2024 Big Ideas: AI-First Games That Never End with Jonathan Lai

AI-first games are on the horizon, driven by creators utilizing large models to revolutionize game systems.
a16z General Partner Jonathan Lai shares his Big Idea in Tech for 2024, envisioning a world where never-ending games with personalized characters captivate players. Game worlds will no longer be static; instead, they'll be dynamically generated at runtime using neural networks.
Topics Covered:
00:00 - Big Ideas in Tech 2024
01:39 - Big Idea: AI-First Games That Never End
03:15 - Generative AI in gaming and AI native games
05:15 - Introducing generative agents as game companions
07:26 - Does the gaming community want hyper personalization?
11:04 - New entrants vs the incumbents
16:08 - How business models and cost structures are evolving
21:03 - Advice for game builders in 2024
Find Jonathan on Twitter: / tocelot
View all 40+ big ideas: a16z.com/bigideas2024
Stay Updated:
Find a16z on Twitter: / a16z
Find a16z on LinkedIn: / a16z
Subscribe on your favorite podcast app: a16z.simplecast.com/
Follow our host: / stephsmithio
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.

[@a16z]

Topics Covered: 00:00 - Big Ideas in Tech 2024 01:39 - Big Idea: AI-First Games That Never End 03:15 - Generative AI in gaming and AI native games 05:15 - Introducing generative agents as game companions 07:26 - Does the gaming community want hyper personalization? 11:04 - New entrants vs the incumbents 16:08 - How business models and cost structures are evolving 21:03 - Advice for game builders in 2024

[@real23lions]

Interesting take. An AI guide that customise according to player. I can see this happening where: 1. Remove or lessen annoying guide like in God of War for veteran players. Adjust according to beginner or veteran. 2. Make games difficulty adjustable. Don't just add more health and damage to bosses. There are other ways to make bosses more difficult but fun. Some people like Soulslike games with higher skill ceiling. Make Hogwarts game with Soulslike difficulty. Cool ideas.

[@themore-you-know]

Bro, you don't need AI to remove tutorials from a game based off skill level. That's a barebone condition. "if score is above VALUE, then disable tutorials". Same for game difficulty. It has nothing to do with AI. I'm programming a game myself, with everything you mentioned, and AI would be the worst possible way to implement what you suggest. Made even worst by the fact that the alternative is quite easy to implement, if considered at the beginning of the design process.

[@real23lions]

@@themore-you-know fair enough. That just means those game studios are just bad game designers then.

[@David.Alberg]

This will make gaming the No.1 activity anyone will ro in their freetime

[@themore-you-know]

I have a lot of doubts. There is no way AI-generated content can function beyond identity-ridden games like Fortnite. Imagine the following: Darth Vader starts talking about his morning routine, coffee preference, etc. - Yoda starts talking about his morning routine, coffee preference, etc. - Diablo starts talking about his morning routine, soul roasting, etc. The sheer force of 99% of IP is built around a minority of content distilling an entire character or franchise. - "I'll be back" - "May the force be with you" - etc. AI generation destroys all of that, by allowing every aspect of every game to add everything, at which point every game has every aspect and looks like every other game. Only games built to be ingest every other IP (such as Fortnite) will benefit. Same goes for gameplay, in which most games cannot support adding every other types of gameplay. - otherwise, we may as well add mandatory Tennis rounds in between every 5 moves in Chess.

[@conformist]

idk, as LLMs get better, and how we use them improve, i think it’ll be able to solve the problems which you mention

[@themore-you-know]

@@conformist , them "getting better" is not a quantifiable property. Please explain what quantifiable property will be improve by better LLMs? Because as I see it, the problem lies in: if an amazing monolithical LLM is hooked up to 200 different games, all 200 games risk looking the same. Sure, the characters may not all talk about their morning routines and coffee, but it will be some form of other similarity commodifying the game as a result of the underlying feature (AI generation) being present in all 200 games. In essence, the quantifiable properties that I judge a game by and allude to are: - diversity of distilled speech (both among in-game interactions, and between games) - diversity of gameplay experience (between games, but perhaps in-game too). "getting better LLMs" will not solve any of the issues I brought up.

[@conformist]

@@themore-you-know you can today already make gpt-4 talk differently if you prompt it differently. there are similarities in text still yes, but it changes substantially with different prompts. exactly how LLMs get better is a rather hard to answer question, very hard to predict how they qualitatively differ.

[@praveshsingh3349]

BTC VIP GROUP is real or fake