News JVTech GPT-4: Ranking of AI’s Most Easily Interchangeable Professions, the study that says the terms
If we already knew that GPT-4 could easily replace much of the workload of a variety of professions, a study has just quantified those doubts. So ready for a career change?
AIs, an almost frightening development
The public perception of artificial intelligence has completely changed in less than a year. A model that learns and adapts, we knew, but everything seemed far away. Today it is as if all other trades are handcrafted by a thread and employees could be replaced by a machine at any time.
Researchers at the University of Pennsylvania in the US rolled up their sleeves and gave us one learn Analyzing the current state of Generative Pre-Trained Transformers (GPT) and getting started the impressive speed at which these models are progressing.
To illustrate this, the graphic below shows the Percentage of performance obtained by GPT 3.5 (developed May 2020) and GPT-4 (released March 2023) on a large number of tests. As you can see, most of these tests get much better results with the new version of the AI. Everything is moving fast, very fast, rubbing off from year to year with the micro-novelties of smartphones and other physical technological objects.
There are many tasks where it is utterly impossible to distinguish the work of a human from that of an intelligent machine. Some just can’t run missions for much less than a salary LLM (Learning Language Models) can do it better than a human. How could companies miss such an opportunity in the years to come?
List of professions most likely to be replaced by GPT-4
Now for the most interesting part: What professions could lose in a duel against GPT-4? To answer this question, the researchers determined a unit of measurement: the level of exposure. This level is defined by the potential economic impact of exposure to machine learning without considering the organizational changes that will be forced by the arrival of an AI instead of humans.
The study is extremely complete and I invite you to go through it if you have the time and are fluent in English. Broadly speaking, the researchers tell us the following:
Our analysis shows that approximately 19% of jobs have at least 50% of their tasks open, considering the capabilities of both models and the tools created with them.
Now here is the Table of occupations most affected by exposure to LLM. The study brings together the results of various types of measurement.
It seems that mathematicians have something to warm their butts about in the future. This trade has 100% exposure in no fewer than three metrics. The interpreter’s job is also at risk. In fact, translating live or offline content is something artificial intelligence is doing more and more easily.
Overall, the highest-paying jobs are the ones most likely to change over the coming years. Not surprisingly, it is the physical and manual trades that have no problem with artificial intelligence.
The conclusion of the study is epideictic towards GPT-4:
GPTs generate profound transformations, enable potential technological advances, and have major implications for professions.
While they remain an interesting support for many current professions such as programming, this form of intelligence is not strong enough to kick everyone out. But with speed things move…
