Singer and Demo argue for* a society of diverse specialist AI agents like I have been building.** Each and every living creature resides in a niche. I believe it will be much the same for mechanical life.
* The Future of AI is Many, Not One, arXiv:2603.29075, 30 March 2026
** See, for example, my blogs of 16 September 2010, 1 February 2024, 20 November 2023, 1 October 2021, etc.
I have previously described* how to give A.s.a. H. AI robots most of the Toki Pona language vocabulary. Sonja Lang's Toki Pona Dictionary** can then be used to chain these core concepts together to build a large concept structure and an English vocabulary of many thousands of words. For example: snow=(cold, moist, white, powder), learn=(information, approaching), brain=(thinking, object), etc., etc.
* See, for example, my blogs of 1 October 2015 and 4 September 2024 and Trans. Kansas Acad. Sci., vol. 120, pg. 108, 2017.
** Tawhid, 2021.
Although I am not a capitalist I would point out that another advantage of specialist A.I.s is the possibility that different specialists might be produced by different manufacturers. Each manufacturer can, itself, be specialized. (This would possibly also reduce the cutthroat hypercapitalist competition between manufacturers.)
Some sensor fusion is done by preprocessors. Conductivity and temperature are used to measure salinity. The product of voltage and current measurement is used to indicate power. Etc.
A good deal of sensor fusion is done by A.s.a. H.'s concept structure* (at various different levels of abstraction).
* as an example see my blog of 4 September 2024.
Be it hardware, software, experimental procedure, etc. you need to provide enough detail that someone else could reproduce your work. Only then can others build upon and advance our science. You will then benefit in turn from THEIR contributions. i.e., Nash's cooperative games.
A non-Markovian process is one where future states depend upon the history or past, not just upon the current state. Jacob Barandes suggests** that "...perhaps when one takes physically fundamental non-Markovian processes and tries to shoehorn them into a Markovian paradigm, the result is quantum theory..."
* or by emergence at higher levels of complexity?
** Pilot-Wave Theories as Hidden Markov Models, arXiv:2602.10569, 12 Feb. 2026
It makes sense to develop industrial robots first*, their environment is simpler, more structured. And, again, how intelligent the robot needs to be will depend upon the task and the environment.
* Before home robots, for instance. (Of course many tasks at home are automated already: dish washing, clothes washing, drying, some cooking, robot vacuums, ....)
