Screenshot of a VisiCalc Emulator. It was the first spreadsheet and the first ‘Killer App’.
A couple of quick thoughts on the current ‘Artificial Intelligence’ (AI) debate:
The purpose of a computer is automation. So the whole point of computers is the ability to automatically process specific tasks. As we can see on a daily basis, these tasks can be enormously complex. Nonetheless — everything that happens on a computer is automation. This includes the processes that we are currently calling ‘Artificial Intelligence’. There is no fundamental difference between automatically sorting numbers and automatically drawing pictures.
It has been noted before — and I am happy to repeat it again: the terms ‘Artificial Intelligence’ and ‘Neural Networks’ are misleading. They suggest a biological agency where there is none. While the concepts are certainly inspired by nature, we are not dealing with biology. ‘Neural Network’ is a fancy name for a set of algorithms that can also be described as ‘non-linear statistical data modelling tools’. Less fancy, more accurate.
With ChatGPT, ‘Artificial Intelligence’ has reached its VisiCalc moment. In case you don’t know what VisiCalc is: it was the first spreadsheet. In 1979, VisiCalc enabled users to compute and re-compute a large set of numerical relations. Change one value and the whole table is automatically updated. For its time, VisiCalc was extremely powerful and practical. It was the first application on a home computer that was really useful. The same is true for ChatGPT. It is the first AI application on a home computer that is truly useful. VisiCalc saved you the tedium of re-computing numeric data, ChatGPT saves you the tedium of writing text.
In this sense, I suggest to treat ChatGPT simply as a productivity tool. It helps you to produce more text in a shorter amount of time. The quality varies — but it is good enough for most purposes. (I know this is a bit of an oversimplification. But at the same time we should not underestimate the impact of a good productivity tool.)
The entire history of technology is about annihilating jobs. The cotton mills destroyed traditional spinning and weaving communities. The steam engine replaced muscle power. The word processor eliminated the typist. Desktop Publishing put typesetters and lithographers out of business. ChatGPT and other ‘AI’ tools will certainly destroy a lot of jobs. Which makes me wonder: is ‘AI’ economically feasible? Certainly for individual companies — but what about the entire economy? What will the new jobs look like? Where is the occupational opportunity?
Up until now the narrative of technological progress was always that technology and automation would make our lives and our jobs easier, more comfortable and more productive. Initially, machines replaced manual labour. Then early computers replaced intellectual labour. And now ‘AI’ replaces social and creative labour. But what does this imply? Do we want to have machines that replace social and creative labour?
What are the positive narratives for ‘AI’? I am really surprised that the industry is so bad a coming up with scenarios where AI can make a positive impact. All the examples for a positive use of ‘AI’ — like fighting cancer or climate change — are pretty vague. Perhaps there are no positive narratives? Is ‘AI’ a solution looking for a problem?
The progress of ‘AI’ technology is always framed as inevitable. Now that the genie is out of the bottle, there is no way to stop it. Because ‘AI’ exists, it must be used and it will change everything. I don’t think that is necessarily true. While technological development has certainly a powerful dynamic, open and democratic societies are able to assess, negotiate and regulate the use of technology.
It is intriguing that up until now software has always been fairly unregulated — at least compared to hardware. Machines (like planes) and technological systems (like aviation) are strongly regulated in order to ensure a high degree of safety and security. Maybe the rise of ‘AI’ will finally lead to a more regulatory approach to software in general.
Training data is the foundation for any ‘AI’ algorithm and as such more important than the algorithm itself. Data is not neutral. It reflects the world view of its creators and as such contains biases and misconceptions. ‘AI’ algorithms expand, amplify and reiterate these biases and misconceptions. Better ‘AI’ starts with better data.
Completely different perspective: let’s assume that ChatGPT is similar to photography. ChatGPT is able to automatically generate a text on any topic — photography is able to automatically generate an image of any sight. Both use technical automation for replacing intellectual and creative labour. Photography was initially considered to be just a scientific tool — an apparatus not an art form. It took many years before the artistic merits of photography were recognised. Maybe ‘AI’ tools will allow for new forms of creative and artistic expression that we simply don’t perceive right now.
We are scared of the wrong things.
Press enter or click to view image in full size
‘The Boulevard du Temple’ by Louis Daguerre from 1838. One of the first photographs ever taken.
