By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
CHAPTERS
- 0:00 – 3:26
Big Ideas 2024 Intro and the AI-First Games Pitch
Steph Smith introduces a16z's 2024 big ideas list. Jonathan Lai then lays out his idea: AI-first games that never end, built around generative agents, personalized narratives, runtime-generated worlds, and AI copilots.
- •a16z released a list of over 40 big ideas for 2024
- •Lai expects the first cohort of AI-first games in 2024, built on large language models
- •The largest long-term opportunity is reinventing the nature of games, not only making creators more efficient
- •Examples include lifelike NPC companions, personalized character and narrative systems, and worlds generated at runtime
- •Every game could be redesigned around an AI copilot: good alone, great with AI, best with friends
- 3:26 – 5:36
Two Waves of AI in Games
Lai describes the first wave, AI tools that help studios build the same games faster, which is already happening. The second wave is AI-native games that are uniquely enabled by AI, and it is still wide open.
- •A survey of 230 game studios found 87% already use generative AI tools
- •Studios named tools such as ChatGPT, Midjourney, Eleven Labs, Adobe Firefly and GitHub Copilot
- •The majority of studios rated the impact as transformational
- •The second wave includes real-time dynamic worlds, lifelike companions and never-ending choose-your-own-adventure games
- •The playing field is wide open and could create genres that captivate hundreds of millions of people
- 5:36 – 7:26
Good Alone, Great with AI, Best with Friends
Lai explains the mantra by describing how big games use AI bots mainly for onboarding and practice. Generative agents could move AI from a warm-up into the core game, for example as a customizable AI companion in every game.
- •Games like Fortnite, Call of Duty and League of Legends use AI mainly for onboarding and practice
- •Generative agents could elevate AI from warm-up to core gameplay
- •An AI companion could play on your team or against you, comment on the world and learn your preferences
- •Such a companion would not replace human play but beats playing alone
- •Modes can be mixed and matched depending on the game and moment
- 7:26 – 11:57
Do Players Want AI Companions and Hyper Personalization?
Steph Smith asks what gives Lai confidence that players want hyper personalized, AI-driven games, given the communal nature of games. Lai points to existing consumer behavior: beloved scripted companions like Cortana and Ellie, avatar customization, skins, and Dungeons & Dragons.
- •The games industry has embraced AI for decades, so the consumer behavior already exists
- •Cortana in Halo and Ellie in The Last of Us are pre-scripted companions players grew deeply attached to
- •Baldur's Gate 3 character builders and Fortnite and League of Legends skins show appetite for personalization
- •Dungeons & Dragons is a personalized adventure told by a human narrator
- •Generative AI makes it easier to do what players already want rather than inventing new behavior
- 11:57 – 16:08
New Entrants vs. Incumbents
Asked who wins, Lai says incumbents will benefit from wave-one tools but he is skeptical they will lead on AI-native games. He cites IP-driven business models, the risk of early AI games looking silly, and the mobile precedent where new companies won.
- •Large developers are already adopting wave-one tools like ChatGPT, Midjourney and Adobe Firefly
- •Their businesses revolve around sequels to key franchises, which doesn't support much risk-taking
- •Early AI-native games may look silly to traditional developers, as Candy Crush did
- •Mobile winners such as Supercell, King and Zynga were new companies, and each paradigm brings new players and expands the pie
- •Character AI is an engaging experience most users wouldn't call a game, and it is likely off incumbents' radar
- 16:08 – 19:01
Business Models and Inference Costs
Lai explains that free-to-play and premium models would likely be unprofitable for games running LLMs in the cloud because of inference costs. He guesses subscriptions will become the default, as with Midjourney and Character AI.
- •Free-to-play has roughly 90–95% of users playing for free while 5–10% monetize
- •Premium games take payment up front and nothing after
- •Inference costs could make either model unprofitable for AI games
- •Subscriptions may become the default so power users pay for their own inference
- •Subscriptions have grown in games through Xbox Game Pass, Apple Arcade and Fortnite's Battle Pass
- 19:01 – 21:13
Designing Around Time vs. Money Trade-offs
Smith asks whether tying inference to cost could discourage deep engagement. Lai says games must design around it, noting the time-versus-money trade-off already exists in free-to-play, as with FarmVille's energy meter, and suggests a virtual currency could approximate inference cost.
- •Smith raises a possible negative feedback loop when play time is linked to inference cost
- •Lai says the same time-versus-money trade-off exists in free-to-play games
- •FarmVille-style energy meters gated play and gave rise to appointment gaming
- •Players could spend to refill energy to keep going
- •An in-game virtual currency could approximate inference cost for AI games
- 21:13 – 23:41
Advice for Game Builders in 2024
Lai suggests builders take a favorite genre and reimagine it from first principles with generative AI in the core gameplay loop, then prototype it. He notes that iconic games came from small teams and that new paradigms favor new teams, with the potential to disrupt a $200 billion market.
- •Pick a favorite genre such as sports, MMO, chess, poker or D&D and reimagine it with generative AI in the core loop
- •Hack together a prototype or team up with friends, then show it to a16z
- •Dota, PUBG and Counter-Strike came from very small teams
- •New paradigm shifts favor new teams with fresh perspectives
- •The opportunity is disrupting a $200 billion market in the fastest-growing category of media
