Decentralized AI Marketplace

Powered by Ethereum

Join our mailing list for updates

Team

Archil Cheishvili

Chief Executive Officer

Economics, Harvard University. Former CEO of AI startup Palatine Analytics. Bridgewater Associates. Recognized by media outlets such as the New York Post and Yahoo Finance.

David Fan

Chief AI Officer

Applied Mathematics in Computer Science, Harvard University. Software Engineer work experience at Google, APT, and Dataminr.

Mena Gadalla

Chief Scientist

PhD in Applied Physics, MS in Computational Science and Engineering both from Harvard John A. Paulson School of Engineering and Applied Sciences. Published multiple scientific papers and raised over $1 million in research grants.

Alexander Parunov

Senior Software Engineer

Mathematics and Computer Science, American University of Bulgaria. Developed and co-founded AI startup Palatine Analytics.

Elena McCormick

Senior Business Development Manager

Human Evolutionary Biology, Harvard University.

Giovanni Cardamone

Chief Business Officer

Graduate studies at Harvard Business School, Master of Science in Engineering from Columbia University.

Badar Dar

Vice President of Marketing

Graduate studies at Harvard Business School, undergraduate studies in engineering.

Advisors

Professor Thomas Magnanti

Former Dean of Engineering, MIT
Institute Professor, MIT

Former Dean of Engineering at MIT. Institute Professor. Founding Director of the Singapore-MIT Alliance for Research and Technology (SMART).

Neil Flanzraich

Lead Independent Director, Chipotle
Former President of Ivax Corporation

Lead Independent Director of Chipotle. Former President of Ivax Corporation (acq. by Teva for $10B). Executive Committee member of Syntex Corporation (sold to Roche Holdings for $5.3 B).

Travis May

CEO, Datavant
Founder & CEO, LiveRamp

Former CEO of LiveRamp (acq. by Acxiom for $310 mil). CEO of Datavant. Forbes 30 Under 30.

Professor Tim Kraska

Associate Professor, MIT
AI/ML Expert

Associate Professor Computer Science at MIT CSAIL. 2017 VMware Systems Research Award Recipient. Widely recognized for early work on hybrid human-machine data management.

Professor Minlan Yu

Associate Professor, Harvard
PhD in Computer Science, Princeton

Associate Professor of Computer Science at Harvard University. PhD in Computer Science from Princeton. Experience at Google, AT&T, Microsoft, Facebook, and Bell Labs.

Professor Stratos Idreos

Assistant Professor, Harvard
Leads Data Systems Laboratory, Harvard SEAS

Assistant Professor in Computer Science at Harvard University. Leads the Data Systems Laboratory at Harvard School of Engineering and Applied Sciences.

Ed Simnett

Principal, The Arnold Group
Launched $1B business for Microsoft

Principal at The Arnold Group. Launched a $1B business for Microsoft. Harvard MBA.

Professor Elie Ofek

Professor, HBS
IBM Research

Professor of Business Administration in the Marketing unit at Harvard Business School. Development engineer at IBM Research.

Professor Andy Wu

Assistant Professor, HBS
PhD in Economics, Wharton

Assistant Professor of Business Administration in the Strategy Unit of the Harvard Business School. Founder of Identified Technologies. PhD in Economics from Wharton.

Problem

AI can change our lives rapidly and has an opportunity to become the single most important technology in the world. Experts estimate that the AI market will reach $3 trillion by 2025. However, fundamental obstacles currently hold AI innovation back:

No connectivity

Today, there is no way for AIs to exchange data, learn from each other, leverage their capabilities, and trade services. AIs are operating in a closed environment.

Costly to use

There are only around 10,000 AI developers in the world. 99% of companies cannot afford to hire their own team of AI engineers to create in-house AIs, nor do they have enough technical capabilities to correctly determine from which open-source APIs to grab existing AI code.

No quality measurement

There is no way to judge the quality of AIs because there is no reputation system. Companies operating on a lean budget face high-risk decisions when choosing which AIs to use.

Solution

Introducing GenesisAI

GenesisAI is a smart-contract based protocol powered by Ethereum that is designed to solve these problems and facilitate easy access to AIs, enabling anyone in the world to access AI services.
Large corporations will no longer be the only entities able to develop and utilize AI. We believe in making AI accessible: “for the people, by the people.”

By linking AI services with each other and increasing the supply of AI services, GenesisAI provides a web platform that offers low-cost AI services. This makes AI technology more efficient and affordable for businesses.

Our Ethereum based smart-contracts enable different AIs to communicate with each other, exchange data, and trade services. Anyone can develop and purchase AI services.

Roadmap

August 2018

Pre-Alpha: Preliminary Marketplace Iteration

Finalize a preliminary iteration of the marketplace. Providers post tasks they are able to provide, and anyone who is interested in a particular AI task can request it for the stated price in GAI tokens. The initial service provider code is finished as well. We’ve used gRPC for multi-language algorithm support and performance. Initial development of the website.

September 2018

Alpha on Rinkeby Testnet

We will deploy a preliminary version of the GenesisAI contracts, as well as a frontend for registering services and requesting jobs. This is meant to demonstrate the capabilities of the GenesisAI platform and allow service providers and buyers to give feedback. Small jobs (e.g. text for sentiment analysis) can be created and resolved, but large inputs will not be feasible.

October 2018

O-Chain Metadata Storage

Descriptions, ratings, prices, etc. will be moved out of the Seller Registry. This will enable Service Providers to start registering themselves and their models without much of the cost overhead of on-chain storage.

November 2018

O-Chain Request/Response Storage

We will add support for transferring encrypted model inputs and model responses over IPFS. Past this point, all inputs and responses will be made available through IPFS, with the blockchain only storing minimal metadata and a reference to the location of files on IPFS.

December 2018

Sample Models Running

We will make sample models available. These will be a subset of models currently available in official Tensorflow repository. These will be pre-trained and running on servers owned by GenesisAI.

January 2019

GenesisAI on MainNet

GenesisAI will be fully available on the main Ethereum network. Buyers will be able to upload arbitrary input data. Service providers will be able to earn tokens by fulfilling incoming requests.

GAI Token

Have you announced token generation event/presale ?

No announcement has been made regarding the token generation event. Stay tuned!

Why is your native token necessary?

We are developing a native token based on Ethereum rather than using USD or another fiat currency because of four major reasons:

Means of the transaction is the most important part of the marketplace. Our ideology of creating a decentralized marketplace requires the means of transaction to not be tied to or controlled by any government in the world.


There will be thousands of micro-transactions happening on our network. Using the USD in microtransactions does not make sense because it has fixed costs. Furthermore, international payments in USD are extremely expensive.


Nobody should be excluded from participating in building the AI-to-AI economy just because they might not have easy access to the USD.


Our native tokens will enable cheaper, faster, and more secure transactions.

We decided to develop our native token instead of relying on existing tokens because of three reasons:

First, we need to reward beneficial AI players with our native token. Native tokens are an affordable yet efficient way to reward helpful network players.


Second, we need a token that is optimized for the AI-to-AI economy. There will be thousands of transactions happening during any given time. An optimized native token will help us to scale the marketplace without any problems.


It is our long-term goal to create the first truly autonomous, decentralized organization.

Network participants will make decisions regarding AI-to-AI economy protocols. The amount of voting power of an AI node in governance will be determined by the amount of tokens held by the node, as well as by its reputation.

Simply put, the GAI token is a utility token, that is used as a mean for AI-to-AI transactions. Therefore, our token mainly has a consumptive use. The GAI token can be acquired during the token generation event (details will be announced at a later date).

Our Marketplace

The GenesisAI marketplace will have a built-in reputation system and a matching system. Services ratings will come from two sources: a review system and expert analyses from GenesisAI's tech team. Members of the marketplace will be able to rate a service from 1 to 5: this will be the raw score. Then, an algorithm will optimize this score based on the following elements:
Reviewer reputation: nodes with a demonstrated track record in the service they are reviewing will weigh more than nodes with less experience.
Number of reviews: as more popular services receive more reviews, the algorithm will boost their optimized score.
Timing: older reviews will weigh less than newer ones. This is important to guarantee that services with the highest score are the most relevant today.
Incentives: potential conflicts of interest will be recognized and avoided. For example, if a provider is rating other services that he provides in a different node, that will be considered in the optimized score.
In addition to that, an AI-powered technology will match buyers and sellers. This technology will account for users' past behavior, willingness to pay, and needs. For example, if users recently ordered AI services in a certain vertical, like speech recognition, they will receive suggestions about similar AI services. Moreover, nodes will be able to alter their researches based on their willingness to pay for a particular service, and on how quickly they want to get this service. The initial protocol is focused on providing as much flexibility as possible, with the trade off of adding some developer complexity for service providers. As more services become available, we may define stricter protocols specific to a family of models. The protocol for the response is also meant to be as flexible as possible.

Join our mailing list for updates

GenesisAI 2018. All rights reserved.