The emergence of Giant Artificial Intelligences (gAIs) like ChatGPT has garnered significant attention in recent times. However, there are concerns related to High Modernism in designing these AIs.
What is ChatGPT?
ChatGPT is a variant of GPT (Generative Pre-trained Transformer) which is a large-scale neural network-based language model developed by OpenAI.
GPT models are trained on vast amounts of text data to generate human-like text.
It can generate responses to a wide range of topics, such as answering questions, providing explanations, and engaging in conversations.
In addition to being able to "admit its mistakes, challenge false premises, and refuse unsuitable requests," ChatGPT can also "answer follow-up questions."
The chatbot was also trained using Reinforcement Learning from Human Feedback (RLHF).
How High Modernism Poses Challenges in Designing AI?
About High Modernism:
High modernism refers to a top-down ideology driven by a faith in order and measurable progress. It is characterized by an unfaltering confidence in science and technology as means to reorder the social and natural world.
It often disregards local knowledge and lived experiences, leading to unintended consequences.
This approach, when applied to AI design, can overlook the complexity and diversity of human thought, resulting in biased and incomplete systems.
Challenges of High Modernism in AI Design:
Loss of Diversity: gAIs are trained predominantly on internet text, which is biased towards certain languages, religions, races, and cultures, risk perpetuating these biases.
The lack of diverse training data can lead to language loss and hinder the richness of human thought and expression.
Reduction of Local Knowledge: By abstracting away the nuanced knowledge acquired through direct experience, gAIs prioritize the "atlas view" of information available on the internet.
This approach neglects local context and the territory-specific insights necessary for accurate and multi-dimensional understanding.
What is the Role of Diversity in Designing AIs?
Lack of diverse AI models may result in standardized, one-size-fits-all solutions that fail to account for regional, cultural, or individual variations.
Promoting diversity in AI development can yield multiple perspectives, encouraging innovation and tailored solutions.
Diversity in AI models helps capture a broader range of perspectives and knowledge, reducing biases and enhancing the models' understanding and response capabilities.
It allows for the inclusion of diverse languages, cultures, and experiences, fostering a more inclusive and equitable AI landscape.
What are the Ways to Thwart the Risks Posed by gAIs?