🗞️ AI superintelligence is coming: How this impacts game development

Get ready, former OpenAI superalignment researcher Leopold Aschenbrenner suggests "we are in for a wild ride"

Jun 22, 2024

Hey Gamemakers,

As mentioned in our last newsletter, we are changing things up. News is now separate from Marketwire and will no longer be short summaries of top 10 news. Instead, we will comment on the past week's top 2-4 news items. With a focus on the news we believe will have the most significant impact on game development and with a bit of commentary on why.

Also, I’m traveling this week, so publishing a little late. In the future, look for this segment on Fridays!

A controversial 165-page paper on the future of AI titled “Situational Awareness” written by former OpenAI Superalignment team head Leopold Aschenbrenner outlines massive potential impacts for future society, and we highlight potential implications for game development.

Plus:

The upcoming launch of Dark and Darker Mobile (published by Krafton) will have major challenges and significant implications if it is successful.
Executive compensation comes under scrutiny as EA’s Andrew Wilson earns $25M despite company layoffs.

🤖 Situational Awareness warns of a future dramatically changed by superintelligence

Leopold Aschenbrenner, formerly with the Superalignment team at OpenAI, recently left (or was fired from) OpenAI. Ilya Sutskever (Chief Scientist) and Jan Leike (Co-lead of the Superalignment team) have also left OpenAi.

In a 165-page treatise called “Situational Awareness,” Leopold discusses the potential implications of AI for our future and how technology and its impact on the world could be unlike anything we’ve ever seen.

According to Leopold: “Everyone is now talking about AI, but few have the faintest glimmer of what is about to hit them.”

Before long, the world will wake up. But right now, there are perhaps a few hundred people, most of them in San Francisco and the AI labs, that have situational awareness. Through whatever peculiar forces of fate, I have found myself amongst them. A few years ago, these people were derided as crazy—but they trusted the trendlines, which allowed them to correctly predict the AI advances of the past few years.
…
If they are seeing the future even close to correctly, we are in for a wild ride.
Let me tell you what we see.

Knowing most game developers and busy execs and product managers don’t read long essays, let me give you a good TLDR version of the paper from this YouTube video:

Note that AGI and the concept of superintelligence are controversial. Some AI experts, like Yann LeCun (from Meta), disagree with the conclusions drawn from Leopold and others.

But for now, let’s consider Leopold's implications. I’ll simplify the setup (but you should read the paper yourself) and explain what I believe to be the potential implications for game development.

The Setup

Leopold argues that AI advancements are unlike anything we’ve ever seen. Further, we will reach a specific point where AI can improve itself.

“It is strikingly plausible that by 2027, models will be able to do the work of an AI researcher/engineer.”

What happens when AI can do the work of an AI researcher/engineer that can help it improve itself? It can then deploy millions, billions, or more of these researchers and thereby dramatically increase the rate at which AI and technology, in general, can be improved.

GPT-2 to GPT-4—from models that were impressive for sometimes managing to string together a few coherent sentences, to models that ace high-school exams—was not a one-time gain. We are racing through the OOMs extremely rapidly, and the numbers indicate we should expect another ~100,000x effective compute scaleup—resulting in another GPT-2-to-GPT-4-sized qualitative jump—over four years. Moreover, and critically, that doesn’t just mean a better chatbot; picking the many obvious low-hanging fruit on “unhobbling” gains should take us from chatbots to agents, from a tool to something that looks more like drop-in remote worker replacements.

As a game development-focused newsletter, let’s get to the point: how could this potentially impact game development?

Regardless of whether you believe in all of the implications of Leopold’s paper, a few things we can conclude:

AI's capabilities to impact generative content across the board (art, engineering, 3D models, etc.), including potentially making games whole scale, will dramatically improve by many OOMs (orders of magnitude).
Many jobs will be displaced.
There will be new winners and losers in a big way; the winners will be those who embrace AI, and the losers who ignore AI will be left behind.

To my readers, get on board now. Learn now. This is the time—right now!

Finally, let me pose an optimistic view for game industry players.

Many of Leopold's trends suggest a future abundance in reasoning, capability, innovation, and generative content. In such a world, it is reasonable to believe that humanity will have less of a role in many jobs and tasks required, leading to dramatically higher leisure time for humans.

I believe this has implications for the demand and supply of entertainment content.

Content Demand

What happens in a world where humans have 50%+ more free time?

I believe this will lead to the value of entertainment increasing significantly.

Hence, games and other entertainment creators should become much more valuable.

We may find ourselves demanding an order of magnitude or more entertainment content.

In a “situational awareness” world where AI and robots do most of the work humans require and do, many jobs suitable for humans will disappear, but I believe entertainment work will continue to hold and likely increase in value.

Content Supply

Even further, game development should become dramatically easier. At some point, AI may completely eliminate the need for humans to make games. However, in the interim (and potentially always), humans should play a role in game development, but the requirement for large development requirements will disappear.

We will increasingly move towards a “small vector” development model from “large vector” development. By vector, I mean time, cost, people, resources required, etc.

This should dramatically increase the number of people who can become game creators. In this case, creativity may become more important than any other skill.

By dramatically reducing the barrier to creation, the number of game creators could increase by 2 or more orders of magnitude.

The intelligence explosion and the immediate post-superintelligence period will be one of the most volatile, tense, dangerous, and wildest periods ever in human history.

Get ready, everybody. We’re gonna be in for a wild upcoming decade.

🕶️ Dark and Darker mobile is likely to face major challenges but has high potential

To date, extraction gameplay on mobile has faced a very tepid response, and almost every extraction game launched on mobile has failed.

For example, almost every extraction shooter game for mobile has already shut down or is considered a failure.

The two most currently prominent games left in the market that have achieved some commercial scale are Lost Light (not looking good) and Arena Breakout (limited commercial scale).

Before June 6, Lost Light had only generated a few thousand USD in net revenue daily. Since then, there has been a big push to $100K daily (according to AppMagic estimates), but it’s starting to drop.

Arena Breakout has more scale, regularly generating about $150K —$300K daily net revenue (AppMagic estimates). However, I suspect most of the revenue scale for this game is due to Tencent’s distribution power, and most of the revenue is from China.

While most games have focused on extraction shooters for mobile, there is no reason other genres of games could not successfully leverage the extraction model.

Many of you are likely aware of the game Dark and Darker by the Korean game studio Ironmace. This PC game has stumbled upon a great gameplay model focused on extraction. It has gained fairly significant traction on PC and regularly sits in the top 25 CCU Steam games with 40K+ CCU.

Krafton recently announced the launch of Dark and Darker Mobile and, even more recently, a global beta test in August. Currently, the game seems to have very positive player reviews and has attracted a core audience of believers.

This is a great move for Krafton and the original developer, Iron Mace. Having said that, as someone working on extraction for mobile, I believe a few major issues will need to be addressed for the game to achieve greater commercial scale.

Key Challenges:

#1. Tuning Difficulty

Many of the failed mobile extraction shooter games tried to take “Escape from Tarkov” and do a direct translation to mobile. This was a mistake.

Historically, many of the big hits on mobile have come from simplifying and casualizing mobile experiences. Hence, I believe a big detractor to success on mobile has been delivering the same hardcore experience for players on mobile.

Many mobile game players are not looking for quite the same experience as PC or console players are looking for.

For example, many mobile players aren’t looking for hardcore gameplay, and competitive integrity is not as critical. They just want a fun gameplay experience in which they can see progress. Hence, the difficulty level of Dark and Darker Mobile may be a major issue for many mobile game players.

#2. Platform Choice in the West

In Western markets, almost every home has a PC, and many high-skill games are preferentially played on PCs instead of mobile devices. When we look at MOBAs, for example, we see a dramatic drop in the commercial success of these games in Western markets relative to Asian markets or those in which mobile devices are primary and many homes may not have PCs.

Without differentiated gameplay, many Western players will likely choose to play Dark and Darker on PC instead of mobile.

#3. Limited Progression

For mobile, I firmly believe and have proselytized that progression and compulsion are dramatically more important than fun or competitive integrity.

For mobile, Dark and Darker’s progression depth is too shallow. It’s good because there are upgradeable skills and some progression. However, IMHO, there won’t be enough to support a truly F2P model with long-term engagement and progression. The game will bank on “fun” instead of progression/compulsion.

Anyway, I’ll leave it with just these few thoughts. I also have a lot more nuanced opinions regarding challenges, which I’d prefer not to publish. To be clear, these are just the humble opinions of another game developer.

If this game is as successful on mobile as I hope, we should see many more extraction shooters take another crack at this model for mobile. Dark and Darker does a few simple things extremely well that other extraction shooters have not done. Further, it will show a viable model that will likely get emulated significantly.

Just to conclude, I'm a fan of the game. I wish them the best and am excited about the game's awesome upside potential despite some challenges adapting more appropriately to mobile.

💰 The War Between Employee vs. Executive Compensation

How valuable are executives, really?

I would argue that the right leadership can make all the difference in a game company's outcome. However, in many cases, leadership does not matter as much.

How much compensation do CEOs and other executives deserve compared to a typical employee?

In the case of EA, Andrew Wilson is worth $25M for the year, and EA’s senior leadership team was paid $60M in executive compensation for FY2024.

This compensation comes in a year when EA laid off 5% of its global workforce and closed studios and some beloved projects. Understandably, the company faced much criticism and public backlash.

“Imagine what an INVESTMENT such money would've been on a next heavy-hitting, trail-blazing game."

-@OwlBasketTeam on X

Especially in the upcoming age of AI, we should expect this discrepancy between executive leadership and line workers will only get worse, not better.

Andrew Wilson has already spoken about the potential for AI to impact 60% of development and reduce development costs by 30%.

We’re in the era of generative AI which is the most exciting yet by a fairly wide margin and something that we’re embracing deeply. We think about it in three core vectors: efficiency, expansion, and transformation.
- Andrew Wilson

All of the changes coming to our industry and the forces at work in our society today suggest a growing disparity between the rich and the poor and between employees and executives.

What’s fair? And how will our society and game companies react?

Ray Dalio has suggested that populism emerges in these times when wealth disparity begins to get out of whack. You can see the concentration of wealth to the top 0.1% from Dalio’s chart below:

This is nothing new; we are experiencing a cyclical trend that will inevitably lead to conflict.

What’s fair?

GameMakers

Discussion about this post