This note is to store TM’s current/latest conception of AI, and related musings.
Conception
TM wants a section to discuss what “AI” even is, and to clarify usage of the term, and related terms such as:
- GenAI
- LLMs
As TM currently understands, “AI” is best thought of as a container term for all human attempts to emulate intelligence (typically of a form recognisable as human intelligence); it’s very woolly, imprecise, general, but in this imprecision and generality is the term’s power - its ability to represent and unify all endeavours to “replicate” human intelligence (and perhaps even create new forms of intelligence), and hence remind us of these broader aims and mitigate any tunnel vision / one-track mindedness in how specifically we pursue it (by e.g. concentrating only on the scaling up of current LLM technology, as opposed to exploring alternative methods/paradigms).
In TM’s current worldview, GenAI (generative AI) refers to a specific subset/form of AI which uses LLMs to generate responses to prompts. Probably since the launch of ChatGPT late 2022, GenAI has become the most prominent, dominant paradigm/approach to/within AI.
TM would like to clarify how best to conceive of the difference between “GenAI” and “LLMs” - e.g., could an LLM be considered one way to implement GenAI and/or are the terms effectively interchangeable in most worldviews, etc.
Good
Helps clarify thinking The process of writing a prompt to summarise what help I need, incorporating all relevant information, I feel has developed my ability to more quickly and succinctly describe things in life in general, and to adopt the perspective of someone who has “no context”, so I can provide that context when verbalising something.
Bad
TM acknowledges there are many harms associated with the current implementation of AI, which is characterised by LLM “services” such as Claude, Gemini, ChatGPT, etc. Would like to cultivate an overview/scope of them here. This is by no means exhaustive/comprehensive, and only serves to reflect the things TM is currently aware of / has thought to include. The aim is to greatly broaden and deepen this scope as TM incrementally decreases ignorance. A current overview:
- Environmental harm and resource (over)usage
- Data centres consume swathes of potable water for cooling purposes ?