Automation of Human Tasks: Typology with Examples of ‘Writing Text’, ‘Calculating’, and ‘Planning’

This text is part of the text “Rebooting Humanity”

The German version can be found HERE.

Author No. 1: Gerd Doeben-Henisch

(Start: June 5, 2024, Last updated: June 6, 2024)

Starting Point

In the broader spectrum of human activities, there are three ‘types’ of action patterns that are common in everyday life and crucial for shared communication and coordination: (i) writing texts, (ii) describing (calculating) quantitative relationships, and (iii) planning possible states in an assumed future. All three types have been present since the beginning of documented ‘cultural life’ of humans. The following attempts a rough typology along with known forms of implementation.

Types and their implementation formats

(a) MANUAL: We write texts ‘manually’ using writing tools and surfaces. Similarly, in quantitative matters, there is manual manipulation of objects that represent quantitative relationships. In planning, there is the problem of how to represent a ‘new’ state: if the ‘new’ is ‘already known’, one can revert to ‘images/symbols’ of the known; if it is ‘truly new’, it becomes difficult; there is no ‘automatism of the new’. How do you describe something that has never existed before? Added to this is the — often overlooked — problem that ‘objects of planning’ are usually ‘value and goal dependent’; needs, intentions, expectations, factual necessities can play a role. The latter can be ‘socially standardized’, but given the ‘radical openness of the future’, history has shown that ‘too strong standardizations’ can be a shortcut to failure.

(b) MANUAL SUPPORT: Skipping the phase of ‘partially mechanical’ support and moving on to the early phase of ‘computer support’, there are machines that can be ‘programmed’ using ‘programming languages’ so that both the writing instrument and the surface are represented by the programmed machine, which allows many additional functions (correcting texts, saving, multiple versions, automatic corrections, etc.). However, one still has to write oneself: letter by letter, word for word, etc. In ‘calculating’, writing down is still very laborious, but the ‘calculation’ then takes place partially ‘automatically’. Planning is similar to writing texts: the ‘writing down’ is supported (with all the additional functions), but ‘what’ one writes down is left to the user. Apart from ‘quantitative calculating’, a ‘projection’, a ‘prediction’ is generally not supported. An ‘evaluation’ is also not supported.

(c) LANGUAGE-BASED SUPPORT: The phase of ‘language-based support’ replaces manual input with speaking. For selected areas of texts, this is becoming increasingly successful. For ‘quantitative matters’ (calculating, mathematics, etc.), hardly at all. For planning also only very limited, where it concerns already formulated texts.

(d) ARTIFICIAL INTELLIGENCE ENVIRONMENTS: The Artificial Intelligence (AI) environment is considered here in the context of dialogue formats: The user can ask questions or send commands, and the system responds. The relevant formats of AI here are the so-called ‘generative AIs’ in the chatbot format. Under the condition of ‘existing knowledge’ of humans in the format ‘stored documents/images/…’ and under the condition of ‘dialogue formats’ of humans (also through explicit training), these generative AIs can process questions and orders in ‘formal proximity’ to the known material in the context of a dialogue so that one does not have to intervene oneself. Corrections and changes in detail are possible. Both in ‘text creation’ and in ‘calculating’, this can function reasonably well within the realm of the ‘known’. However, the actual accuracy in the ‘real world’ is never guaranteed. In ‘planning’, the specific problem remains that for the AI, ‘truly new’ is only limited possible within the framework of combinatorial matters. The truth reservation remains, but also applies in the ‘manual case’, where the human plans themselves. The evaluation problem is also limited to already known evaluation patterns: the future is not the same as the past; the future is more or less ‘different’. Where should the necessary evaluations come from?

An interesting question remains, in what sense the advancing support by generative AIs actually supports communication and coordination among people, and specifically, whether and to what extent the central challenge of ‘future planning’ can be additionally supported by it. The fact is that we humans struggle with future planning in the areas of social life, community life, and larger contexts such as counties, states, etc., ranging from difficult to very difficult. But, for a sustainable future, successful planning seems to be indispensable.