What is the biggest problem with the AI? The answer is trust. The AI requires systems. That drives information into the system so that the AI can analyze information. Without information AI makes nothing. Another thing is that the AI collects information from very large areas. The AI can give trusted answers only if somebody can confirm that data. The AI follows the same rules as other systems. There must be error detection and correction algorithms.
If we want to use AI, we must know the topics, if we want to use that system effectively. The AI itself is well-protected, but manipulating answers can happen by manipulating sources. Sometimes the AI gives wrong answers because it selects the wrong homepages.
When AI generates answers to questions, it uses the X number of homepages, and then it connects data from those homepages or other databases. The problem is that the AI doesn't think. And if there is some kind of false information. The AI will not realize that thing.
The big problem for the AI, and especially, large language models, LLMs are these: they require precise and well-written orders to make answers. In the new models to use the AI, LLMs are the system used to generate limited, more effective, precise AIs that operate as the RISC models. The RISC is like a pocket calculator. It has a limited operational sector, but it handles those operations very fast.
It's an algorithm that controls some machines and communicates with larger-scale An idea is this: The LLM operator describes the need that the new AI must make to LLM that generates the code.
One example of this kind of AI model is the DNA analyzing AI. These kinds of systems are meant for one purpose. They might be complicated and involve many modules. But they just analyze the DNA and then compile the results with people with a certain detail in their phenotype.
The system can cut the DNA into multiple analyzers. Then each one of those analyzers operates with a certain DNA bite. The system operates like a virtual quantum computer. It drives DNA through the laser spectrometer, which opens the DNA chemical code by reading its base-pair orders.
The laser system can read the DNA and the laser spectrometer can tell if there is thymine, adenine, guanine, or cytosine that the laser illuminates. Maybe those kinds of systems are effective. But the thing is that those systems are tools that highly trained professionals use. Those people know how to ask things from the AI. But the main thing is this. The AI that makes DNA analysis makes only one thing.
It illuminates the DNA with a laser and reads the data using microscopes and spectrometers. The thing that searches for things like hereditary diseases is another AI. The idea is that the system uses RISC processors. And DNA analyzers as groups. The system searches for similarities in the DNA of people, who have a certain ability.
https://bigthink.com/the-present/when-to-trust-ai/
Comments
Post a Comment