ATHERXPERIMENT
May 13, 2024
ONCHAIN UTILITY AI: A LITEPAPER
Written by: @Tin Nguyen
TLDR:
This litepaper discusses the development of Utility Artificial Intelligence (Utility AI) in video games with the application of blockchain and Non-Fungible Tokens (NFTs). Utility AI is an advanced paradigm in AI that involves decisions being derived from dynamic evaluations of potential actions based on their perceived utility or "usefulness". It's applied in interactive environments such as video games, where characters make decisions based on a utility score.
The paper details the concept and mathematical models behind Utility AI, its applications in video games and other sectors, and its future improvement possibilities.
It further discusses the use of NFTs in gaming, enabling functionalities like provenance, ownership, and interoperability. Practical implementations of Utility AI in games, specifically in the Sipher Odyssey game, are also discussed in depth.
I. Foreword
As we delve into the fascinating world of Utility AI, it is vital to remember that this concept, while not new, represents a significant innovation in the gaming industry. The integration of Utility AI with blockchain, specifically in the form of Non-Fungible Tokens (NFTs), is set to unlock unknown value and bring a fresh wave of innovation. Looking ahead, with the current predictive and non-deterministic facets of Generative AI, coupled with decision-making frameworks like Utility AI, hold immense potential.
We are on the brink of an age where AI agents can serve as copilots and semi-autonomous assistants, not only within gaming but also in a broader context. This litepaper seeks to shed light on this exciting frontier and the opportunities it presents.
II. Primer on Utility AI
1. Introduction to Utility AI
Utility Artificial Intelligence, often referred to as Utility AI, represents a sophisticated paradigm in the realm of artificial intelligence, particularly within interactive environments such as video games. This approach diverges from traditional AI systems that rely on pre-defined rules or complex decision trees. Instead, Utility AI introduces a framework where decisions are derived from dynamic evaluations of potential actions based on their perceived utility, measured by specific, context-dependent criteria.
Most notable implementation of Utility AI in gaming is The Sims, alongside with Dragon Age Inquisition and Total War Three Kingdomws to name a few.
2. Conceptual Framework
At its core, Utility AI operates on the principle that agents—entities capable of making decisions, such as characters in a game—assess actions based on a utility score. This score quantifies the desirability of outcomes associated with each possible action, contingent upon the agent's current state and environmental factors. The fundamental mathematical model used in Utility AI involves calculating a utility value for each potential action, and then selecting the action with the highest utility.
The utility score of an action is computed through a function that aggregates multiple considerations, each representing a distinct aspect of the agent’s needs or goals, such as safety, hunger, or social interaction. These considerations are quantified using context-specific formulas that often incorporate a multitude of variables.
3. Mathematical Underpinnings
Utility functions typically involve weighted sums of these considerations, where the weights signify the relative importance of each factor in a given situation:
where:
Consideration functions can be linear or non-linear, designed to capture the varying impact of different states or resources on the agent’s decision-making process. For instance, a consideration function for hunger might exponentially increase as the agent's food level decreases, reflecting increasing desperation.
4. Dynamic Decision-Making
What sets Utility AI apart is its adaptability and flexibility. Unlike static rule-based systems, Utility AI can adjust to fluctuating game dynamics and player actions. This adaptability is facilitated by the potential for real-time recalibration of the weights and functions used in utility calculations, enabling AI agents to evolve their strategies and responses.
5. Applications in Production
Utility AI has seen applications across various sectors, not limited to gaming. In video games, its use is most pronounced in titles that require nuanced NPC behavior, such as strategic simulations and role-playing games. For instance, The Sims series by Electronic Arts utilizes a form of Utility AI to manage the complex social interactions and needs of its characters, making each sim's behavior seem uniquely reactive and life-like.
In addition to gaming, Utility AI has applications in robotics, where robots must make decisions that optimize their tasks while adapting to new inputs and environments dynamically. Ambitiously, in the onchain world, AI-driven virtual agents can leverage Utility AI to handle requests based on the urgency and nature of user-explicit requests, contextual environment, and historical artifacts.
III. Practical Implementations in Sipher Odyssey
1. Utility Functions
Each potential action that an agent can take is assigned a utility score through a utility function. This score represents how desirable or beneficial the action is in the agent's current context. The general form of a utility function can be expressed as:
where:
- U(as) is the utility of action a in state s.
- f represents the utility function, which is a combination of different considerations C.
- C are the different considerations or factors that affect the utility of the action (like casting skill, attack, melee, range, emote, etc.).
- Example of Player Bot's Tactical Selection
Player Bot AI Architecture
The Player Bot's Behavior Tree is composed of numerous Tactics. Each major Tactic contains smaller Tactics or task nodes to execute specific tasks (Move To, Fire, Melee...).
Each Tactic of the same level will use a Utility Value to choose the appropriate Tactic. For example, the Attack Tactic will choose one of the three sub-Tactics: SkillAttackTactical, MeleeAttackTactical, and WeaponRangedAttackTactical.
In the bot's Behavior Tree, all main Factors are updated first when the BT is browsed and saved in the Blackboard so that the Utility Service can read and calculate the corresponding Utility Values when needed. Currently, the update of Factors is separated into three main Services:
- Update Attack Decision Factors: Update factors to assist in selecting strategies for Attack.
- Update Allies Status: Update Factors to assist in selecting strategies to cover allies.
- GetCharacterFactors: Update static Factors of the subrace character that is configured and balanced by the Game Designer...
In BT of the bot, most of the Utilities of the Attack Tactical will depend on a factor called Threat. The Threat Factor reflects the bot's fear at the present time and depends on the following 3 Factors:
- Distance To Enemy: Values from 0 - 1, if = 1 then the enemy is outside the detection range and conversely very close to the bot.
- Health Percentage: Values from 0 - 1, if = 1 then the bot's current health = max health
- Health Lost Per Second: Values from 0 - 1, representing the Health Percentage lost per second
Utility AI Debugger Tool
Visual debugging the AI process is an important part to validate the Utility AI decisions and values.
2. Consideration Functions
Each consideration 𝐶C in the utility function is often modeled using its function based on specific attributes or states of the agent. These functions can be simple linear functions, polynomial equations, or more complex non-linear functions like sigmoid or exponential functions, depending on the desired behavior complexity.
3. Combining Considerations
The overall utility is often computed by combining multiple considerations. This can be done through weighted sums, multiplicative forms, or even more complex aggregations like prioritized weighting where certain needs must be satisfied before others are considered
where w are weights that signify the relative importance of each consideration.
4. Decision Rule
Once each potential action has a calculated utility, the decision rule can be one of the 4 types below.
- Max: Choose the child node with the highest Utility Value
- Min: Choose the child node with the lowest Utility Value
- Random: Randomly select any child node
- Random By Weight: Random selection based on probability.
Example of the Agent selects the action with the highest utility score. Mathematically, this is:
where a* is the chosen action that maximizes the utility function U.
IV. Possible In-game Implementations
1. Specific Adaptive AI Roles
Dynamic Squad Leader AI | Adaptive Support NPC | Strategic Enemy Boss |
Scenario: an NPC squad leader must make strategic decisions that affect the safety and effectiveness of the entire squad during combat. | Scenario: support NPCs assist the player by providing healing, buffs, or combat support depending on the situation. | Scenario: the enemy commander NPC orchestrates attacks, adapting to both the player's strategies and the unfolding dynamics of battle. |
Utility Functions:
Evaluate actions like flanking, retreating, or charging based on battlefield conditions such as the squad’s health, ammunition levels, and enemy positioning.
Considerations: The squad leader's decisions are influenced by various considerations:
Safety of Squad: Actions that minimize risk to squad members are prioritized.
Enemy Threat Level: The strength and positioning of enemy forces are assessed.
Mission Objectives: Balance between aggressive tactics and primary mission goals is maintained.
Decision-Making: The Utility AI system recalculates utilities for potential actions continuously as the battle scenario evolves. This allows the squad leader to adapt strategies dynamically and direct squad members to execute maneuvers that optimize survival and mission success. | Utility Functions:
Assign scores to actions such as healing a player, attacking an enemy, or deploying a defensive buff.
The scores are based on the immediate needs of the player and the NPC.
Considerations for Utility Evaluation:
Player Health:
Healing actions have higher utility if the player’s health is low.
NPC's Safety:
Evaluate the risk of the NPC entering combat when weighing potential actions.
Effectiveness of Support:
Consider how much an action would change the tide of battle.
Dynamic Response:
Utilize real-time feedback from the combat environment to prioritize the NPC’s actions.
Seamlessly shift between roles such as healer, attacker, or protector, based on which action offers the highest utility at any given moment. | Utility Functions:
The commander evaluates potential strategic options like launching an ambush, retreating, or calling for reinforcements.
Considerations for Decision-making:
Current Resources:
Assess available troops and materials when evaluating strategic options.
Player’s Strength and Position:
Evaluate the player’s force size and tactical positioning.
Long-term Goals:
Align tactics with overarching strategic objectives.
Proactive Strategy Adjustment:
Enable the Enemy Boss to anticipate player moves and adjust strategies proactively.
Optimize decision outcomes based on evolving battle conditions, making enemy units appear more intelligent and formidable. |
These examples demonstrate how Utility AI can enhance NPCs to adapt their behaviors to a wide range of scenarios, making each encounter feel natural, not too predictable, and dynamically responsive to the player's actions and environment.
2. Utility AI as Companion Stereotypes
Utility AI enables the design of highly adaptive and context-sensitive NPCs in action games. By configuring utility functions tailored to specific archetypes, we aim to create nuanced behaviors that enhance the strategic depth and interactivity of gameplay. Here’s an in-depth look at potential utility function setups for two NPC stereotype examples:
Tanker NPC | Assassin NPC |
Role Description: The Tanker NPC is designed to absorb damage and protect other members of the player’s team, acting as a shield in critical combat situations. | Role Description: The Assassin NPC specializes in high damage output, targeting high-value targets quickly and efficiently, usually at the risk of being vulnerable due to lower health. |
Attract Aggression (Tanker):
Utility Calculation: U(a) = Enemy Focus × Distance from Allies
𝑈(𝑎)=Enemy Focus×Distance from Allies
Purpose: Prioritizes positioning the Tanker between enemies and allies, maximizing enemy focus on the Tanker while minimizing risk to allies.
Defensive Posture (Tanker):
Utility Calculation: U(a) = (1 - Tanker Health Percentage) × Available Defenses
𝑈(𝑎)=(1−Tanker Health Percentage)×Available Defenses
Purpose: Increases the utility of using defensive abilities (like shields or area control weapons) as the Tanker’s health decreases, ensuring survival in the thick of battle.
Protect Ally:
Utility Calculation: U(a) = Ally Health Deficit × Proximity to Threatened Ally
𝑈(𝑎)=Ally Health Deficit×Proximity to Threatened Ally
Purpose: Triggers protective maneuvers or abilities when nearby allies are at low health, focusing on those who are most vulnerable. | Target High-Value Enemy:
Utility Calculation: U(a) = Enemy Value / Distance to Enemy
𝑈(𝑎)=Enemy Value/Distance to Enemy
Purpose: Maximizes the impact of attacks by prioritizing enemies based on their strategic value (like enemy leaders or healers) and their accessibility.
Stealth Engagement:
Utility Calculation: U(a) = Cover Availability × Enemy Alertness - 1
𝑈(𝑎)=Cover Availability×Enemy Alertness−1
Purpose: Enhances the likelihood of engaging enemies while in stealth mode, particularly when cover is available and enemy alertness is low, ensuring safety and surprise.
Escape to Safety (Assassin):
Utility Calculation: U(a) = (1 - Assassin Health Percentage) × Escape Routes
𝑈(𝑎)=(1−Assassin Health Percentage)×Escape Routes
Purpose: Prioritizes disengagement and retreat tactics as the Assassin’s health decreases, allowing them to recover or avoid knockdown. |
Tanker in Action: In a scenario where multiple allies are under fire, the Tanker’s AI would calculate which protective action—either shielding a specific ally or using a group defensive ability—has the highest utility based on the current health and position of the team members relative to enemies. | Assassin in Action: When tasked with eliminating a high-value target, the Assassin would assess the potential paths to approach the target, the likelihood of being detected, and the impact of eliminating the target versus the risk. The chosen path would be the one where the calculated utility of a stealthy, effective strike is highest. |
3. Examples of the mathematical utility functions
Here's the table utility function for examples of the Tanker and Assassin NPC stereotypes, where they must consider their current equipped ranged weapon (like revolvers, shotguns, and rocket launchers) and behave accordingly:
NPC Stereotype | Utility Function | Utility Calculation | Description |
Tanker | Attract Aggression | 𝑈(𝑎)=Enemy Focus×Distance from Allies | Prioritizes positioning to attract enemy focus and protect allies by acting as a frontline barrier. |
Tanker | Defensive Posture | 𝑈(𝑎)=(1−Tanker Health Percentage)×Available Defenses | Increases use of defensive abilities as health decreases to ensure survival in combat. |
Tanker | Protect Ally | 𝑈(𝑎)=Ally Health Deficit×Proximity to Threatened Ally | Activates protective maneuvers when nearby allies are at low health, focusing on the most vulnerable. |
Tanker | Weapon-Based Positioning | 𝑈(𝑎)=Weapon Range×Enemy Proximity | Adjusts positioning based on weapon range, seeking optimal distance to maximize effectiveness while maintaining defense. |
Assassin | Target High-Value Enemy | 𝑈(𝑎)=Enemy Value/Distance to Enemy | Maximizes attack impact by prioritizing high-value targets that are within accessible range. |
Assassin | Stealth Engagement | 𝑈(𝑎)=Cover Availability×Enemy Alertness−1 | Enhances the likelihood of engaging in stealth mode, exploiting cover and low enemy alertness. |
Assassin | Escape to Safety | 𝑈(𝑎)=(1−Assassin Health Percentage)×Escape Routes | Prioritizes disengagement and retreat to safe positions when health is low. |
Assassin | Weapon Utilization Tactic | 𝑈(𝑎)=Stealth Efficiency×Weapon Silence Factor | Chooses attack methods based on weapon noise and efficiency, favoring silent engagements to maintain stealth. |
Additional Details:
- Tanker - Weapon-Based Positioning: This function ensures that the Tanker NPC adjusts its position based on the effective range of the equipped weapon, such as staying at mid-range with shotguns or keeping distance with rocket launchers, while still fulfilling its role as a protector.
- Assassin - Weapon Utilization Tactic: For the Assassin, this utility function evaluates how well a weapon supports stealth tactics. Silent weapons like suppressed revolvers increase the utility of staying undetected, whereas louder weapons might shift the utility towards quicker, more aggressive attacks before escaping.
V. Possible Advance Improvements to Utility AI
1. Learning from Human Feedback
Integrating machine learning techniques can greatly enhance utility functions. By allowing the AI to learn from the outcomes of its actions, you can refine utility calculations based on success or failure. Reinforcement learning, for example, can be used to adjust the weights or shapes of utility functions based on rewards received from the environment. This can help the AI adapt to changing game dynamics and player interactions more effectively.
2. Contextual and Temporal Dynamics
Utility functions often benefit from incorporating more contextual and temporal factors. This means not only looking at the current state but also considering past states and predicting future states. For example, using decay functions to model how the importance of needs changes over time or integrating predictive models to anticipate future states can make decisions more strategic and forward-thinking.
3. Multi-Attribute Utility Theory (MAUT)
This approach involves more sophisticated mathematical modeling to combine multiple criteria into a single, composite utility score. It allows for more nuanced trade-offs between competing needs or goals. By defining utility functions that account for multiple attributes and using techniques like exponential or logistic functions, you can create more realistic decision-making processes that mimic human-like trade-offs.
4. Emotion and Personality Modeling
Incorporating psychological factors such as emotions and personality can dramatically improve the realism of utility-based AI. For example, defining utility functions that change based on an agent’s emotional state or personality traits can lead to behavior that better reflects complex human interactions.
5. Group Agent Dynamics and Social Influence
Utility functions can also be extended to consider social influences and group dynamics. This is particularly relevant in games like The Sims where social interactions play a crucial role. Agents could adjust their utility calculations based on the actions, reactions, and presence of other agents, aligning their behaviors with social norms or relationships.
6. Hybrid Approaches
Combining utility-based AI with other decision-making paradigms can lead to more robust systems. For instance, integrating rule-based systems for certain predictable behaviors and using utility functions for more complex, context-dependent decisions can optimize both performance and realism.
7. Continuous Optimization
Finally, the parameters of utility functions (like weights and shapes) can be continually optimized through algorithms. In more advanced implementations, the utility functions can be adjusted dynamically based on learning or changes in the game environment. This could involve adjusting the weights 𝑤 or even the form of the consideration functions based on past outcomes or evolving game states.
VI. Blockchain implementations
1. NFTs Explained
Non-fungible tokens (NFTs) are digital assets under the ERC-721 standard, ensuring that each token is one of a kind. Unlike ERC-20s, which are interchangeable, each NFT has a distinct data structure and unique data record, making them ideal for representing rare or unique items, such as digital art, collectibles, and, in the context of gaming, AI brains.
2. Benefits for Gaming
Integrating NFTs with blockchain technology in gaming enables several innovative functionalities:
- Provenance and Ownership: Players can verify the history and ownership of digital assets (like AI brains) transparently, enhancing trust and value perception.
- Interoperability: Assets represented as NFTs can potentially be used across different games and platforms, assuming common standards, and increasing their utility and long-term value.
- Security: Blockchain’s decentralized nature prevents tampering and fraud, ensuring that player assets are secure.
3. Storing and Transfering Utility AI
Each AI brain in the game is stored as an NFT on the blockchain. This method provides several advantages:
- Unique AI Capabilities: Each NFT can represent a unique set of AI behaviors and strategies, ensuring a diverse range of AI interactions within the game.
- Transferability: Players can transfer and/or trade AI brains on NFT marketplaces, based on the rarity or effectiveness of the AI strategies encapsulated within each NFT.
4. Traceability and Transparency
Blockchain technology allows for the tracking of NFT transactions (AI brains) with high fidelity. This traceability ensures that:
- Players can see the complete history of any Utility AI brain, including its creation, previous owners, and changes over time.
- Trading and Modifying AI Brains: Explain how players can trade and modify AI brains to suit their gameplay strategies.
VII. Utility AI for games outside of games
1. Interoperability due to the design of Utility AI
Core Principles of Utility AI Transferability
Utility AI systems make decisions by evaluating the utility of various actions based on a set of input parameters and then choosing the action with the highest utility. Since these decisions are based on abstract criteria rather than specific game mechanics, the fundamental logic of a Utility AI system can be applicable across multiple game types. Here’s how this transferability could manifest:
- Generic Decision Frameworks: Utility AI systems are built on frameworks that assess situations using general criteria like distance, health, enemy presence, available skills, ammos, standard move sets, etc., which are common across many games.
- Adaptability of Criteria Weights: The weights assigned to different criteria in Utility AI can be adjusted for different game dynamics or player strategies without altering the core decision-making algorithms.
Challenges in Utility AI Interoperability
Despite the inherent adaptability of Utility AI, some challenges need to be addressed to fully realize its cross-game interoperability:
- Context-Specific Adjustments: While the decision-making framework might be generic, specific utility functions might need adjustments to better fit different game contexts or genres. For example, the utility of "distance to enemy" in a strategy game versus a first-person shooter may require different interpretations and weightings.
- Integration with Game-Specific Mechanics: Each game has unique mechanics and systems that the Utility AI must interact with. Adapting the AI to understand and respond to these unique elements is necessary for effective functionality.
Technological Facilitators for Interoperability
Several technological approaches can facilitate the interoperability of Utility AI:
- Utility AI Framework: Developing a unified framework that can translate between the game-specific contexts and the AI’s decision-making processes can help adapt AIs to new games more easily.
- Standardized AI Interfaces: Creating standardized interfaces for AI inputs and outputs can help developers design Utility AIs that are easily pluggable into different games, leading to emergent behaviors.
Possible examples of Utility AI interoperability can be:
- Cross-Genre AI: An AI designed for a tactical role-playing game could be adapted to provide tactical support in a real-time strategy game, with minimal changes to its core utility assessments.
- Community-Driven AI Improvement: Players and developers could propose modifications to an AI’s decision-making parameters based on its performance in different game scenarios, which could then be voted on or implemented via decentralized governance mechanisms.
2. Fostering further Interactions and Player Engagement external to the core game loop
NFT technology facilitates a dynamic market for trading AI brains among players. This market could function as an economic game outside of the action game, examples are as follows:
- Onchain Marketplace: Players list their AI brains for sale or auction. Smart Contracts automatically execute the terms of a trade when conditions are met, ensuring fairness and eliminating the need for intermediaries.
- Pricing Mechanisms: AI brains could be priced based on their proven battle effectiveness, the rarity of the strategies they employ, and their historical performance in various game scenarios.
- Economic Benefits: Players can attempt to make a profit by trading AI brains based on their market value, which could fluctuate based on their rarity, effectiveness, or demand. This introduces an economic strategy component where players must decide the best time to buy or sell AI assets.
VIII. Strategic Depth Through Utility AI
1. Diverse AI Capabilities and Player Specialization
Each AI brain, stored and traded as an NFT, possesses unique utility functions that dictate different strategies and behaviors. This diversity allows players to specialize by acquiring and deploying AI brains that complement their play style or tactical needs. For example:
- Tactical Players: Might seek AI brains that excel in strategic positioning and team support.
- Aggressive Players: May prefer AI brains that enhance offensive capabilities and direct confrontation.
By acquiring AI brains, players can fine-tune their strategic approach to matches, choosing AIs that best suit specific roles, missions, or the meta game.
2. Evolving Meta
The ability to trade AI brains encourages an evolving game meta where certain AI strategies may become popular as gameplay trends shift. Players must continuously adapt, acquiring new AI brains to stay competitive or to counter prevailing strategies. This fluid meta-game adds a layer of strategic decision-making tied not only to in-game actions but also to player choices in the marketplace.
3. Tactical Advantages
Acquiring a rare or highly effective AI brain can provide a significant tactical advantage in gameplay, making the timing and selection of trades crucial strategic decisions.
4. Customization and Adaptability
Trading allows players to customize their strategic toolkit, choosing from a broader array of AI behaviors and tactics than they could develop on their own. This level of customization fosters a more adaptable approach to gameplay, where players can swiftly change tactics by swapping AI brains to better handle new challenges or unexpected situations in the game.
5. Player Engagement and Longevity
The trading aspect encourages ongoing player engagement and game longevity. Players are motivated to continue playing to collect valuable AI brains, engage with the community in trading activities, and experiment with different strategic combinations. This not only keeps the game fresh and interesting but also helps build a vibrant community around AI brain collection and trading.
6. Strategic Trading and Collaboration
Trading of Utility AI ‘brains’ can foster collaboration and strategic alliances among players. Trading marketplaces or guilds will form, where players exchange information and AI brains to mutual benefit, adding a layer of social strategy to the game.
IX. Integration of Blockchain technology
Leveraging the [Redacted Protocol], Utility AI can be implemented within various static NFTs that are game or even non-game-related, breathing life into what is lifeless.
In overview, [Redacted Protocol] creates the scaffolding data structure for NFT characters. This data structure is extendable and configurable, whilst maintaining the ownership connection with the underlying original object or character.
This allows various game worlds and projects to have different configurations and applications of data, including the data structure used for Utility AI.
This flexible system also allows for different game developers to post updates to the Utility AI prevalence in their own universe, allowing near real-time updates and changes without impacting the original or other ‘verses’.
[Redacted Protocol] Implementations
- A framework for projects to build their interactions with blockchain using standardized configurations. Configurations can vary in complexity to match the casual or technically savvy nature of web3 users.
- Give project owners the right to own their Utility AI at the contract level. Facilitating the ability to develop a healthy market for Utility AI in gaming and other aspects.
- Through this framework, project owners will have additional solutions to provide a transparent system to incentivize their respective community of contributors and modders. All participants will have their contributions tracked, recorded, and rewarded onchain.
Modding of games has been around for a long time, [Redacted Protocol] along with Utility AI will allow a new era of modding AIs, with the long-term visions that everyone will have multiple AI Agents as copilots and companions in their gaming and onchain adventures.
X. Integration Challenges and Solutions
Challenge 1: Interfacing game engines (Unreal Engine, Unity, etc) with blockchain
Problem: Game Engines do not natively support blockchain functionalities, which means there is no out-of-the-box solution for integrating blockchain data or executing transactions directly within the engine.
Possible Solution:
- API Integration: Develop or utilize existing APIs that allow Unreal Engine and other game engines to communicate with blockchain networks. This can be achieved through middleware that acts as a bridge between the game engine and the blockchain.
- Smart Contract Interaction: develop plugins that enable game clients and servers to interact with smart contracts on the blockchain, allowing for operations like minting, transferring, and querying NFTs.
Challenge 2: Real-Time Performance and Scalability
Problem: Blockchain transactions may not meet the real-time performance requirements of fast-paced games developed in Unreal Engine such as Sipher Odyssey.
Solution:
- Off-Chain Operations: Perform game logic calculations off-chain within Unreal Engine and only interact with the blockchain for critical transactions such as finalizing ownership transfers of NFTs of Utility AIs or recording significant game events that impact the AIs.
- Layer-2 Solution: L2s and other similar infrastructure should be designed to handle transactions faster and more efficiently, reducing latency and improving AI performance.
Challenge 3: Data Synchronization and Consistency
Problem: Ensuring that the game state and blockchain state are synchronized can be challenging, especially in multiplayer settings where players might interact with the blockchain at different times or from various locations.
Solution:
- Preload: Fast-paced action games require pre-set-up event listeners or webhooks that trigger updates in the game state in the game client and DGS whenever relevant game Utility AI inquiry events occur, ensuring consistent AI states across the game and blockchain.
- Post-change Update: Once changes have been made to the Utility AI, signed transactions with attestations by the DGS will be triggered that will update any changes to the Utility AI NFTs on the chain.
XI. Future Outlook and Development Roadmap
Near-term and Medium-term Objectives:
- Develop and release Utility AI in Sipher Odyssey as companion NPCs that will assist the players in their adventure across Sipheria
- Develop and release Utility AI as Digital Assets that are of variant complexity, traits, and adaptability.
- Integrate transferability and trading of Utility AIs
- Integrate Utility AI from Sipher Odyssey to other games within the Ather Collections and partners
Long-term Vision:
- Develop a shared framework that will allow the modifications and growth of Utility AIs
- Develop SDKs that allow the integration of Utility AIs on chain into various Game Engines and Onchain Autonomous Worlds
- Develop a shared framework that will allow the personalization and permanence of Utility AI, on-chain.
- LLMs Integrations
- Contextual Interaction: LLMs can process and respond to player inputs or environmental narratives in natural language, providing a rich layer of interaction that Utility AI can use to refine its decision-making.
- Enhanced Decision Frameworks: By feeding LLM-generated insights into a Utility AI’s decision matrix, the AI can make more informed choices that consider broader contextual information, not just game mechanics.
- Dialogue and Interaction: LLMs already can generate dialogues and interactive scripts for characters, which will then be assessed by the Utility AI to determine the most appropriate responses and actions based on the current state and goals of the AI character. This gives an additional layer of ‘control’ over the AI interactions.
- Dynamic Storytelling: In narrative-driven games, combining LLMs with Utility AI allows non-player characters (NPCs) to not only participate in the story based on predefined scripts but also dynamically alter their behavior and responses based on player interactions and game developments.
- Advanced Simulations: In training or simulation applications, various Tool Use integrations can lead to more useful scenarios where AI Agents react both logically and contextually to human input requests, serving almost as ‘helpful assistants’
XII. Final Words
Utility AI is not new, but the integration of Utility AI with blockchain technology represents a novel advancement in the realm of gaming onchain. This combination not only enhances the dynamic capabilities of in-game characters through advanced decision-making algorithms but also revolutionizes how game assets, particularly AI brains, are managed, traded, and valued. Key benefits include:
- Enhanced NPC Behavior: Utility AI enables NPCs to make more sophisticated and contextually appropriate decisions, creating a richer and more immersive player experience.
- Digital Asset Ownership: By storing AI brains as NFTs on the blockchain, we ensure transparent and secure ownership, allowing players to own & trade in these digital assets genuinely.
- Increased Player Engagement: The ability to trade and modify AI brains as NFTs adds a layer of strategic depth to the game, keeping players engaged through both gameplay mechanics and economic incentives.
- Innovation in Game Development: This integration pushes the boundaries of what's possible in game development, setting a new standard for creativity and technological integration.
These frameworks allow Utility AI to simulate a wide range of realistic behaviors and ensure that agents make decisions that are contextually appropriate and diverse, mimicking a semblance of rationality and autonomy seen in human behavior while given the chance of pseudo-emergent or slightly unexpected behaviors. This approach is particularly powerful in environments that simulate complex social interactions and personal needs, such as onchain worlds full of ever-growing classes of Generative AI Agents.
Note: These suggestive improvements although exciting, one must keep in mind and consider the practicality and relevancy of each implementation. The ultimate goal of any AI system should be enhancing positive experiences for human users, whether social, interactive, peace of mind, time-saving, or others.
We invite the community, developers, players, and potential partners to join us in this exciting journey as we continue to explore and expand the possibilities of Utility AI and blockchain in gaming. Whether playing the game or contributing ideas, your participation is vital to our collective success.
- Players: We encourage you to experience Sipher Odyssey and other games in the Ather Collective ecosystem, provide feedback, and become active contributors to the development of gaming AI brains.
- Developers: Collaborate with us, and build together with us to create new novel gaming and social applications
LIKE WHAT YOU READ? CHAT WITH US!
- Join Sipher’s discord: @playSipher
- Follow us and say hello on X: @playSipher @AtherLabs
SUBSCRIBE
Subscribe to our newsletter today and be the first to receive Sipher news, AtherXplorer weekly issues, and new AtherXperiment posts. Sign up now and get ahead of the game!