iwca-logo
Cellar Door Visits

Visits are by prior appointment on weekdays only. Please click here for bookings and information
Menu
 
Winery Solar

On AI

AI, as it is called, is a synonym for ‘large language models’, which is the crude idea that if you get a computer to read enough documents and books, regardless of what they are, it can create a statistical model of the thing we call knowledge and regurgitate it on demand. The flaw is self evident: large language models cannot separate the wheat from the chaff: true knowledge from random, often corrupted, ramblings. Even 10 or 20 years ago, this would not have been that important as the sum record of ‘truth’ far exceeded that of ‘delusion’. That was before Trump, Fox News, and the host of other disciples of ‘alternative truth’. Today, delusion, or, worse: deliberate contamination of reality, has reached a point where pollution of a LLM is inevitable. In fact, there are active campaigns to create enough of this spam to deliberately skew LLM’s to believe Far Right mythology, or indeed Far Left mythology, given that the Russians, Chinese, North Koreans etc.,  are furiously getting onto this as a priority mission. 

The current vision of LLM was built on the naive idea, that what was out there on the internet was broadly true. Clearly those behind it hadn’t ever watched ‘Love Island’. Creating LLM is rather like creating a mineral water company. You need volume of product but no pollution. At small scale this isn’t so hard but you can’t build without taking pollution on board. 

Ted Chiang: a writer and thinker of remarkable insight and something of an expert on AI, has made the comparison of Tom Hanks in the film ‘Cast Away’ finding, then befriending, a volleyball, inevitably called Wilson. He finds solace in talking to this ball in his loneliness. Ted makes the point, (and I paraphrase) “It sounds clever, it sounds plausible, it even sounds like it understands you… but in the end, you’re talking to a volleyball.”

What to do? LLM’s, as I prefer to name them over the term AI, can be both useful and constructive. But they fundamentally corrupt the idea of truth. We can’t control that. It has been said that in the 3rd Millennium, water will be the most precious resource. Perhaps it will be truth. Either way, you can’t blame AI: you’re blaming a volleyball. 

Cheers, Nigel

Winemaker's comment
Nigel's musings
Sustainability
Blair Walter
Nigel Greening
Sustainability
Nigel's
Winemaker's

On AI

AI, as it is called, is a synonym for ‘large language models’, which is the crude idea that if you get a computer to read enough documents and books, regardless of what they are, it can create a statistical model of the thing we call knowledge and regurgitate it on demand. The flaw is self evident: large language models cannot separate the wheat from the chaff: true knowledge from random, often corrupted, ramblings. Even 10 or 20 years ago, this would not have been that important as the sum record of ‘truth’ far exceeded that of ‘delusion’. That was before Trump, Fox News, and the host of other disciples of ‘alternative truth’. Today, delusion, or, worse: deliberate contamination of reality, has reached a point where pollution of a LLM is inevitable. In fact, there are active campaigns to create enough of this spam to deliberately skew LLM’s to believe Far Right mythology, or indeed Far Left mythology, given that the Russians, Chinese, North Koreans etc.,  are furiously getting onto this as a priority mission. 

The current vision of LLM was built on the naive idea, that what was out there on the internet was broadly true. Clearly those behind it hadn’t ever watched ‘Love Island’. Creating LLM is rather like creating a mineral water company. You need volume of product but no pollution. At small scale this isn’t so hard but you can’t build without taking pollution on board. 

Ted Chiang: a writer and thinker of remarkable insight and something of an expert on AI, has made the comparison of Tom Hanks in the film ‘Cast Away’ finding, then befriending, a volleyball, inevitably called Wilson. He finds solace in talking to this ball in his loneliness. Ted makes the point, (and I paraphrase) “It sounds clever, it sounds plausible, it even sounds like it understands you… but in the end, you’re talking to a volleyball.”

What to do? LLM’s, as I prefer to name them over the term AI, can be both useful and constructive. But they fundamentally corrupt the idea of truth. We can’t control that. It has been said that in the 3rd Millennium, water will be the most precious resource. Perhaps it will be truth. Either way, you can’t blame AI: you’re blaming a volleyball. 

Cheers, Nigel