====== Artificial General Intelligence (AGI) ======
{{tag>Undefined}}

Research into computer-based A//rtificial Intelligence// (AI) began over sixty years ago with the development of //Expert Systems// ([[https://en.wikipedia.org/wiki/Expert_system#History|ref.]]) but AI only began to (appear to) show some parallels with human intelligence with the deployment of workable //Large Language Models// (LLMs)([[https://en.wikipedia.org/wiki/Large_language_model|ref.]]) in the late 2010s.

A proposed derivative of AI, called// Artificial General Intelligence// (AGI) currently lacks an agreed definition - but is broadly taken to mean an AI-based system which equals or surpasses human mental capacities across a wide range of cognitive abilities.

Despite its lack of definition, recent progress with AI LLMs has prompted some leading computer scientists ([[https://www.cs.toronto.edu/~hinton/|ref.]]) and philosophers to suggest that Artificial General Intelligence //could //oneday exist (note there is also very little agreement on a timescale).

A 2024 [[http://arxiv.org/pdf/2311.02462|paper]] in //Proceedings of the 41st International Conference on Machine Learning,// authored by one of the firms involved in promoting the research - //Google Mind// - reviewed nine 'case studies' of previous definitions, and created six 'principles' by which (they say) AGI might be judged.

>We hope our framework will prove adaptable and scalable – for instance, how we define and measure progress toward AGI might change with technical advances such as improvements in interpretability that provide insight into models’ inner workings."

Note that AGI's definition is greatly complicated by the lack of agreement about what biological [[content:psychology:general:intelligence]] itself actually is, and how to define it.

Some proponents of AGI suggest the possibility that systems could oneday become 'conscious' (by some unknown means) though again, there is currently no agreement about what [[content:psychology:general:consciousness]] itself actually is.

To recap the unknowns surrounding AGI :

  * How can AGI be defined?
  * What is 'Intelligence'?
  * What is 'Consciousness'?
  * When (if ever) might AGI happen?

**Notes : **

1) The exact workings of the computer programming constructions known as [[content:computer_science:neural_networks]] - on which LLMs are fundamentally based - are unclear.

2) Because of Neural Network obscurities and the complexities of their enormous database inputs, it's not always straightforward to understand exactly how or why LLM queries produce the results which they do.

3) Although extremely useful and fast for solving many real-world problems, LLMs are //never 100% reliable// - sometimes giving entirely ludicrous answers, but nevertheless presented with great 'confidence'. A syndrome known in AI jargon as 'Hallucinations'. Presumably, an AGI system would also 'hallucinate'.

4) Various skeptical observers of what they call the 'AGI Hype Cycle' have pointed out that most of the groups involved in AGI research are fully commercial enterprises which could stand to financially gain from investments, sales, and promotions of systems which they create - whether they are truly AGI ( according to a yet-to-be-agreed definition ) or simply appear to be,