Br Bringing inging Gaming, ing, VR, and nd AR to L to Life fe - - PowerPoint PPT Presentation

▶

Sep 07, 2023 108 likes •384 views

Br Bringing inging Gaming, ing, VR, and nd AR to L to Life fe Wi With th D Deep L Learn rning Danny Lange Vice President of AI & ML Unity Technologies What Was Before Machine Learning? Clockwork Universe FEEDBACK

SLIDE 1

Br Bringing inging Gaming, ing, VR, and nd AR to L to Life fe Wi With th D Deep L Learn rning

Danny Lange Vice President of AI & ML Unity Technologies

SLIDE 2

What Was Before Machine Learning?

PROGRAM “ALL-KNOWING PROGRAMMER” DATA RESULTS FEEDBACK Clockwork Universe

SLIDE 3

What is Machine Learning

MODEL LEARNER DATA PREDICTIONS HISTORIC DATA Indeterminism

SLIDE 4

OODA Loop (John Boyd)

Heisenberg's Uncertainty Principle

There is a limit on our ability to observe reality with precision.

Gödel's Incompleteness Theorem

Any model of reality is incomplete and must be continuously

refined in the face of new observations.

Second Law of Thermodynamics (Entropy) - Ludwig Boltzman

Any given system is continuously changing even as we try to maintain order

SLIDE 5

Multi-armed Bandit & Reinforcement Learning

Objective: Maximize winnings Exploration vs Exploitation

Gaining knowledge
Max payout with current knowledge

Reinforcement Learning

Actions
Rewards

X% Y% Z%

AG AGENT EN ENVIRONMEN ENT ACTION STATE & REWARD

SLIDE 6

DeepMind Playing Atari

SLIDE 7

SLIDE 8

DeepMind Lab

Current AI Research Landscape

OpenAI Gym/Universe

SLIDE 9

About Unity

Leading global game industry platform

16 Billion Downloads in 2016 – up 31%
2.6 Billion Unique Devices
700 Million Gamers
38% of top 1000 free mobile games

SLIDE 10

The Unity Ecosystem

SLIDE 11

Bringing Reinforcement Learning to Unity

SLIDE 12

How did the Chicken Cross the Road?

Actions Rewards

Fatal penalty (being hit by a car)
Positive reward (collecting gift packet)

Exploration Exploitation

SLIDE 13

SLIDE 14

Navigating by reading maps

SLIDE 15

SLIDE 16

Network Architecture

Deep Recurrent Q-Network
Dual-stream input
Train network with additional

auxiliary losses Actions Inputs Auxiliary signals

SLIDE 17

SLIDE 18

Synthetic Data

SLIDE 19

SLIDE 20

SLIDE 21

Bringing it to developers and researchers

SLIDE 22

Unity Reinforcement Learning API (Coming Soon)

Action State & Reward

SLIDE 23

SLIDE 24

SLIDE 25

Bringing it together

SLIDE 26

SLIDE 27

Danny Lange Vice President, AI & Machine Learning +1 425.463.5801 dlange@unity3d.com

@danny_lange dannylange