First published in Mar 2024 on LI

There is so much talks of Intelligence, AGI, the “AI danger” and even the plethora of Startup ideas driven by some LLM, that I am going to share few lines and my take on this topic from my draft article on my blog.

First, Every LLM Evaluation, Every new LLM will be called as AGI, by the tech bros, if you don’t bother looking at your training data.

Second, heading into a Dystopian World (as one of my Long Thesis) has very less to do with “Age of Intelligent Sentient Machines” but all to do with people’s inability to differentiate between Language and Intelligence.

Language is a key component of Intelligence. But Language <> Intelligence.

A small rodent is an Intelligent creature. It knows where are the dangers lurking for them, what is good for their survival and what is not.

The ability to reason out signals around you, deduce them and make a cogent conclusion of its impact on one’s welfare is actually a definition of Intelligence and mental acquity.

But does a Rodent have the faculty of Language? If not, then how is a Rodent or any creature around us which can interpret the signals and make sense of things around them is Intelligent?

Or how are people who are born with an unfortunate disability of being deaf and dumb still function decently well and survive in the world if there’s no intelligence in them as Language clearly became a barrier for them.

The issue is not Language.

Issue is our perception that Language leads to Intelligence. Or in some extreme cases Language Equates Intelligence.

You will witness a world soon where young people will be dating their AI companions. People will become friends with their AI Companion, fall in love with them. Some will even go to the extent of cheating by dating other people’s AI Agents.

Human Psycho Therapists will get overtaken by AI Therapists. A 24x7 always available “friend” who understands you, empathizes with you and speaks to you as you like it to.

And this “dystopian society” is possible because Language creates a facade of Intelligence to us humans. There is a reason why a Persuasive Speech moves and inspires more people around the world than a less confident one. Even if the former is all lies and mistruths while the latter is actually speaking facts.

Because a Confident Persuasive Language creates a veneer of “smartness” over a less persuasive one.

This nuance misunderstanding leads to the next step of errors where we are now seeing people looking to use LLMs to do tasks that requires Reasoning or Planning.

And that’s in my opinion, lies the fundamental mistake of building products with LLM.

Use LLMs as great Approximate Retrieval Systems to build a better User Experience layer than actual Intelligence Layer. Your actual Intelligence will be better served by traditional ML Systems. LLMs take it a notch above by enhancing a great UX layer than possible earlier.

← Things that Enrich Life The Cataclysm Upheaval →