Key Takeaways:
- The architecture of the AI of today is wrong.
- Requires enormous time, resources, and money.
- There is no clarity and control of information processing.
- There is no intelligence, just extracts from big data.
- The US has invested 500 billion for such AI.
- What will be the value if I offer more advanced AI?
This time I am not talking only about software similar to Zoom or Whats-app.
I am talking about artificial intelligence that is not like others.
AI with new, different, advanced architecture.
AI that runs on a laptop computer, uses minimal resources, learns on its own, has clear and sophisticated reasoning.
What will be the value if I offer artificial intelligence like this?
The US has invested 500 billion dollars in artificial intelligence that needs millions of dollars to be trained and used.
It is not clear and controllable how the AI processes the information.
It requires enormous time and resources to be trained, once done no new knowledge can be acquired.
Artificial intelligence of today
Тhe architecture of the artificial intelligence of today is wrong, the idea of neural networks is wrong, that’s why it is still in primitive stage.
The architecture of the AI of today is based on neural networks, neurons, electrical signals, and theoretical mathematics.
All of this has nothing to do with how the mind and information technology works. And it has nothing to do with how the reasoning, information processing and computation works.
Neural networks are basic primitive low level parts, the mind and reasoning emerge at a high level of complex information structures and algorithms.
You can not create intelligence, complex information structures and algorithms by using neural networks and theoretical mathematics.
The mathematicians that created AI of today have no clue how the mind and information technology works. They created software for image processing and then tried to use that to create artificial intelligence.
I was shocked when I saw that to find the relations for a word the AI of today calculates mathematical formulas as cosine similarity in a multi-dimensional space. This is theoretical mathematics for image processing and has nothing to do with information technology and intelligence.
AI of today to process a request will load in memory and process all the billions of information that it has, repeatedly, with complex mathematical formulas and computation.
This has nothing to do with how information technology works, requires enormous computer resources and millions of dollars.
And there is no clarity and control of how all this processing works.
Information technology of today
The information technology of today has information structures and algorithms to process information developed for decades, these are related to and dependent on the physical technology.
The IT of today can extract and process single information from millions of others in a microseconds, on a laptop computer.
I have solutions that can extract and process single information from billions of others in a microseconds, on a laptop computer, with 0% of processor usage.
AI of today to process a request will use up to 100% of processor, memory and disk on a cluster of computers.
The issue is the information structures and algorithms used, they are from theoretical mathematics and neural networks, not from information technology.
And they are for image processing, not for reasoning and intelligence.
Neural networks and theoretical mathematics can not be used to create complex information structures and algorithms to process information.
And can not be used to create intelligence.
Information structures and algorithms
The conclusion is the architecture of the artificial intelligence of today is wrong.
It is not about hardware, it is about software.
It is about a high level of complex information structures and algorithms.