Need for billions of ethical cases and examples for Safe Artificial Intelligence simulation and training

Ray Kurzweil has spoken about how current Deep Learning systems needs billions of examples to get trained to human level of capabilities or beyond. The Deep learning system AlphaGop Zero taught itself to play the game of Go better than any human or computer program before it by can create billions of examples by simulating the game.

This would seem to be something needed for making or proving that there is safe AI.

Deep learning AI used the many photos on the internet to self-train to categorize which were pictures cats or other objects.

There is a deep-learning training data set of cat and dog pictures.

There need to be data sets of ethical versus unethical and law-abiding versus criminal. There needs to be jurisdiction-specific datasets and culture specific data sets and some attempts at universal data sets.

The history of legal cases can be used in the simulation and to seed examples.

Legal systems in countries around the world generally fall into one of two main categories: common law systems and civil law systems. There are roughly 150 countries that have what can be described as primarily civil law systems, whereas there are about 80 common law countries. The main difference between the two systems is that in common law countries, case law — in the form of published judicial opinions — is of primary importance, whereas in civil law systems, codified statutes predominate. Many systems are also a hybrid of common law and civil law.

Having many ethical and legal examples would allow measurement against people and allow for automated generation of more situations.

Ultimately there would need to be AI components for ethical choices and ethical choice competitions. These components would be available on AI marketplaces where AI components are made available for including in larger AI systems.

Singularitynet and other companies are creating AI marketplaces.

Experts estimate that AI market will increase from the $200 billion it is valued today to $3.1 trillion by 2025.

SingularityNET is a protocol for coordinating and transacting AI services.

Currently AI is a decentralized market. The SingularityNET protocol enables a global AI marketplace. As it matures, it will contribute to a decentralized, market-based artificial general intelligence for the benefit of all.

AI service marketplace would make the task of creating Sophia level AI far easier

Sophia the robot is a case of an AI. She is the most advanced robot of her kind made up of dozens of algorithms. All algorithms are coordinated. It took 30-40 different AI developers to create Sophia.

Sophia uses dozens of algorithms, many of which are licensed from third parties. Today, integrating extern AI resources and functionality is slow, costly, and simply too difficult for most organizations. AI developers, including the hundreds that we work with, are demanding better ways to access and integrate AI technologies. SingularityNET is designed to meet that demand, providing the only protocol for AI to AI communication, transaction, and market discovery.

Soon, Sophia’s entire mind will live on the network, letting her learn from every other AI in the SingularityNET. And the whole world will be able to talk to her.

Protocols and Interfaces to enable AI modules to work together

If there were protocols and standardized interactions then AI services could be added via a plug and play model. This would be like internet services or programming objects.

If the standards interfaces were created and adopted then the barrier to entry in Artificial intelligence would drop.

AI services would become modules.

Standardize the data passing.
Inside processing remains hidden.

Image recognition modules could be combined with risk analysis.

Many standard components could be used and an AI developer or smaller team of AI developers would only focus on improving one module.

AI Marketplace Whitepaper

There are several aspects to this
* Language for the APIs.
* Reputation system on the blockchain. This would help provide feedback on the quality of processing and honesty, verification of completion. Arbitor of truth.
* Create a settlement layer using tokens

There can also be a staking mechanism. Voting with their cash.

There is a 51 page whitepaper. SingularityNET: A decentralized, open market and inter-network for AIs, Ben Goertzel, Simone Giacomelli, David Hanson, Cassio Pennachin, Marco Argentieri and the SingularityNET team, November 16, 2017

The value and power of Artificial Intelligence is growing dramatically every year, and will soon dominate the internet – and the economy as a whole. However, AI tools today are fragmented by a closed development environment; most are developed by one company to perform one task, and there is no way to plug two tools together. SingularityNET aims to become the key protocol for networking AI and machine learning tools to form a coordinated Artificial General Intelligence.

Open AI funded for a billion dollars seeking Safe AI

OpenAI was funded for a billion dollars by Elon Musk to try to solve safe AI.

There are many dozens of research projects on AI safety and AI ethics.

Vincent Conitzer, a Professor of Computer Science at Duke University, had received some funding to work these problems. He had people make ethical choices in order to find patterns and then figuring out how that can be translated into an artificial intelligence. They actually had people make ethical decisions, or state what decision they would make in a given situation and then use machine learning to try to identify what the general pattern is and determine the extent that we could reproduce that kind of decisions.

In short, the team is trying to find the patterns in our moral choices and translate this pattern into AI systems. Conitzer notes that, on a basic level, it’s all about making predictions regarding what a human would do in a given situation. They want the system to become very good at predicting what kind of decisions people make in these kinds of ethical circumstances and then make those decisions ourselves in the form of the computer program.