Recently, much has been written about so-called artificial intelligence (or AI). However, most of what has been said:
So-called AI has developed slowly over the past 70 years, and has been overhyped since the release of ChatGPT in November 2022. Thankfully there have been some honest assessments of this technology, especially this summer (2025). While there are positive aspects, the technology also has negative aspects.
The phrase “Artificial Intelligence” and its abbreviation “AI” does not have a universally accepted definition. The definition seems to vary depending on context. In practice the best definition might be “computer software that is more capable than expected.”
In my opinion, the word intelligence should not be applied to inanimate objects such as computers, no matter what kind of software might be running on that computer. This also applies to the following phrases:
The people who are the strongest proponents of so-called AI seem to have developed a philosophy with the following parts:
I disagree with all of these points.
75 years ago, Alan Turing suggested a computer that simulates the human brain is possible (Turing, A. M., 1950). He started with assumptions on the storage capacity of the human brain, and then concluded that building a computer with similar capacity is possible. More recently, AI proponents have made similar arguments: An adult human brain contains approximately 100 billion neurons and 100 trillion synapses. While these numbers might seem large, the AI proponents argue building a system with this number of components is not unreasonable with current technology.
However, this relies on an invalid assumption. It only works if neurons and synapses are simple objects that can be easily simulated with simple electronics or a small amount of computer code. We now know this is not the case. Neurons and synapses are complex things that are not completely understood. It is clear that even if we did understand them, simulating either neurons or synapses would be difficult. Even if we did understand them, simulating large numbers of these objects is well beyond current technology.
The argument that it is important to move as quickly as possible is not rational and is wasteful. Since the goal is well beyond current technology, taking a slow approach makes much more sense. To get to that goal requires new ideas which will take time to discover.
Consciousness is not completely understood. The “sufficiently complex” idea is simply too imprecise to be an adequate explanation for consciousness; we need to wait for an adequate explanation.
Much of the software described as “AI” uses a technology known as “machine learning” (see machine learning).
See Book List for AI, Machine Learning and Quantum Computing