November 13, 2023

OpenAI's goal for ChatGPT is to have it become a super intelligence. Or, rather, the Superintelligence.

What is that? The silly notion of Sam Altman (founder of OpenAI) that computers can be as smart as humans if you feed the large language models enough of the world's data.

What would you do with a Superintelligence? Put people out of work, according to Musk and Altman.

Is that what is going to happen?

I am not sure. What do we do with some of the smartest and most skilled people under capitalism right now? Do a super intelligent individuals in a workplace put people out of work?

There are many very capable and very intelligent people with access to the internet right now. Do they put people out of work in those places? Do we see workplaces filled with people trying to be more intelligent that other people in offices?

While the threat of AI systems to some jobs are real, The Dystopian visions from Altman and Musk—who themselves are "trying to save humanity"—are a strange thing to behold. It is unclear if either of them really understand how the world works, how the economy functions, or how things are actually made.

One ironic situation stands out to me: Nvidia chips.

OpenAI is not really helping the supply of Nvidia GPUs, which are necessary for the production of the datasets that run ChatGPT and other language models.

And, while other companies are developing (and have developed) their own AI-focused processors, the supply of these is limited by real world things.

Altman said there had been “a brutal crunch” all year due to supply shortages of Nvidia’s $40,000-a-piece chips. He said his company had received H100s, and was expecting more soon, adding that “next year looks already like it’s going to be better”.

However, as other players such as Google, Microsoft, AMD and Intel prepare to release rival AI chips, the dependence on Nvidia is unlikely to last much longer. “I think the magic of capitalism is doing its thing here. And a lot of people would like to be Nvidia now,” Altman said. (FT)

The "magic of capitalism" is not the only magic that Altman and others in his circle believe in. The belief that the magic of their engineering teams to conjure "intelligence" from statistical models.

Intelligence is a process, not a thing that is created. That process is a combination of social context and many individual actions. Large language models have part of this creation of knowledge, that is they have the combined history of written stuff on the internet and it can summarize that and do some fancy restructuring of that information. But, it is not capable of generating new knowledge. It is not engaged in discovery and it is not inspired to seek-out additions from the universe.

I think that this is part of the problem generally in society. The difference between engineering and science is an important distinction to be made.

The difference between "innovation" and curiosity-driven discovery is the difference between using the information provided to you and the search for and testing of new information's correctness.

I am not naive to think that we cannot eventually create an artificial version of the human mind and our social structures, but it is well known that the very specific limitation of large language models is that they cannot tell the difference between truth and lie. It is at least this "telling the difference" that is required to even attempt to be intelligent.

The growing threat of cyber attacks on critical infrastructure should be a concern for workers when it comes to health and safety.

Australia shipping ports are one of the more recent targets of an attack that brought down infrastructure reliant more and more on instructions provided a central system. If a cyber attack happens on critical infrastructure it can lead to a rather sudden stop of operations, something most operations cannot do quickly safely.

Large transport and production infrastructure shutting down can put workers lives at risk who are use to a constant movement or process where manual operations continue to operate through lack of communication, only altering that process if a warning is received.

These kinds of "no word is green" processes need to be examined in operations reliant on automation or hybrid work environments.

Collective agreement language needs to bargained to contemplate these and other scenarios in the face of increased digital automation and the increased attack surface such systems create.

He added that the government had not yet identified the perpetrators of the cyber-attack, which caused the firm to disconnect its ports from the internet.

DP World said it halted internet connectivity at its ports on Friday to prevent "any ongoing unauthorised access" to its network.

Going offline meant trucks had been unable to transport containers in and out of the affected sites. (BBC)