# Some Notes on Pros of Open Science and Open Source
- **Pooling Resources**: Building off of one another’s strengths; learning from one another’s failures.
- **Accessibility**: Anyone can use the models, regardless of budget or affiliation.
  - This also helps to ensure diversity of contributors.
- **Lowering Barriers**: You don’t need to have a tech job to explore how AI works.
- **Innovation**: High-value applications are possible for more people to discover and create.
  - Relatedly, advancements in **addressing bias/harms** become more possible.
- **Economic Opportunity**: More access leads to more businesses and jobs.
- **Transparency**: Users and those affected have full visibility on the model and the training data. They can better identify potential biases or errors.
- **Accountability**: Provenance to trace who-did-what; independent auditing possible.
- **Privacy**: Users don't have to send their data to black box APIs.
- **IP protection**: Users train their models on their data, and own them.
- **Freedom of choice**: Users are not locked in. They can switch models anytime.
- **IT flexibility**: Users can train and deploy models anywhere they like.
- **Tailored use**: Users can train/fine-tune for their specific needs.
- **Safety**: More mechanisms available.
- **Speed**: Good ideas can quickly flourish and be built on. Security issues can be quickly addressed.
- **Diversity** of options.

# Cons of Closed Source
- **Centralization** of power.
- **Opacity** of subtle bias/harm issues.
- Hiding **illegal** or problematic data.
- **Bare minimum of legal compliance** as opposed to good practices.
- Fostering **misunderstanding for hype and profit**.
- **Insularity of thinking** creates "groupthink" technology issues (such as harming people with marginalized characteristics).
- **Security issues** not addressed quickly.
- Consumer apps **can’t be flexible** and become dependent on a single model: Consumer apps built on top of closed source must “lock-in” their code based on what an API outputs; as closed source internal models are updated or changed, this can completely break the consumer’s system, or the consumer’s expectations of behavior.

# Common Misunderstandings
## There’s an idea that open source is “less secure”.
- Misses that closed software has just as dire (or more so) security concerns as open source.
- Misses the fact that the diversity of options available with open source limits how many people will be affected by a malicious actor.

## There’s an idea that open source will help China to “beat us”.
- Misses that part of why U.S. technology has flourished due to open science/open source.
- Misses that U.S. dominance is a function of how friendly the U.S. is to companies: There is more to success than the code itself, the socioeconomic variables that the U.S. provides is particularly well-placed to help open companies flourish.