The AI debate continues to dominate public discourse, with the emphasis now being on the job market, and while there are going to be some effects, they are unlikely to result in millions of white-collar workers being out of work.
The problem with the AI debate
The other problem with this debate is that it is already becoming highly politicised, meaning that facts are quickly giving way to tribal rivalry, which will help no one.
From a very high level, this fear is very easy to understand.
Machines have been created that can discourse like humans meaning that many white-collar jobs that humans do will shortly become obsolete.
However, I think that the mistake that the doom-mongers are making is that they assume that because the machines can discourse like machines, they can also think like them.
This idea is demonstrably false, as evidenced by many laboratory experiments, the hallucinations of generative AI chatbots and the failure of machines to be able to learn how to drive vehicles.
Machines are purely statistical in nature
As I have stated many times, the heart of this weakness lies in the fact that these machines are purely statistical in nature, and as such, they have no causal understanding of anything that they do.
Unlike humans, they are unable to distinguish between a relationship that is casual and one that is purely down to correlation.
This means that they don’t think like humans, and, in fact, I would argue that they do not think at all merely compute statistics.
Incapable of thinking creatively
Consequently, they are incapable of thinking creatively because true creativity requires the creation of something that has never been seen before.
Even the largest language model with the most compute horsepower behind it will be quickly outclassed by my 8-year-old when both are confronted with a task that neither of them has seen before.
These massive networks are now many times bigger than the human brain and consume kilowatts of power (the human brain runs at 20 watts), but still, they are hopelessly outclassed.
Consequently, I have no fear for the creative industries, although we may see a large increase in the mindless copycat pap that is available.
The impact of these generative AIs over the next few decades could be to take away many of the mindless tasks that humans are forced to endure in order to put food on the table, and I suspect that they will be replaced by much better positions.
Furthermore, I think that the right way to deal with this innovation is to use the technology to do more.
For example, a legal clerk will be able to produce five or 10x more legal briefs than he did in the past, but he will still be needed because generative AI is completely unreliable, and its work must be checked.
This unreliability is not going to disappear because the way that these models are trained has not materially changed for the last 11 years, and, as of yet, no alternative has emerged.
Companies will face a choice between doing much more with the humans that they already have or doing the same with fewer people.
Business owners and shareholders want growth
Almost all business owners and shareholders I know want growth, and so I suspect that only a few companies will choose to cut staff massively should generative AI prove to be as useful as everyone is promising.
Given the nature of the hype cycle, I suspect that their real capability is being overstated, and so their impact on the economy is likely to be less than everyone is enthusiastically espousing.
I also think that this technology is likely to create a whole host of new professions as, after all, no one had heard of a social media influencer or YouTube creator 20 years ago.
The limitations of these algorithms are going to ensure that a large number of jobs now considered to be extinct will last for much longer than people expect, meaning that there is time to transition.
The fact that all the empirical evidence suggests that the machines are as dumb as ever despite giving the very convincing illusion to the contrary is why I am not overly concerned with its impact on the job market over the next five years or so.
Some workers will have to adapt or learn new professions over time, but these are likely to be better paid and more fulfilling than the drudgery that they currently endure.
Consequently, in the long term, these AIs may make the quality of work done to put food on the table better rather than worse.
Risks with generative AI
There are plenty of risks with generative AI, but I have argued that these are mostly due to misuse by bad humans rather than the machines suddenly deciding that they want to see the world burn.
Hence, I continue to think that the real threat of generative AI will be driven by humans who desire to commit bad deeds and use it to do so.
This is where the focus on prevention needs to be, and I think it also needs to bear in mind that blanket regulation has the potential to damage the development of AI that has legitimate, lawful and very profitable use cases.
This, in turn, would slow the creation of new and more fulfilling professions meaning more mindless paper shuffling and less use of the best organic intelligence system ever created.