IOSG Ventures
Published in

IOSG Ventures

I, Robot (2004) : Crypto version

What is the Turing Test?

Let’s assume a simple experiment: you are in a closed room with a small hole in it, and a mysterious visitor in another room connected by the small hole. By your side, there is a pile of paper and a pen. You can write down any question on the note, stuff it through a small hole and the other person will hand back the note with the answer written on it.

Then the question arises, how can you tell whether the person is a real person or a computer by means of paper and pen? This is the classic (and simplified) “Turing Test”, also known as the “The Imitation Game”. Turing believed that a computer could be described as intelligent if it was able to impersonate a human being and make the questioner believe that he or she was actually a human being. As early as four hundred years ago, Descartes, the proponent of dualism, had also suggested that machines could in fact interact with humans. And the philosopher Diderot even remarked, slightly tongue-in-cheek, that “if they found a parrot that could answer all questions, I would not hesitate to declare that it has intelligence”.

The Turing Test was considered by many to be a fantasy for half a century. Even today, most people would still laugh at Siri for being too rigid, as if AI that could pass for the real thing was still a long way off.

AI development in the gaming sector

As a veteran gamer, my first contact with AI in a game was in StarCraft. Twenty years ago, the bandwidth resources were very limited and real-time online battles were unattainable. Starcraft, as a strategy game, developed the earliest AI for the player. I remember clearly the time when the AI was so clumsy and slow to react to the player’s strategy that the game company had to give the AI extra in-game resources to make them an even match. No one fails to tell if their opponent is an AI or a real person, because they are just too dumb. Interestingly, despite the fact that Blizzard (the company behind Starcraft) has since made a fortune on Starcraft and Warcraft, and Blizzard has continued to pour profits into R&D with funding that would make even many academic institutions envious, the AI development of Blizzard’s games has been very slow and stagnant.

Why is the development of AI under the big tech companies so slow? Apart from the theoretical bottleneck of the hardware, we also should talk about the business model of Starcraft. In the early gaming market, due to the lack of online payments, the vast majority of gaming products were profitable through offline buyouts. Since it was a one-shot deal, game companies naturally lacked the economic incentives to optimize the gaming experience in the long term.

The advent of the mobile era has significantly accelerated the development of gaming AI. Thanks to the widespread use of online payments, mobile games have tapped into a more efficient business model — the in-app purchase system. In games like Arena Valor or Game for Peace, game companies can bury payment points at many nodes in the user lifecycle. The one-shot deal becomes a long-term repurchase, and the game’s DAU (Daily active users) becomes the most crucial business metric. In order to improve player retention, game companies need to find ways to create mind-flow and excitement for players. Specifically, game companies need to interfere with and control the user’s gameplay. In the case of Arena Valor, for example, each player’s first game is critical, and if a player fails in the first game, there is a higher probability of losing that player outright. As a result, game companies need to deploy carefully designed bots for the first game, or even the first few dozen games, so that players win more than they lose, and are always able to meet comparable ‘opponents’ and win by neck in thrilling encounters.

What converges with the new business model is the rapid development of hardware and machine learning theory. The creation of Tensorflow has made it possible for even the most mundane programmer to efficiently train AI. with dramatically reduced development costs and exponentially increased benefits, the invisible hand has driven the development of AI to new heights. To give an example from my side, my wife rarely engages in gaming, but I recently took her through Arena Valor or Game for Peace. She quickly caught on and even became somewhat addicted. As a diehard gamer and gaming investor, I was naturally aware that most of our opponents were bots, but she was oblivious. To give an example from my side, my wife rarely engages in gaming, but I recently took her through Arena Valor or Game for Peace. She quickly caught on and even became somewhat addicted. As a diehard gamer and gaming investor, I was naturally aware that most of our opponents were bots, but she was oblivious. It probably didn’t occur to Mao that “it’s fun to fight with (a) robot” as well.

Crypto is the golden age for AI

During the last ten years of mobile gaming development, we witness how in-game economic exuberance impact on AI. When it comes to crypto, AI will definitely be growing in an exponential way.

A few points will be discussed as below:

The application layer/entity is enormous in scale: for a non-official AI, using a less proper analogy, it is much easier to make profit out from Axie Infinity if comparing to Fortnite

Technically speaking, KYC is classified as off-chain data, which cannot effectively interact with blockchain

The excess computational power due to the transformation from PoW to PoS will lower the bar of AI development

As I mentioned in previous passage, the P2E model represented by Axie Infinity is fundamentally a short-term business strategy to compensate users. The underlying revenue supposedly collected by traditional gaming companies is redistributed to active gamers by Axie Infinity, effectively lowering the threshold to attain new comers. Meanwhile by such mechanism, user stickiness is expected to realise in business valuation within long-term horizon. For instance, when Sky Mavis (team behind Axie) launches new game, users will be more likely to pay for subscription if he/she was previously a loyal gamer of Axie. Even if gamers refused to do so, revenue can be compensated substantially in other ways such as ads and streaming services. However, if it is bots who are playing behind these unique addresses, then the above assumptions are no longer valid. On one hand, bots will not pay any subscription fee. On the other hand, other revenue streams will be net zero. An real example is the recent price crash of SLP, which could be explained by the claim that roughly 30% of the SLP is drained by bots and dumped by then.

From project’s perspective, how to distinguish bots from real users is always tricky. Axie Infinity is the first unicorn dApp, but will not be the last one, that has to put real thoughts on Turing test, if reflecting on latest game named Sunflower Farmers which is destroyed holistically by bots. In web2, such moat is established by ID card and Bank account which are guaranteed by national credit system, preventing AI from largely involving in economic activity. Not to even mention that the generalised AI has been rapidly replacing human beings. For instance, quant bots have made numerous traders redundant. The absence of KYC in crypto makes it even easier for AI to claim their territory.

The thing can even get worse. When Ethereum transitioned to PoS, a large amount of GPU will be idle. A proportion of it will flow back to individual computer, while another proportion will be consumed by bots, through Tensorflow or RandomForest.

Defend humans, starting from Proof of Personhood

The rapid growth of AI will put most of the xx-to-earn model in threat. From a crypto gaming investor’s perspective, we keep a close watch on DID and POP protocols. DID will not be addressed in details in this article given it is not a new idea anymore, As below I will dive deeper into the POP development.

Before we understand POP, we need to reach a consensus on why Poof plays a huge part here. First of all, within any distributed network, Sybil Attack is disastrous — malicious act targeted towards distributed network, by attaining fake and ineffective accounts. A quick imagination can play out — if Bitcoin network no longer requires hash computation, and thus any node can participate in block generation, which makes hackers can easily run hundreds of thousand of nodes with little cost and in turn attain control of the whole network. We can tell from the above example that for either PoW of Bitcoin network or PoS of Ethereum 2.0, only one goal is deemed to meet — raise the cost of Sybil Attack and effectively resolved the problem of bots and any other problems associated with manipulation of fake accounts.

However, POW and POS both brought up a serious side effect, which is plutocracy. Rich people has much more control over distributed network comparing to average people, which contradicts to the narrative of web3. web3 intends to strive for equality and inclusiveness, while the situation for now is another story.

Maybe Proof of Personhood is the best answer to the prevention of Sybil Attack and promotion of average people. POP derives from the distributed network, built on biological definition. Two criteria has to be met for a POP network — 1. only human can attain identity and become a node operator 2. each one of human can only claim one identity. When the above two criteria are met, a native web3 Dapp can be successful globally.

POP is still developing at nascent stage. Majority of people don’t smell the urgency yet while builders have been committing in establishing POP network already. Except Proof of Humanity which is raised by Vitalik lately, there are many other excellent projects out there leaving their own mark down the road, such as Idena Network, Humanity DAO, Kleros, Upala, BrightID, Duniter, Equality Protocol, etc…

It becomes harder and harder to establish an effective POP network, considering the fast-paced development of AI technology. Here we go again, facing same problem raised by Turing half an century ago — how to see through the real identity via screen. Without any doubt the above POP protocols mentioned adopt unique designs and technologies, while they all encounter the trilemma of decentralisation, privacy, and scalability.

Congratulations if you make this far. To make people’s life easier, more is expected to delve and explore in the next research paper.

Like, Subscript and turn on your notifications!

🦄 About IOSG

IOSG Ventures, founded in 2017, is a community-friendly and research-driven early-stage venture firm with offices in China and Singapore. We focus on open finance, Web 3.0 and infrastructure for a decentralized economy. Our portfolio covers more than 60 projects, including Layer 1 (NEAR, Polkadot, Cosmos), DeFi (1inch, Synthetix, UMA, Dodo, Liquity, Gelato). We commit ourselves to work alongside various developer & DAO communities and helping the most aspiring founding teams to achieve success. As a developer-friendly fund with long-term values, we have launched Kickstarter Program which offers capital and resources for innovative and courageous developers. Since we consistently cooperate with our partners and connect with communities, we work closely with our portfolio projects throughout their journey of entrepreneurship.

If you would like IOSG Ventures to consider your project, please send a summary of your project along with a pitch deck and/or white paper to

Follow us on Twitter.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


Community & Thesis Driven Investing iosg.eth