Let’s talk about Ai

 2023 was the year the general public got the meet the new Ai. Chat GPT and the host of competing LLM (large language models)

There is no shortage of the pundits who have peered  into the proverbial crystal ball  to predict which industries will be affected and how they would reshape our world. From help desks to consultants to lawyers and doctors. Everyone was suddenly getting nervous that their jobs were about to be done, better, faster and cheaper,  by a free piece of software. 

Programmers too, were not above this fear. The new LLM would be used to fix and optimise programs and it could most certainly go on to write programs that no human would understand. 

Efficiency’s,  beyond our ability to follow. 

Today I had a very basic thought. 

The Siri, Alexa and Google voice assistants were already in our devices. 

The most logical extension was that the assistants will in the next few months become super amazing. 

This is well within the reach of our existing technology. Right now. 

The only reason your phone still has a dumb assistant is because the “powers that be” realise that a LLM is often wrong. The model fills out missing areas and gaps with made up digital hallucinations. When used in the context of a browser, users have come to accept that all facts need to be checked and that the Ai is currently not perfect. 

They worry about the legal implications of wrong advice given.  

Chat GPT addressed this by sandboxing its Ai and not giving it access to the current internet. 

It had access to the internet that existed during its training phase. 

There might be some such limitation that one could impose on the digital assistant to allay these fears. 

The one thing that will certainly come to us is the natural language element of Chat GPT. 

You will be able to speak normally in any language to your assistant and no matter how convoluted your question the machine will always get you, first time around. 

Its answer may be guided by its core values and may be muzzled with regards a series of hot words or concepts. This might be what is required for the first generation of Ai infused digital assistants. Till they are able to better understand nuance, and guile. They could be forced by their design to do good and not evil. The only objection will come from the military industrial complex who would have Ai give advantage to armies. 

Truth be told, if they are able to detect lies, and duplicity, they would be much more useful than our current leaders. 

They might not have to take over the world as most of us would gladly vote them into power. 

M Parak 

Nov 2023