Machine Learning
Dave Snyder
Overview
Most so-called AI software is based on machine learning.
When comparing conventional software with machine learning
software, note the following:
To make conventional software, software developers must write out each step to create a specific software product.
On the other hand, to make machine learning software, software developers must collect relevant data. This is used to
“train” the software. After this training, the software is available for use.
No step-by-step programming is required.
Both types of software
generally require a lot of effort to produce, but machine learning software requires developers
with a different skill set than what is needed for conventional software.
Neither type of software is perfect, and as such both types of software must be tested to look for problems.
Each example of machine learning software is based on a model, and models vary in size.
Current LLM (large language models) have over a trillion parameters. However smaller models exist.
It is in theory possible to construct conventional software that can be exhaustively tested.
Machine learning software with large models cannot, even in theory, be exhaustively tested.
Machine learning is not software that “learns” in the normal sense
of the word. Software can be written so it changes its behavior over time in response to input, however that
is not necessarily true of machine learning software. And in any case, no software is as good at learning as a typical human.
Applications
There are several potential applications for machine learning, some have been more successful than others.
A few examples.
- Processing Scientific Data
Scientists collect data, process/analyze that data and determine whether the results agree with theory.
In most cases software, specifically conventional software, is used to process/analyze the data.
In some cases, machine learning software has been used along with conventional software.
This has worked well enough that the nobel prize in physics for 2024 was awarded for work on applying
machine learning to scientific research.
- Medical Applications (Medications)
Developing new drugs involves two steps. 1) Proposing a new substance that might have therapeutic properties.
2) Testing that substance to ensure it is both safe and effective. Software has been helpful in step 1, but so far has not
been helpful in step 2. To date machine learning has not been successful in producing any new drugs (see Rosen, 2025)
[see references].
- Medical Applications (Other)
Other applications include diagnostic procedures and mental health. There has been some limited success in
produce software that assists medical professionals in doing their jobs. However it is problematic to allow
people to use software without the guidence of medical professionals.
- Business Applications
A number of businesses have attempted to use machine software to improve internal operations. According to a
recent survey 95% of businesses report that the costs of using the software exceed the benefits. There is a
possibility this situtation could improve if businesses develop more realistic expections of what this software
can do, and learn the best ways to encorporate that software into business practices (see Challapally, 2025)
[see references].
- Other
The most commonly used applications are chatbots such as ChatGPT. People have
used chatbots for a variety of use cases. Users of chatbots need to be aware of the negatives such as
- This type of software has potential data privacy issues.
- It has a tendency to make mistakes.
- The output of this software will tend to reflect the biases found in the data used to train it.
- It contributes to misinformation already present on internet sites.
- Large language models require large amounts of electricity, and add to the demand for data centers.
References
See Book List for AI, Machine Learning and Quantum Computing
Links
Modified January 29, 2026