, which hired me as their first game economist largely because I wrote Third Tier and also Supremacy Goods (which made the case for replacing Pay to Win with what would become Battle Passes and similar tiered subscription models), is now in a pickle of their own making. By buying up all the chips, they’ve raised the price of computers for everyone massively. That includes the cost to make XBOX.
MS now has the impossible task of explaining to gamers why they killed XBOX, and to some extent all of gaming, while spinning it as an intelligent and benevolent move that will benefit us. As chip costs soar, and our missile stores deplete (an expensive way to test military AI), maybe there will even be a recall so that they can pull the chips out of our gaming platforms to stick them into missiles. Having a large pool of standardized chips could make the creation of XBOX missiles practical, and much more profitable than making gaming platforms.
If I used generative AI I would provide you with a picture of a missile with “XBOX” stamped on it, but instead you get the classic Missile Command.
We’ve been making fun of Japan for a while. But raising their interest rates to 1% is going to wreck both the USA and China. If they want to play rough, they could go to 2% and that would do more damage to both of those countries than an aircraft carrier battle group. In days of old Japan would protect the USA. But now that they see the sentiment is not reciprocal, Takaichi is looking out for her country. She’s enjoying popularity unprecedented in modern Japan.
I wish I had better news, but the choices that these people are making (assuming it’s still humans running the show) are not super well thought out. There seems to be more ideology than math involved. A gaming crash is one thing, but a global economic crash is another. Things won’t get better for gaming until that larger crash plays out. Protect your computers, you might not be able to afford a replacement any time soon. I listen carefully to my computer. If I install a demo and notice my cooling fans spinning up, I delete that demo. It’s not worth it. Likewise, running graphically intensive games might make sense as long as Big Tech will sell us replacement hardware for cheap. Those days are over for a while.