Artificial Intelligence

• Slides
• Audio/Video Presentation
• Surveillance Presentation

The analysis of automation in a particular workplace starts by identifying whether new technology is being deployed as a tool used by craft labour or as an industrial machine.

Craft tools are used by the worker to complete their skilled labour. Upgrading tools can lead to an increase in efficiency, safety, or ease of work.

We call the implementation of this kind of technology augmentation.

An example for a factory worker in an industrial work environment might be a heavy-lift support arm, something that supports the manipulation of a piece of a product being assembled. For a so called white collar knowledge worker, it might be the addition of spell checker or a central storage and versioning system for their documents.

Depending on the workplace, these upgraded craft tool can be brought into a workplace by the individual worker (like a better hammer) or can be brought by the employer if use is more of an argument for productivity gains and/or health and safety.

Sometimes there is government (or third party) certification or standards when they are used for certain industrial practices setting out best practices for implementation. Think certified safety work boots or welding equipment. Other times, the tool is simply upgraded as a semi-consumer-level tool like mobile phones.

For industrial machinery, the economics are somewhat different. Industrial machines are designed and implemented by firms to replace workers to purposely automate a process. There is a specific goal of increasing productivity. These are capital intensive, very expensive, require financial risk analysis as there are additional long-term service costs associated with their implementation. As such, knowing that they will increase productivity and profitability is the question.

While tools and machines are discussed as different processes in a business and are themselves different applications of technology, this is much more of a spectrum than most realize.

The process of automation happens through the capturing of the logic and motion of labour and infusing that knowledge into a machine. Automation is designed to replace that work (if not the worker).

When we talk about innovation for productivity, that is the explicit goal. The process of automation for replacement must be examined over a longer time frame and not as a snapshot of this or that technology being implemented.

Many successful augmentation tools get upgraded by incorporating more and more information about the work.

Add a computer to the pneumatic lift assist, some sensors, and electric motors you get closer to the robot arms that replaced the worker on the assembly line. The process is one where the craft tool is upgraded through capturing movement information from the worker into the tool.

Every step of this process involves deskilling. Augmentation is always going to make the job of the worker easier as it incorporates and speeds up parts of the task.

This type of augmentation provides the pathway to the development of industrial machines and replacement. Over time, it is easy to see how these tools for augmentation become the machine for replacement.

Before replacement though, augmentation can be designed to drive speed-up.

Speed-up can be driven by augmentation at any point along production. Allow a worker to work faster at the beginning and there are more of a product to deal with at every other step in a day. This goes for industrial processes and white collar work.

There is barely any difference between augmentation of industrial process and when the same augmentation is applied to white collar knowledge work. When dealing with LLM/GPT/Generative AI systems, the process is very clear.

The augmentation is possible through learning the process used by the worker and directly incorporated into the craft tool. AI tools for augmentation are the very same tools used to learn the specific tasks that are being automated. Speed-up is inevitable as the tools make some jobs easier. Some workers incorporating LLMs into their workflow leads to speed-up down the line for your coworkers.

Slop or not, there is more produced over a period of time. (See our previous comments about value decline per unit not being a limit to introduction of automation.)

We know the goal here, it is to replace tasks of workers in the office in the same way as automation in industrial workplaces reduced head count in industry in North America.

The question, then, is not "can LLMs replace jobs?" It is what are the economics involved driving and limiting craft tools becoming industrial machines in capitalist firms? We have talked about those economics before and it has to do with profitability and competition driven by the regulating capital.

The point here is that the economics are the same no matter where you work. White collar or industrial worker, the entire picture must be seen to push back against speed-up and capture of any increased surplus value created through the introduction of better tools.