Machine learning models draw much of their inspiration from biological brains. Composed of millions of neurons working in unison, these models are extremely powerful. However, like biological intelligence, they are susceptible to occasional errors or confusion. By structuring data, enriching it with metadata, or setting guardrails on the way it's processed, we can significantly enhance their accuracy. This improvement is feasible because we're enhancing a system with excellent computational abilities and superior memory (traditional computers) with robust machine learning models capable of solving problems that cannot be easily described with mathematical formulas or code.

This scenario arises because AI and traditional computing are based on fundamentally different computing models. AI exhibits the capability for "Fuzzy Logic", which relates to understanding problems with imprecise inputs and outputs, such as recognizing a human face in various positions or transcribing sound from a noisy recording. In contrast, traditional computers operate on "Exact Logic", in which both the inputs and outputs can be described with high precision, and any variation must be explicitly encoded. Examples of such operations include performing mathematical computations or removing all uppercase letters from a paragraph.

Consequently, when AI models and traditional computers collaborate, their combined effectiveness significantly increases. Leveraging their individual strengths, they form a synergy analogous to our relation with the applications we use every day. The apps on our mobile phones and computers give us almost superhuman abilities, like perfect and virtually limitless memory, the potential to visualize and act on unusual health data, or even talk in real-time with someone on the other side of the globe. Unlike us, AI applications don't require user interfaces as they can swiftly process and operate on raw data. However, they can use APIs and databases, in the same way we use apps and notepads.

In my view, AI is not here to replace traditional software. Instead, it brings it to a new level by giving us tools to introduce features that were unthinkable just a year ago. AI will empower us to build more intelligent machines with impeccable memory that can interact with humans in more natural ways while retaining existing computer capabilities.

So software will likely remain the same under the hoods, but the ways we interact with computers will likely change radically.

We can now create machines that can talk.