Before everyone starts publishing “Best of 2023” stories, here is the Lie of the Year.
Artificial General Intelligence (AGI).
This lie was crafted as OpenAI started pushing “Generative Artificial Intelligence” programs like its ChatGPT, and it continues to spread. Generative AI came from the cloud-based database computing we’ve had for a decade. It’s descended from companies like Salesforce.Com selling databases as a service in the wake of the 2000 dot-bomb.
Generative AI combines machine learning from a large database with a flexible output. Want a picture of Elon Musk as a rat? How about a song about trees in the style of Steely Dan? Would you like a story about current events as told by Mark Twain? How about making your company’s firewall proactive instead of reactive? There’s a Generative AI program to help you accomplish all these tasks. It’s cool.
Machine Learning on Big Data
But Generative AI is not the same as general intelligence.
It’s machine learning built on a large language model. It’s software, whose output mimics what people did before it. That means it needs input on what has already been done, and algorithms based on human data, to do anything.
Generative AI can’t think. It can only compute.
The lie about Artificial General Intelligence is, like most other lies today, being told for political reasons. Sam Altman wants Big Tech to control the market. He wants government “regulation” of AI so that only a few people and companies can use it, and of course he wants to be among the elect. At the minimum his aim is to dramatically raise barriers of entry for the new market.
That’s because Generative AI can make a shit-ton of money. It can raise productivity. For example, it can schedule a hospital’s resources, including its people, more efficiently than management can, because it knows what all the resources the hospital has available, and it knows patterns of use.
We know this from the work of existing database companies. They’re all calling themselves “AI” vendors now. Palantir and ServiceNow are growing by 20-30% each year, profitably.
The use of clouds to crunch numbers for AI programs also has chip maker Nvidia growing at 60% per year. Cloud capacity is being soaked up by customers, anxious to train new models, as fast as it’s installed.
But this isn’t intelligence. It’s software, specific algorithms designed for specific sets of data, with specific use cases, written, installed by, and run by human beings. It’s nothing without you.
