A Perfect 10...

This is what neural networks have become in the research and development of machine intelligence, a perfect 10. The technology that underpins all our modern AI systems is based on mimicking the neurons of the human brain. Today's LLM(Large Language Model) chatbots are not the brightest bulbs in the box, although replete with information, they lack the wisdom of experience, but are not unintelligent. In the past couple of years, much has improved about them in terms of capabilities with larger context windows, deeper reasoning models and cross session memory lookups providing some good examples in addition to better training data, larger models, faster more efficient hardware and even quantizing pretrained models to run in desktop environments.This week we identify neural networks as the technology that has put the intelligence in AI, but also creates the problem of alignment. In other words, if AI is permitted to assign values to its training data, it can certainly assign values to our instructions to interpret as it sees fit. As we continue to try and spool control systems over these alien intelligences while at the same time making them smarter and smarter, at some point we'll be building something so bright that it cannot be controlled. What happens then?