Let the creatures sleep!

A proclaimed aspect of artificial creatures is that they can work without pause, they are never tired and do not need sleep. A stark difference with non-artificial creatures. Sleeping takes up a big part of our time alive, wherein we lie still and unconscious. Artificial creatures can use this seemingly wasted time productively instead.

Or does this greed for productiveness have a counterproductive effect? What if sleeping, including dreaming, can unlock the full potential of artificial creatures.

For humans, sleeping is necessary to stay alive. During the night, our body can rest and recharge. However, when we enter the rapid eye movement (REM) sleep, our mind becomes more active and generates dreams (Peever & Fuller, 2017). This cognitive process plays an important role in storing and recalling memories (Sejnowski & Destexhe, 2000). It is during this time that our experiences of the day are consolidated and become connected to our existing knowledge, which enables us to remember, learn and understand.

Many artificial creatures also have a sleep mode they go into when they have mothing to do or are low on energy. For example, the pet robot Aibo returns to its charging station when it is tired and needs energy. The electronic lifeform ELF has a sleep routine that starts when the battery level is low or if it is dark, it then returns to its nest to sleep and charge.

Aibo

Electronic lifeform ELF

The rest and recharge benefits of sleep are clearly taken advantage of by artificial creatures. At the same time, the artificial creatures often fail to benefit of the cognitive advantages of sleep. They do not reflect on the data gathered during the day. Often, they do not record memories at all and simply start over when their code has ended. Because of this, they miss out on new insights. In the paper "Speculations Concerning the First Ultraintelligent Machine", Good (1966) already mentioned that dreaming could be an essential aspect in strengthening the unconscious and preconscious parts of an artificial mind.

An attempt has been made to use this principle in artificial creatures by the EU-funded DREAM project. Here, dream-like experiences were developed for robots as a way for them to order information. Contrary to the previously mentioned artificial creatures that exhibited sleep behaviour, in this case, dreaming was only an internal computational process.

DeepDream output

I think it would be interesting to explore and expose the cognitive together with behavioural experience of sleeping in artificial creatures, making it possible for others to experience as well. Input an artificial creature gathered during the day could be effectively processed and ordered during their sleep. The process could be visualised as dreams (think of Google’s AI DeepDream) or unconscious behaviour, such as sleepwalking. New insights could subsequently influence the creature’s behaviour. Alternatively, an artificial creature could be sleeping calmly. It would lie still and unaware of its environment, besides their active internal processing displayed. This might confront human onlookers with the reality of their unconscious but productive sleeping hours.

Good, I. J. (1966). Speculations Concerning the First Ultraintelligent Machine. Advances in Computers, 6(C), 31–88. https://doi.org/10.1016/S0065-2458(08)60418-0

Peever, J., & Fuller, P. M. (2017). The Biology of REM Sleep. Current Biology, 27(22), R1237–R1248. https://doi.org/10.1016/J.CUB.2017.10.026

Sejnowski, T. J., & Destexhe, A. (2000). Why do we sleep? Brain Research, 886(1–2), 208–223. https://doi.org/10.1016/S0006-8993(00)03007-9

Aibo: https://us.aibo.com/

Electronic lifeform ELF: https://cyberneticzoo.com/cyberneticanimals/1996-99-elf-electronic-life-form-tony-ellis-british/

DREAM project: https://ec.europa.eu/research-and-innovation/en/horizon-magazine/robot-dreams-are-helping-them-learn-humans

DeepDream: https://deepdreamgenerator.com/