3D computer animation is a time-consuming and highly technical medium — to complete even a single animated scene requires numerous steps, like modeling, rigging and animating, each of which is itself a sub-discipline that can take years to master. Because of its complexity, 3D animation is generally practiced by teams of skilled specialists and is inaccessible to almost everyone else, despite decades of advances in technology and tools. With the recent development of tools that facilitate game character creation and game balance, a natural question arises: is it possible to democratize the 3D animation process so it’s accessible to everyone?
To explore this concept, we start with the observation that most forms of artistic expression have a casual mode: a classical guitarist might jam without any written music, a trained actor could ad-lib a line or two while rehearsing, and an oil painter can jot down a quick gesture drawing. What these casual modes have in common is that they allow an artist to express a complete thought quickly and intuitively without fear of making a mistake. This turns out to be essential to the creative process — when each sketch is nearly effortless, it is possible to iteratively explore the space of possibilities far more effectively.
In this post, we describe Monster Mash, an open source tool presented at SIGGRAPH Asia 2020 that allows experts and amateurs alike to create rich, expressive, deformable 3D models from scratch — and to animate them — all in a casual mode, without ever having to leave the 2D plane. With Monster Mash, the user sketches out a character, and the software automatically converts it to a soft, deformable 3D model that the user can immediately animate by grabbing parts of it and moving them around in real time. There is also an online demo, where you can try it out for yourself.
Creating a 2D Sketch The insight that makes this casual sketching approach possible is that many 3D models, particularly those of organic forms, can be described by an ordered set of overlapping 2D regions. This abstraction makes the complex task of 3D modeling much easier: the user creates 2D regions by drawing their outlines, then the algorithm creates a 3D model by stitching the regions together and inflating them. The result is a simple and intuitive user interface for sketching 3D figures.
For example, suppose the user wants to create a 3D model of an elephant. The first step is to draw the body as a closed stroke (a). Then the user adds strokes to depict other body parts such as legs (b). Drawing those additional strokes as open curves provides a hint to the system that they are meant to be smoothly connected with the regions they overlap. The user can also specify that some new parts should go behind the existing ones by drawing them with the right mouse button (c), and mark other parts as symmetrical by double-clicking on them (d). The result is an ordered list of 2D regions.
Stitching and Inflation To understand how a 3D model is created from these 2D regions, let’s look more closely at one part of the elephant. First, the system identifies where the leg must be connected to the body (a) by finding the segment (red) that completes the open curve. The system cuts the body’s front surface along that segment, and then stitches the front of the leg together with the body (b). It then inflates the model into 3D by solving a modified form of Poisson’s equation to produce a surface with a rounded cross-section (c). The resulting model (d) is smooth and well-shaped, but because all of the 3D parts are rooted in the drawing plane, they may intersect each other, resulting in a somewhat odd-looking “elephant”. These intersections will be resolved by the deformation system.
Layered Deformation At this point we just have a static model — we need to give the user an easy way to pose the model, and also separate the intersecting parts somehow. Monster Mash’s layered deformation system, based on the well-known smooth deformation method as-rigid-as-possible (ARAP), solves both of these problems at once. What’s novel about our layered “ARAP-L” approach is that it combines deformation and other constraints into a single optimization framework, allowing these processes to run in parallel at interactive speed, so that the user can manipulate the model in real time.
The framework incorporates a set of layering and equality constraints, which move body parts along the z axis to prevent them from visibly intersecting each other. These constraints are applied only at the silhouettes of overlapping parts, and are dynamically updated each frame.
Meanwhile, in a separate thread of the framework, we satisfy point constraints to make the model follow user-defined control points (described in the section below) in the xy-plane. This ARAP-L method allows us to combine modeling, rigging, deformation, and animation all into a single process that is much more approachable to the non-specialist user.
Animation To pose the model, the user can create control points anywhere on the model’s surface and move them. The deformation system converges over multiple frames, which gives the model’s movement a soft and floppy quality, allowing the user to intuitively grasp its dynamic properties — an essential prerequisite for kinesthetic learning.
To create animation, the system records the user’s movements in real time. The user can animate one control point, then play back that movement while recording additional control points. In this way, the user can build up a complex action like a walk by layering animation, one body part at a time. At every stage of the animation process, the only task required of the user is to move points around in 2D, a low-risk workflow meant to encourage experimentation and play.
Conclusion We believe this new way of creating animation is intuitive and can thus help democratize the field of computer animation, encouraging novices who would normally be unable to try it on their own as well as experts who often require fast iteration under tight deadlines. Here you can see a few of the animated characters that have been created using Monster Mash. Most of these were created in a matter of minutes.
All of the code for Monster Mash is available as open source, and you can watch our presentation and read our paper from SIGGRAPH Asia 2020 to learn more. We hope this software will make creating 3D animations more broadly accessible. Try out the online demo and see for yourself!
Acknowledgements Monster Mash is the result of a collaboration between Google Research, Czech Technical University in Prague, ETH Zürich, and the University of Washington. Key contributors include Marek Dvorožňák, Daniel Sýkora, Cassidy Curtis, Brian Curless, Olga Sorkine-Hornung, and David Salesin. We are also grateful to Hélène Leroux, Neth Nom, David Murphy, Samuel Leather, Pavla Sýkorová, and Jakub Javora for participating in the early interactive sessions.
In March 2020 we introduced the Research Scholar Program, an effort focused on developing collaborations with new professors and encouraging the formation of long-term relationships with the academic community. In November we opened the inaugural call for proposals for this program, which was received with enthusiastic interest from faculty who are working on cutting edge research across many research areas in computer science, including machine learning, human-computer interaction, health research, systems and more.
Today we are pleased to announce that in this first year of the program we have granted 77 awards, which included 86 principal investigators representing 15+ countries and over 50 universities. Of the 86 award recipients, 43% identify as an historically marginalized group within technology. Please see the full list of 2021 recipients on our web page, as well as in the list below.
We offer our congratulations to this year’s recipients, and look forward to seeing what they achieve!
Natural language processing (NLP) models based on Transformers, such as BERT, RoBERTa, T5, or GPT3, are successful for a wide variety of tasks and a mainstay of modern NLP research. The versatility and robustness of Transformers are the primary drivers behind their wide-scale adoption, leading them to be easily adapted for a diverse range of sequence-based tasks — as a seq2seq model for translation, summarization, generation, and others, or as a standalone encoder for sentiment analysis, POS tagging, machine reading comprehension, etc. The key innovation in Transformers is the introduction of a self-attention mechanism, which computes similarity scores for all pairs of positions in an input sequence, and can be evaluated in parallel for each token of the input sequence, avoiding the sequential dependency of recurrent neural networks, and enabling Transformers to vastly outperform previous sequence models like LSTM.
A limitation of existing Transformer models and their derivatives, however, is that the full self-attention mechanism has computational and memory requirements that are quadratic with the input sequence length. With commonly available current hardware and model sizes, this typically limits the input sequence to roughly 512 tokens, and prevents Transformers from being directly applicable to tasks that require larger context, like question answering, document summarization or genome fragment classification. Two natural questions arise: 1) Can we achieve the empirical benefits of quadratic full Transformers using sparse models with computational and memory requirements that scale linearly with the input sequence length? 2) Is it possible to show theoretically that these linear Transformers preserve the expressivity and flexibility of the quadratic full Transformers?
We address both of these questions in a recent pair of papers. In “ETC: Encoding Long and Structured Inputs in Transformers”, presented at EMNLP 2020, we present the Extended Transformer Construction (ETC), which is a novel method for sparse attention, in which one uses structural information to limit the number of computed pairs of similarity scores. This reduces the quadratic dependency on input length to linear and yields strong empirical results in the NLP domain. Then, in “Big Bird: Transformers for Longer Sequences”, presented at NeurIPS 2020, we introduce another sparse attention method, called BigBird that extends ETC to more generic scenarios where prerequisite domain knowledge about structure present in the source data may be unavailable. Moreover, we also show that theoretically our proposed sparse attention mechanism preserves the expressivity and flexibility of the quadratic full Transformers. Our proposed methods achieve a new state of the art on challenging long-sequence tasks, including question answering, document summarization and genome fragment classification.
Attention as a Graph The attention module used in Transformer models computes similarity scores for all pairs of positions in an input sequence. It is useful to think of the attention mechanism as a directed graph, with tokens represented by nodes and the similarity score computed between a pair of tokens represented by an edge. In this view, the full attention model is a complete graph. The core idea behind our approach is to carefully design sparse graphs, such that one only computes a linear number of similarity scores.
Extended Transformer Construction (ETC) On NLP tasks that require long and structured inputs, we propose a structured sparse attention mechanism, which we call Extended Transformer Construction (ETC). To achieve structured sparsification of self attention, we developed the global-local attention mechanism. Here the input to the Transformer is split into two parts: a global input where tokens have unrestricted attention, and a long input where tokens can only attend to either the global input or to a local neighborhood. This achieves linear scaling of attention, which allows ETC to significantly scale input length.
In order to further exploit the structure of long documents, ETC combines additional ideas: representing the positional information of the tokens in a relative way, rather than using their absolute position in the sequence; using an additional training objective beyond the usual masked language model (MLM) used in models like BERT; and flexible masking of tokens to control which tokens can attend to which other tokens. For example, given a long selection of text, a global token is applied to each sentence, which connects to all tokens within the sentence, and a global token is also applied to each paragraph, which connects to all tokens within the same paragraph.
With this approach, we report state-of-the-art results in five challenging NLP datasets requiring long or structured inputs: TriviaQA, Natural Questions (NQ), HotpotQA, WikiHop, and OpenKP.
BigBird Extending the work of ETC, we propose BigBird — a sparse attention mechanism that is also linear in the number of tokens and is a generic replacement for the attention mechanism used in Transformers. In contrast to ETC, BigBird doesn’t require any prerequisite knowledge about structure present in the source data. Sparse attention in the BigBird model consists of three main parts:
In the BigBird paper, we explain why sparse attention is sufficient to approximate quadratic attention, partially explaining why ETC was successful. A crucial observation is that there is an inherent tension between how few similarity scores one computes and the flow of information between different nodes (i.e., the ability of one token to influence each other). Global tokens serve as a conduit for information flow and we prove that sparse attention mechanisms with global tokens can be as powerful as the full attention model. In particular, we show that BigBird is as expressive as the original Transformer, is computationally universal (following the work of Yun et al. and Perez et al.), and is a universal approximator of continuous functions. Furthermore, our proof suggests that the use of random graphs can further help ease the flow of information — motivating the use of the random attention component.
This design scales to much longer sequence lengths for both structured and unstructured tasks. Further scaling can be achieved by using gradient checkpointing by trading off training time for sequence length. This lets us extend our efficient sparse transformers to include generative tasks that require an encoder and a decoder, such as long document summarization, on which we achieve a new state of the art.
Moreover, the fact that BigBird is a generic replacement also allows it to be extended to new domains without pre-existing domain knowledge. In particular, we introduce a novel application of Transformer-based models where long contexts are beneficial — extracting contextual representations of genomic sequences (DNA). With longer masked language model pre-training, BigBird achieves state-of-the-art performance on downstream tasks, such as promoter-region prediction and chromatin profile prediction.
Main Implementation Idea One of the main impediments to the large scale adoption of sparse attention is the fact that sparse operations are quite inefficient in modern hardware. Behind both ETC and BigBird, one of our key innovations is to make an efficient implementation of the sparse attention mechanism. As modern hardware accelerators like GPUs and TPUs excel using coalesced memory operations, which load blocks of contiguous bytes at once, it is not efficient to have small sporadic look-ups caused by a sliding window (for local attention) or random element queries (random attention). Instead we transform the sparse local and random attention into dense tensor operations to take full advantage of modern single instruction, multiple data (SIMD) hardware.
To do this, we first “blockify” the attention mechanism to better leverage GPUs/TPUs, which are designed to operate on blocks. Then we convert the sparse attention mechanism computation into a dense tensor product through a series of simple matrix operations such as reshape, roll, and gather, as illustrated in the animation below.
Recently, “Long Range Arena: A Benchmark for Efficient Transformers“ provided a benchmark of six tasks that require longer context, and performed experiments to benchmark all existing long range transformers. The results show that the BigBird model, unlike its counterparts, clearly reduces memory consumption without sacrificing performance.
Conclusion We show that carefully designed sparse attention can be as expressive and flexible as the original full attention model. Along with theoretical guarantees, we provide a very efficient implementation which allows us to scale to much longer inputs. As a consequence, we achieve state-of-the-art results for question answering, document summarization and genome fragment classification. Given the generic nature of our sparse attention, the approach should be applicable to many other tasks like program synthesis and long form open domain question answering. We have open sourced the code for both ETC (github) and BigBird (github), both of which run efficiently for long sequences on both GPUs and TPUs.
Acknowledgements This research resulted as a collaboration with Amr Ahmed, Joshua Ainslie, Chris Alberti, Vaclav Cvicek, Avinava Dubey, Zachary Fisher, Guru Guruganesh, Santiago Ontañón, Philip Pham, Anirudh Ravula, Sumit Sanghai, Qifan Wang, Li Yang, Manzil Zaheer, who co-authored EMNLP and NeurIPS papers.
A general goal of robotics research is to design systems that can assist in a variety of tasks that can potentially improve daily life. Most reinforcement learning algorithms for teaching agents to perform new tasks require a reward function, which provides positive feedback to the agent for taking actions that lead to good outcomes. However, actually specifying these reward functions can be quite tedious and can be very difficult to define for situations without a clear objective, such as whether a room is clean or if a door is sufficiently shut. Even for tasks that are easy to describe, actually measuring whether the task has been solved can be difficult and may require adding many sensors to a robot's environment.
Alternatively, training a model using examples, called example-based control, has the potential to overcome the limitations of approaches that rely on traditional reward functions. This new problem statement is most similar to prior methods based on "success detectors", and efficient algorithms for example-based control could enable non-expert users to teach robots to perform new tasks, without the need for coding expertise, knowledge of reward function design, or the installation of environmental sensors.
In "Replacing Rewards with Examples: Example-Based Policy Search via Recursive Classification," we propose a machine learning algorithm for teaching agents how to solve new tasks by providing examples of success (e.g., if “success” examples show a nail embedded into a wall, the agent will learn to pick up a hammer and knock nails into the wall). This algorithm, recursive classification of examples (RCE), does not rely on hand-crafted reward functions, distance functions, or features, but rather learns to solve tasks directly from data, requiring the agent to learn how to solve the entire task by itself, without requiring examples of any intermediate states. Using a version of temporal difference learning — similar to Q-learning, but replacing the typical reward function term using only examples of success — RCE outperforms prior approaches based on imitation learning on simulated robotics tasks. Coupled with theoretical guarantees similar to those for reward-based learning, the proposed method offers a user-friendly alternative for teaching robots new tasks.
Example-Based Control vs Imitation Learning While the example-based control method is similar to imitation learning, there is an important distinction — it does not require expert demonstrations. In fact, the user can actually be quite bad at performing the task themselves, as long as they can look back and pick out the small fraction of states where they did happen to solve the task.
Additionally, whereas previous research used a stage-wise approach in which the model first uses success examples to learn a reward function and then applies that reward function with an off-the-shelf reinforcement learning algorithm, RCE learns directly from the examples and skips the intermediate step of defining the reward function. Doing so avoids potential bugs and bypasses the process of defining the hyperparameters associated with learning a reward function (such as how often to update the reward function or how to regularize it) and, when debugging, removes the need to examine code related to learning the reward function.
Recursive Classification of Examples The intuition behind the RCE approach is simple: the model should predict whether the agent will solve the task in the future, given the current state of the world and the action that the agent is taking. If there were data that specified which state-action pairs lead to future success and which state-action pairs lead to future failure, then one could solve this problem using standard supervised learning. However, when the only data available consists of success examples, the system doesn’t know which states and actions led to success, and while the system also has experience interacting with the environment, this experience isn't labeled as leading to success or not.
Nonetheless, one can piece together what these data would look like, if it were available. First, by definition, a successful example must be one that solves the given task. Second, even though it is unknown whether an arbitrary state-action pair will lead to success in solving a task, it is possible to estimate how likely it is that the task will be solved if the agent started at the next state. If the next state is likely to lead to future success, it can be assumed that the current state is also likely to lead to future success. In effect, this is recursive classification, where the labels are inferred based on predictions at the next time step.
The underlying algorithmic idea of using a model's predictions at a future time step as a label for the current time step closely resembles existing temporal-difference methods, such as Q-learning and successor features. The key difference is that the approach described here does not require a reward function. Nonetheless, we show that this method inherits many of the same theoretical convergence guarantees as temporal difference methods. In practice, implementing RCE requires changing only a few lines of code in an existing Q-learning implementation.
Evaluation We evaluated the RCE method on a range of challenging robotic manipulation tasks. For example, in one task we required a robotic hand to pick up a hammer and hit a nail into a board. Previous research into this task [1, 2] have used a complex reward function (with terms corresponding to the distance between the hand and the hammer, the distance between the hammer and the nail, and whether the nail has been knocked into the board). In contrast, the RCE method requires only a few observations of what the world would look like if the nail were hammered into the board.
We compared the performance of RCE to a number of prior methods, including those that learn an explicit reward function and those based on imitation learning , all of which struggle to solve this task. This experiment highlights how example-based control makes it easy for users to specify even complex tasks, and demonstrates that recursive classification can successfully solve these sorts of tasks.
Conclusion We have presented a method to teach autonomous agents to perform tasks by providing them with examples of success, rather than meticulously designing reward functions or collecting first-person demonstrations. An important aspect of example-based control, which we discuss in the paper, is what assumptions the system makes about the capabilities of different users. Designing variants of RCE that are robust to differences in users' capabilities may be important for applications in real-world robotics. The code is available, and the project website contains additional videos of the learned behaviors.
Acknowledgements We thank our co-authors, Ruslan Salakhutdinov and Sergey Levine. We also thank Surya Bhupatiraju, Kamyar Ghasemipour, Max Igl, and Harini Kannan for feedback on this post, and Tom Small for helping to design figures for this post.
Open-domain long-form question answering (LFQA) is a fundamental challenge in natural language processing (NLP) that involves retrieving documents relevant to a given question and using them to generate an elaborate paragraph-length answer. While there has been remarkable recent progress in factoid open-domain question answering (QA), where a short phrase or entity is enough to answer a question, much less work has been done in the area of long-form question answering. LFQA is nevertheless an important task, especially because it provides a testbed to measure the factuality of generative text models. But, are current benchmarks and evaluation metrics really suitable for making progress on LFQA?
In “Hurdles to Progress in Long-form Question Answering” (to appear at NAACL 2021), we present a new system for open-domain long-form question answering that leverages two recent advances in NLP: 1) state-of-the-art sparse attention models, such as Routing Transformer (RT), which allow attention-based models to scale to long sequences, and 2) retrieval-based models, such as REALM, which facilitate retrievals of Wikipedia articles related to a given query. To encourage more factual grounding, our system combines information from several retrieved Wikipedia articles related to the given question before generating an answer. It achieves a new state of the art on ELI5, the only large-scale publicly available dataset for long-form question answering.
However, while our system tops the public leaderboard, we discover several troubling trends with the ELI5 dataset and its associated evaluation metrics. In particular, we find 1) little evidence that models actually use the retrievals on which they condition; 2) that trivial baselines (e.g., input copying) beat modern systems, like RAG / BART+DPR; and 3) that there is a significant train/validation overlap in the dataset. Our paper suggests mitigation strategies for each of these issues.
Text Generation The main workhorse of NLP models is the Transformer architecture, in which each token in a sequence attends to every other token in a sequence, resulting in a model that scales quadratically with sequence length. The RT model introduces a dynamic, content-based sparse attention mechanism that reduces the complexity of attention in the Transformer model from n2 to n1.5, where n is the sequence length, which enables it to scale to long sequences. This allows each word to attend to other relevant words anywhere in the entire piece of text, unlike methods such as Transformer-XL where a word can only attend to words in its immediate vicinity.
The key insight of the RT work is that each token attending to every other token is often redundant, and may be approximated by a combination of local and global attention. Local attention allows each token to build up a local representation over several layers of the model, where each token attends to a local neighborhood, facilitating local consistency and fluency. Complementing local attention, the RT model also uses mini-batch k-means clustering to enable each token to attend only to a set of most relevant tokens.
We pre-train an RT model on the Project Gutenberg (PG-19) data-set with a language modeling objective, i.e, the model learns to predict the next word given all the previous words, so as to be able to generate fluent paragraph long text.
Information Retrieval To demonstrate the effectiveness of the RT model on the task of LFQA, we combine it with retrievals from REALM. The REALM model (Guu et al. 2020) is a retrieval-based model that uses the maximum inner product search to retrieve Wikipedia articles relevant to a particular query or question. The model was fine-tuned for factoid-based question answering on the Natural Questions dataset. REALM utilizes the BERT model to learn good representations for a question and uses SCANN to retrieve Wikipedia articles that have a high topical similarity with the question representation. This is then trained end-to-end to maximize the log-likelihood on the QA task.
We further improve the quality of REALM retrievals by using a contrastive loss. The idea behind this is to encourage the representation of a question to get close to its ground truth answer and diverge from the other answers in its mini-batch. This ensures that when the system retrieves relevant items using this question representation, it returns articles that are "similar" to ground truth answers. We call this retriever contrastive-REALM or c-REALM.
Evaluation We test the model on long-form question answering using the ELI5 dataset, which is a part of the KILT benchmark, and is the only publicly available large-scale LFQA dataset. The KILT benchmark measures text retrievals using Precision (R-Prec) and text generation using ROUGE-L. The two scores are combined to give a KILT R-L score, which determines a model’s ranking on the leaderboard. We fine-tune the pre-trained RT model together with retrievals from c-REALM on the ELI5 dataset from KILT.
Our submission tops the KILT leaderboard for long-form question answering on ELI5 with a combined KILT R-L score of 2.36. It improves on the previous leaderboard entry of BART + DPR (KILT R-L score of 1.9), while having a similar number of parameters as the other models on the leaderboard. In terms of text generation quality, we see an improvement of +4.11, +5.78 and +9.14 Rouge-L over T5, BART + DPR and RAG, respectively.
Example Generations from the RT Model
Hurdles Towards Progress in LFQA However, while the RT system described here tops the public leaderboard, a detailed analysis of the model and the ELI5 dataset reveal some concerning trends.
We find little to no evidence that the model is actually grounding its text generation in the retrieved documents — fine-tuning an RT model with random retrievals from Wikipedia (i.e., random retrieval + RT) performs nearly as well as the c-REALM + RT model (24.2 vs 24.4 ROUGE-L). We also find significant overlap in the training, validation and test sets of ELI5 (with several questions being paraphrases of each other), which may eliminate the need for retrievals. The KILT benchmark measures the quality of retrievals and generations separately, without making sure that the text generation actually use the retrievals.
Moreover, we find issues with the Rouge-L metric used to evaluate the quality of text generation, with trivial nonsensical baselines, such as a Random Training Set answer and Input Copying, achieving relatively high Rouge-L scores (even beating BART + DPR and RAG).
Conclusion We proposed a system for long form-question answering based on Routing Transformers and REALM, which tops the KILT leaderboard on ELI5. However, a detailed analysis reveals several issues with the benchmark that preclude using it to inform meaningful modelling advances. We hope that the community works together to solve these issues so that researchers can climb the right hills and make meaningful progress in this challenging but important task.
Acknowledgments The Routing Transformer work has been a team effort involving Aurko Roy, Mohammad Saffar, Ashish Vaswani and David Grangier. The follow-up work on open-domain long-form question answering has been a collaboration involving Kalpesh Krishna, Aurko Roy and Mohit Iyyer. We wish to thank Vidhisha Balachandran, Niki Parmar and Ashish Vaswani for several helpful discussions, and the REALM team (Kenton Lee, Kelvin Guu, Ming-Wei Chang and Zora Tung) for help with their codebase and several useful discussions, which helped us improve our experiments. We are grateful to Tu Vu for help with the QQP classifier used to detect paraphrases in ELI5 train and test sets. We thank Jules Gagnon-Marchand and Sewon Min for suggesting useful experiments on checking ROUGE-L bounds. Finally we thank Shufan Wang, Andrew Drozdov, Nader Akoury and the rest of the UMass NLP group for helpful discussions and suggestions at various stages in the project.
Over the years, online multiplayer games have exploded in popularity, captivating millions of players across the world. This popularity has also exponentially increased demands on game designers, as players expect games to be well-crafted and balanced — after all, it's no fun to play a game where a single strategy beats all the rest.
In order to create a positive gameplay experience, game designers typically tune the balance of a game iteratively:
This process is not only time-consuming but also imperfect — the more complex the game, the easier it is for subtle flaws to slip through the cracks. When games often have many different roles that can be played, with dozens of interconnecting skills, it makes it all the more difficult to hit the right balance.
Today, we present an approach that leverages machine learning (ML) to adjust game balance by training models to serve as play-testers, and demonstrate this approach on the digital card game prototype Chimera, which we’ve previously shown as a testbed for ML-generated art. By running millions of simulations using trained agents to collect data, this ML-based game testing approach enables game designers to more efficiently make a game more fun, balanced, and aligned with their original vision.
Chimera We developed Chimera as a game prototype that would heavily lean on machine learning during its development process. For the game itself, we purposefully designed the rules to expand the possibility space, making it difficult to build a traditional hand-crafted AI to play the game.
The gameplay of Chimera revolves around the titular chimeras, creature mash-ups that players aim to strengthen and evolve. The objective of the game is to defeat the opponent's chimera. These are the key points in the game design:
Learning to Play Chimera As an imperfect information card game with a large state space, we expected Chimera to be a difficult game for an ML model to learn, especially as we were aiming for a relatively simple model. We used an approach inspired by those used by earlier game-playing agents like AlphaGo, in which a convolutional neural network (CNN) is trained to predict the probability of a win when given an arbitrary game state. After training an initial model on games where random moves were chosen, we set the agent to play against itself, iteratively collecting game data, that was then used to train a new agent. With each iteration, the quality of the training data improved, as did the agent’s ability to play the game.
For the actual game state representation that the model would receive as input, we found that passing an "image" encoding to the CNN resulted in the best performance, beating all benchmark procedural agents and other types of networks (e.g. fully connected). The chosen model architecture is small enough to run on a CPU in reasonable time, which allowed us to download the model weights and run the agent live in a Chimera game client using Unity Barracuda.
Balancing Chimera This approach enabled us to simulate millions more games than real players would be capable of playing in the same time span. After collecting data from the games played by the best-performing agents, we analyzed the results to find imbalances between the two of the player decks we had designed.
First, the Evasion Link Gen deck was composed of spells and creatures with abilities that generated extra link energy used to evolve a player’s chimera. It also contained spells that enabled creatures to evade attacks. In contrast, the Damage-Heal deck contained creatures of variable strength with spells that focused on healing and inflicting minor damage. Although we had designed these decks to be of equal strength, the Evasion Link Gen deck was winning 60% of the time when played against the Damage-Heal deck.
When we collected various stats related to biomes, creatures, spells, and chimera evolutions, two things immediately jumped out at us:
From these insights, we made some adjustments to the game. To emphasize chimera evolution as a core mechanism in the game, we decreased the amount of link energy required to evolve a chimera from 3 to 1. We also added a “cool-off” period to the T-Rex creature, doubling the time it took to recover from any of its actions.
Repeating our ‘self-play’ training procedure with the updated rules, we observed that these changes pushed the game in the desired direction — the average number of evolves per game increased, and the T-Rex's dominance faded.
By weakening the T-Rex, we successfully reduced the Evasion Link Gen deck's reliance on an overpowered creature. Even so, the win ratio between the decks remained at 60/40 rather than 50/50. A closer look at the individual game logs revealed that the gameplay was often less strategic than we would have liked. Searching through our gathered data again, we found several more areas to introduce changes in.
To start, we increased the starting health of both players as well as the amount of health that healing spells could replenish. This was to encourage longer games that would allow a more diverse set of strategies to flourish. In particular, this enabled the Damage-Heal deck to survive long enough to take advantage of its healing strategy. To encourage proper summoning and strategic biome placement, we increased the existing penalties on playing creatures into incorrect or overcrowded biomes. And finally, we decreased the gap between the strongest and weakest creatures through minor attribute adjustments.
New adjustments in place, we arrived at the final game balance stats for these two decks:
Conclusion Normally, identifying imbalances in a newly prototyped game can take months of playtesting. With this approach, we were able to not only discover potential imbalances but also introduce tweaks to mitigate them in a span of days. We found that a relatively simple neural network was sufficient to reach high level performance against humans and traditional game AI. These agents could be leveraged in further ways, such as for coaching new players or discovering unexpected strategies. We hope this work will inspire more exploration in the possibilities of machine learning for game development.
Acknowledgements This project was conducted in collaboration with many people. Thanks to Ryan Poplin, Maxwell Hannaman, Taylor Steil, Adam Prins, Michal Todorovic, Xuefan Zhou, Aaron Cammarata, Andeep Toor, Trung Le, Erin Hoffman-John, and Colin Boswell. Thanks to everyone who contributed through playtesting, advising on game design, and giving valuable feedback.
