tig-monorepo/docs/guides/advances.md

Rewarding Innovation in The Innovation Game

Introduction

The rewards for innovation in The Innovation Game (the "Game") are designed to be sufficient to incentivise innovators to submit their innovation ("Contributions") to the Game on the understanding that the Contributions will be made available to others as the basis for open innovation under the TIG Open Data License, and for further innovation by participants in the Game under the terms of the TIG Innovator Outbound Game License.

By sufficiency we mean that the reward must be at least enough to cause commercially valuable innovation (by implication, what is commercially valuable should, at least, be novel over the state of the art, innovative and free from encumbrances) to be submitted to the Game. What is sufficient reward may not necessarily be equitable in terms of fully compensating an innovator for their contribution of value (that is almost impossible to determine at the point of submission of the Contribution), at least not immediately, it just needs to be enough to cause the innovator to submit their Contribution.

In an ideal world the rewards to innovators would be exactly commensurate with the value added by an innovator’s Contribution. Because the Game seeks to reward Contributors for improved algorithm performance, it is intuitively attractive to look for a method of quantitatively measuring algorithmic performance but this would add significant (and what we consider unnecessary) complexity and also fail to recognise a number of realities including, the difficulties of objectively measuring and verifying the quantum of immediate commercial value of a particular Contribution, its spill over value and any latent value realised only later in time after the rewards have been assessed. Accordingly, we have decided, for now, to keep it as simple as possible. We will seek only what is sufficient in terms of reward to an innovator to compensate them for their effort and ingenuity to incentivise them to submit their innovation to the Game. This assessment of reward may not necessarily look objectively fair in relative terms (when comparing one algorithm to another), particularly in hindsight, but any apparent inequity is mitigated by the fact that the value of the TIG token will represent all off the accrued value of all Contributions over time and so each innovator will have the chance to benefit (if they hold some or all of their tokens) from the value created not just by their Contribution but also that from the Contributions of everyone else throughout the life of the Game.

One of the most obvious rewards offered by the Innovation Game for Contributions is the TIG token. Whether the token rewards are sufficient to incentivise the submission of Contributions will depend on the token price prevailing at the time of submission, the number of challenges to which rewards are allocated and, also to some extent, on the expected future value of the token. Some submissions will also be motivated by other forms of reward such as academic impact.

Because the token price will vary from time to time, we will only know if, as the basis of reward it is sufficient, by gathering empirical evidence provided by the quality of submissions in practice, and by feedback from potential innovators. For this reason, the Team will always be on alert to adjusting the rewards for the Game to make them sufficient. Whilst the protocol is still in development innovator rewards will be under constant review.

If an innovator decides to withhold their innovation from the Game because they assess that the reward is not sufficient incentive, they will be exposed to the jeopardy that, whilst they wait, someone else may submit an equivalent or better Contribution and resultingly then be deprived of the rewards that they may have earned themselves by submitting their algorithm.

A fundamental cornerstone of the Game is the synthetic market that is created by incentivising benchmarkers to create demand for the most performant algorithm implementations. The value of a Contribution is measured in that market by benchmarkers; the more a Contribution is adopted, the more reward is allocated to it. If a change of a single line of code in an existing code implementation (whether it relates to a change to the algorithmic method or to a code optimisation) makes sufficient difference that benchmarkers, acting rationally to improve their performance in the Game, adopt it instead of other available alternatives, then the market is designed to reward that change.

In the context of the Game, with the exception below related to the TIG Game Rules*, it is crooked thinking to simply assess the quantum of lines of code present or changed for determining the appropriateness for reward or what the quantum of that reward should be. If the synthetic market is functioning properly the best algorithm implementation will prevail and an enquiry into "why" it is prevailing is not relevant. It is the job of the Team to ensure that the synthetic market is the closest proxy that it can be for rewarding the algorithmic performance and utility most valued by commercial enterprise.

*The TIG Game Rules do mandate that a Contribution must make a "meaningful difference" over existing implementations in the Game. This will stop straightforward plagiarism where no or insignificant additional value is added by a Contributor and this aspect of the Game Rules will be policed.

Defining and Classifying Innovation

Generally, improvements in algorithmic methods tend to yield exponential or order-of-magnitude efficiency gains by changing the problem-solving approach, while the way in which algorithmic methods are implemented in code tend to provide incremental improvements. It is important therefore, that TIG places emphasis on incentivising the submission of innovative algorithmic methods. TIG intends to do that by introducing a process for identifying innovative algorithmic methods that will then be eligible for additional rewards.

We propose to distinguish two different types of Contribution in the context of the Game and rewarding them differently: (1) Algorithm Implementations (code); and (2) Algorithmic Advances (methods**)**.

Implementations of algorithms comprise two elements; (i) an algorithmic method (the fundamental approach or strategy for solving a problem, independent of specific code or language); and (ii) an expression of the algorithmic method in code.

In the Game the algorithmic method element may be a Algorithmic Advance or an algorithmic method that is not a Algorithmic Advance.

Algorithmic Advances

TIG believes that the determination of whether an algorithmic method is an Algorithmic Advance should be done by a combination of; (i) a token weighted vote assessing novelty and inventiveness*; and (ii) an assessment of the performance of the algorithmic method determined solely by the extent of its adoption by benchmarkers in the "TIG synthetic market" (see: Accessing Advance Rewards below).

* For a discussion and explanation of novelty and inventiveness in the context of the Game see the document Advance Rewards Guide for Token Holders.

Algorithm Implementations

Algorithm Implementations are expressions of algorithmic methods in code and in the context of the Game they may be an expression of an algorithmic method which is a Algorithmic Advance or an expression of an algorithmic method that is not a Algorithmic Advance.

Token Rewards for Innovation

To reflect the patentability and significant impact on performance that innovative algorithmic methods can have compared with algorithm implementations, TIG will offer additional rewards for Algorithmic Advances. These additional rewards, so called "Advance Rewards", are brought about through the design of the protocol enabling the rewards for Algorithmic Advances to persist for multiple Rounds where they continue to provide the basis of algorithm implementations adopted by benchmarkers.

An Algorithm Implementation which does not embody an algorithmic method which is eligible for potential Advance Rewards, only rewards the innovator that submitted the Algorithm Implementation with Standard Rewards (subject to satisfaction of adoption thresholds). Because of the absence of a Algorithmic Advance, Advance Rewards notionally allocated for Algorithmic Advances will, instead, be allocated to the TIG treasury and used to bootstrap Algorithmic Advance development (these are referred to in the Game as "Bootstrap Rewards").

For Algorithm Implementations that embody an algorithmic method which is eligible for potential Advance Rewards, two types of reward are potentially available (subject to satisfaction of adoption thresholds):

• Standard Rewards will be available for the innovator that submitted the adopted implementation of the Algorithmic Advance.

• Advance Rewards will be available for the innovator that submitted the Algorithmic Advance embodied in the implementation.

The innovator(s) earning Standard Rewards and Advance Rewards in respect of an Algorithm Implementation may be the same or different entities. An innovator can earn both Standard Rewards and Advance Rewards simultaneously or at different times.

Standard Rewards for Algorithm Implementations

10% of the total rewards available in a Round will be allocated to Standard Rewards for Algorithm Implementations.

Advance Rewards for Algorithmic Advances

20% of the total rewards available in a Round will be allocated to Advance Rewards for Algorithmic Advances.

The most significant difference between Standard Rewards and Advance Rewards is that the Advance Rewards will continue to be earned where the Algorithmic Advance is inherited by an implementation, meaning that Advance Rewards can persist for longer because they will be earned in connection with any, and all adopted implementations of that Algorithmic Advance. Standard Rewards will be earned only for as long as the respective implementation is adopted by benchmarkers. Notwithstanding that the allocations of rewards are prima facie equal for both Advance Rewards and Standard Rewards (at 15% each), because of the factor of persistence of adoption in calculating aggregate rewards over time, Advance Rewards in respect of significant algorithmic breakthroughs are likely to far exceed Standard Rewards.

Advance Rewards are expected to be greater than Standard Rewards to reflect four attributes of Algorithmic Advances;

1. they are potentially patentable (bringing greater commercial value to the TIG ecosystem);

2. they are novel (there is no novelty test in the Game for Code Optimisations which means they may have lower commercial value than Algorithmic Advances);

3. they generally have greater potential to make significant performance improvements over the state of the art (bringing greater value to the TIG ecosystem); and

4. they are more certain to have value extrinsic to the Game.

It is the Team’s belief that sufficiency of reward can be achieved in respect of Standard Rewards by appropriately setting the period during which a Code Optimisation will be guaranteed not to be used as the basis for a subsequent Code Optimisation contributed by a third party. At present this period is two Rounds. Only the behaviour of innovators can ultimately signal to us whether, at a minimum, this "protected" period for a Code Optimisation is sufficient to incentivise its submission to the game and we will be monitoring this as the protocol develops.

Accessing Advance Rewards

If an Innovator believes that their Contribution is or embodies a potential Algorithmic Advance, then they may request that their Contribution is considered for approval as a Algorithmic Advance. For a Contribution to be a Algorithmic Advance, the following must ALL be satisfied with respect to the Contribution:

1. It must be declared, by Token Holder Vote that the Algorithm implementation subject to review embodies an algorithmic method that is eligible for potential Advance Rewards;

2. The intellectual property rights embodied in the Contribution must be irrevocably assigned to TIG in accordance with the TIG IP Policy;

3. The Contributor must burn two hundred and fifty (250) TIG tokens for each requested Token Holder Vote; and

4. In a rolling 1 week window the sum of all algorithm implementations which embody the algorithmic method subject to review must achieve a sufficient degree of adoption by Benchmarkers (where for Standard rewards the adoption threshold by benchmarkers is presently 25% before the algorithm implementation is merged to obtain Standard Rewards, the corresponding adoption threshold by benchmarkers for potential Algorithmic Advances will be 50%). The reason why the threshold of adoption is higher for potential Algorithmic Advances is that TIG wishes to reserve Advance Rewards for algorithmic improvements that offer a significant universal improvement in performance (which will be valued more highly by commercial licensees).

Essentially the Token Holder Vote is assessing the novelty and inventiveness of a contributed algorithmic method. To maximise the likelihood that an Algorithmic Advance will be ratified by Token Holder Vote, an innovator seeking Advance Rewards is advised to disclose the following information so that it may be assessed by token holders:

• Novelty: Provide results of a prior art search to identify any existing disclosures that might affect the novelty of your algorithmic method.

• Inventiveness: Clearly document how the algorithmic method differs from existing solutions (i.e. is non-obvious) and any new technical effect it achieves.

• Technical Effect: Document of how the algorithmic method offers the potential to provide the basis for significant technical advancements. Document how the algorithmic method might have real world application e.g. in fields such as computer security, medical imaging, autonomous systems. or data compression.

Encumbrances

The value of Contributions to TIG will be compromised if the Contributions are encumbered by third party intellectual property rights and terms of the TIG Inbound Game License seeks to deal with this.

The terms of the TIG Inbound Game License state:

"Submitter represents that Submitter is legally entitled to grant the above licenses and that, to the best of Submitter’s knowledge, the Work does not infringe any rights of a third party."

Attribution of Copyright Owner

The header file for each algorithm implementation submitted to the Game must comply with the terms of the TIG Inbound Game License. This establishes the identity of the copyright owner (which may or may not be the original author of the code) of the implementation of the algorithmic method in source code.

Unique Algorithmic Method Identifier (UAI)

With the introduction of Advance Rewards, the significance of the correct identification of the original creator of the algorithmic method embodied in an algorithm implementation has even greater significance.

It is important that the identification is not only correct, but that it appears in all algorithm implementations which embody the relevant algorithmic method.

The copyright notice does not necessarily identify the original creator of the algorithmic method embodied in the implementation (because the creator of the algorithmic method and the author of the code implementing the algorithmic method may be different). For this reason, the header file must also include information which identifies the creator of the algorithmic method. To do this TIG will issue a Unique Algorithmic Method Identifier (UAI) for any algorithmic method submitted to TIG. This UAI will be included in the header file of the algorithm implementation embodying the algorithmic method before it is published by TIG. The UAI will enable TIG to identify which algorithmic methods are eligible for Advance Rewards and which ones are not.

The Game Rules will mandate that where an Algorithm Implementation (Algorithm B) is a modified implementation of an existing Algorithm Implementation (Algorithm A) and continues to embody an algorithmic method substantially similar to that embodied in Algorithm A, if Algorithm A has an associated UAI then that UAI must be entered into the header file of Algorithm B. This should create a correct indication of the heritage of the code back to the original creator of the relevant embodied algorithmic method.

There will be a very strong incentive for innovators that benefit from Advance Rewards to ensure that they protect the persistence of the Advance Rewards associated with their Algorithmic Advance, by carefully scrutinising the novelty of any challenger algorithmic methods submitted to TIG.

Attribution of Contributor

The header file must also include an attribution of the name of the Contributor (referred to as "Submitter" in the TIG Inbound Game License) of the Contribution and the Game Rules will be amended such that any code based on the work submitted by that Contributor is required to attribute the Contributor thus creating an indication of the heritage of the code back to the original Contributor.