Rise of AI
Artificial Intelligence (AI) has been around for a while, it was formally founded in 1956 at a conference at Dartmouth College, in Hanover, New Hampshire, where the term “artificial intelligence” was coined.
However, while the basic proof of principle was there, there was still a long way to go before the end goals of natural language processing, abstract thinking, and self-recognition could be achieved.
Graph below demonstrates the ups and down in the journey of AI becoming mainstream so far.
We now live in the age of “big data,” an age in which we have the capacity to collect huge sums of information too cumbersome for a person to process. The application of artificial intelligence in this regard has already been quite fruitful in several industries.
Recent advances in AI have driven an explosion of intelligent applications that will dramatically change the way we live.
Applications can be found in every vertical and functional area, from manufacturing to HR.
For example, manufacturers are using deep neural networks to quickly identify manufacturing flaws, far surpassing the speed and accuracy of their existing techniques, and HR professionals are using AI to help them sift through thousands of resumes to build a short list of candidates efficiently.
Figure below demonstrates the vast array of enterprise companies that use AI in their product or service.
McKinsey estimates[1] that the AI techniques have the potential to create between $3.5 trillion and $5.8 trillion in value annually across nine business functions in 19 industries. This constitutes about 40 percent of the overall $9.5 trillion to $15.4 trillion annual impact that could potentially be enabled by all analytical techniques.
AI is centralized
As the computing power and data storage capacity increases day by day, community as a whole is making very significant advances in the field of AI but as it takes a lot of resources to train models and run them, accessing and taking advantage of the machine learning systems making these developments possible can be challenging, especially for those with limited resources.
Current AI systems are heavily centralized[2], their predictions are often sold on a per-query basis, and the datasets required to train them are generally proprietary and expensive to create on their own. Additionally, published models run the risk of becoming outdated if new data isn’t regularly provided to retrain them.
Issues
AI being centralized is an issue on two fronts. First is ease of access, it’s not available to the masses.
Getting access to data and compute power is easy for big corporations but it’s not that easy for regular data scientists or any developer who is working on to create small AI applications.
On the other hand, since AI has proliferated almost all the industries, almost everyone doing business now should and needs to use some form of AI, but it’s not easily available for smaller firms, but bigger firms such as Google and Apples of the world have no problem developing their own AI tools in house. There are independent developers but how do we connect independent AI tools developers to small businesses?
Second front is development of AGI. AI is not just Artificial Intelligence, there are three[3] broad forms of AI.
- Artificial Narrow intelligence (ANI)
It is a stream of intelligence which is prominent in performing a single task with smartness. The intelligence learns about a single task which it has to perform efficiently and with smartness(intelligently). It is considered to be a basic concept of AI. This is the only type out of the three that is currently around.
Examples include, Speech recognition, Voice assistants (Siri, Cortana, Etc.)
- Artificial General Intelligence (AGI)
As the name suggests, it is general-purpose. Its smartness/efficiency could be applied to do various tasks as well as learn and improve itself. It is comparatively as intelligent as the human brain. Unlike ANI, it can learn and improve itself to perform various tasks.
We have not achieved this level of Artificial Intelligence yet.
- Artificial Super Intelligence (ASI)
Artificial Super Intelligence is an aspect of intelligence which is more powerful and sophisticated than a human’s intelligence. Human’s intelligence is considered to be one of the most capable and developmental. Superintelligence can surpass human intelligence; it can think about abstractions which are IMPOSSIBLE for humans to think. The human brain is made of neurons and is thereby constrained to some billion neurons.
Currently we are working with ANI and there is not a very clear path from ANI to AGI.
How can we solve these two issues?
Decentralized AI
What if we combine AI and Blockchain and make AI decentralized?
Blockchain is important here because it offers participants a level of trust, security and reliability without involvement of any third party.
AI models can be hosted on blockchain and a smart contract can be placed for providing the data. Smart contracts are important here because they are unmodifiable and evaluated by many machines, helping to ensure the model does what it specifies it will do. The immutable nature and permanent record of smart contracts also allows to reliably compute and deliver rewards for good data contributions.
This way we can use power of decentralization to train models and users can get rewarded for their data contribution. Open Mind Architecture[4] works on similar concepts.
Using decentralized data in this fashion will essentially solve the first half of the first problem that we talked about in AI being centralized.
But how do we increase the reach of AI, so that its easily available to small businesses and how do we address the issue regarding making a forward progress from ANI to AGI? One school of thought is to create a marketplace of AI where anyone in the world can launch their AI tools on the networks and it’s all supported by smart contracts. We can envision an open community where independent developers can get connected with paying users without the need of third party. Blockchain and smart contracts will bring in the element of reliability and trust.
On top of this if AI agents can interoperate or cooperate with each other based on task then it can be a step towards AGI. Currently there is no clear path for AGI but if different AI tools start to interoperate and start to share their compute power, data and results, this can be a step in the right direction. Singularitynet[5] is based on similar principals.
[1] https://www.mckinsey.com/featured-insights/artificial-intelligence/notes-from-the-ai-frontier-applications-and-value-of-deep-learning#part3
[2] https://medium.com/@kalehungerson/is-a-centralized-ai-dangerous-5872293056d9
[3] https://medium.com/predict/types-of-artificial-intelligence-and-examples-4f586489c5de
Strategy in the Age of Decentralization 