Linux enthusiast Linus Thorwalds said that 80% of AI is marketing, and 20% is true. That is the normal relation in computer applications. The news about "doomsday AI": or, AI that destroys the world, makes AI look alike more fascinating than it is. But the fact is that the AI and its ability to generate the code should cause re-estimation in data security.
The neural network can attack systems with powerful methods. In neural networks. The AI can change the attacking computers. And their IP's all the time. The AI-based systems also make it possible to use dynamic IPs that make it harder to track those computers. The large-scale neural networks can also operate as the artificial general intelligence, AGI.
The news about the AI's ability to make code and even essays brings new users to those systems. Also, things like weapons and other kinds of stuff make the AI news interesting. Networks are things that make the AI real. They allow the AI to search data from the net. Without them, the developers must program everything to the computer's memories.
Or it makes it possible to generate a language model that can search data from the network. The data that the AI can use limits its abilities. Another thing that determines AI and its skills is the ability to operate in the physical world. The robot is the tool that turns the real actor who can clean the house and make food.
Above:) Neural network.
The fact is this. If we talk about things like AI doctors or AIs that make doctoral theses, we are far away from those things. There have always been people who cheat in universities. And the AI is only one tool for that thing. AI is the tool that makes many things more effective. But it requires that the user knows about the topics. The AI is an ultimate tool for people, who know what they do.
The AI is a tool that can make many things easier. It can handle well-sorted and well-presented information like numeric calculations with very high accuracy. In a neural network, the AI can swap the calculation to another computer if that computer is required for some other mission. When a user takes the workstation in use, the system can swap the ghost or background process to another workstation.
In that kind of network-based system, the system is based on the network of workstations. In the network-based solution, the single workstation can try to solve problems in a certain time. If the problem is not solved the system calls other computers to help. In this case, the system can scale the job theoretically unlimitedly. The only thing that limits the number of computers that the AI can use is the number of workstations and supercomputers in that network.
But the AI cannot make things spontaneously. It always requires humans to begin the operation.
Social media makes the AI more well-known. But AI is also a tool that requires social media. The AI companies use data from internet services like X, and Facebook to develop and train the AI. When R&D people make some new solutions for AI they require words that activate the process. In cases of the AI the word that people say or write causes a process where the AI searches for match. The word that the AI gets is the trigger that activates the process.
The AI is still a regular computer program. Or, actually, it's a group of computer programs and applications. The marketing people call a language model that can connect itself to datasets as AI. Each dataset gives some "skill" to AI. The artificial general intelligence AGI is a large number of datasets that it can use to control things like robot taxis and microwave ovens. Otherwise, we can see a large group of dataset modules as one entirety.
Mesh protocol.
That means when the LLM calls the robot taxi, it just makes a "phone call" to the LLM that operates the taxi. Then the customer's LLM fills the form. And then it delivers responsibility to the AI that controls the taxi. The appearance of the AGI is formed when the cab transforms the voice it uses with the customer similar to the customer's LLM. There are two independently-operating AI-based systems run on two servers. The network-based structure that swaps missions between systems makes it possible to create a system that seems to be one entirety.
In this case, the AGI is the network of independent operating systems that humans can control through the LLM. The network-based systems that use open architecture, or partially connected mesh protocols allow to connection unlimited number of modules to that entirety. And that brings AGI closer than we think.
These kinds of network-based systems can have billions of full mesh networks and in the middle of those systems is the LLM. So each ball in the neural network involves LLM and the mesh network. Those systems can also have independent operating supercomputers.
The thing is that the network of AI-based systems that connect robots with AI and LLM are impressive tools. The fact is that those robots are good marketing tools. People want to see robots in action. And that brings a new audience to the AI.
Comments
Post a Comment