Introduction
In his 2017 talk at Stanford University, Andrew Ng referred to Artificial Intelligence as the “new electricity” and added “Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years”.
A Look Back on Previous Three Industrial Revolutions
As discussed in the upcoming book, “The Fourth Industrial Revolution and 100 Years of AI (1950-2050)”, each of the three previous industrial revolutions had many inventions with one leading to the creation of a new infrastructure and the other having a Cambrian explosion and diversifying in various shapes, forms, and sizes.
For example, in the first three industrial revolutions, the three newly created infrastructures were related to water-and-steam, electricity, and electronic communication, respectively. In the fourth and current one (which started in 2011 and may continue beyond 2050), the new infrastructure will be related to production, ingestion, cleansing, harmonizing, and utilizing of data.
Similarly, in the first three industrial revolutions, steam engines, electric motors, and processing units had their Cambrian explosions and they started being used in various shapes, forms, and sizes. In fact, today, there are at least 3,000 different types, shapes, and sizes of motors; they are everywhere and usually invisible. Their worldwide market is expected to be $220 billion by 2030 and including them in a new gadget is a no-brainer.
Similarly, AI systems are already witnessing a Cambrian explosion. They are being used in more than two thousand use-cases and applications. Indeed, within two decades, these systems are likely to diversify into various forms, shapes and sizes to solve more than a hundred thousand different types of applications and use cases.
However, the analogy between AI systems and electric motors, steam engines, or processing units pretty much ends here. This is because contemporary AI systems are neither explainable nor interpretable and suffer from inherent biases, ethics-related issues, and non-causal behavior.
Limitations of AI Systems
Furthermore, most contemporary AI systems, particularly Deep Learning Networks (DLNs) and their embellishments suffer from the following debilitating limitations:
- They are brittle and break down easily; even when they are wrong, they often provide answers with 99% confidence.
- They often have “machine hallucinations”.
- Because their output is very coherent and polite, they provide “machine endearment” and fool people into trusting them completely and assuming their outputs as facts.
- It is much easier to infuse them with malware (e.g., injecting GPTs and LLMs with “Prompt Injections”), which could cause havoc for the users.
The implications of these debilitating limitations will be discussed in the next four blogs and in the upcoming book.
The book titled “The Fourth Industrial Revolution and 100 Years of AI (1950-2050) will be published in December 2023. For details, see
www.scryai.com/book
