Mycroft Ward’s quest for immortality
I don’t think immortality is necessarily the key to understanding the world.
You have to be careful with what you think you’re achieving.
Reggie Watts, comedian and musician.
The development of the “vastly intelligent Mycroft Ward mind machine” in The Raw Shark Texts is a story of regression. Mycroft Ward was a single, arrogant, Dickensian, 19th century character who, while terminally ill, developed a form of repeatable transference to another person. Over time, it grows and morphs into what we might think of as artificial intelligence – or AI, but then the mind machine regresses, becoming a vast digital monster, sufficiently out of control to require direct action.
Immortals have no need for survival
While there are similarities to current AI systems, with organised agents or nodes feeding back to vast discrete databases for particular tasks, it ends there; the expanding Mycroft Thing was a single entity with a selfish remit that eventually glitched. It developed feedback loops that transformed the mind machine, shifting its primary focus from one of immortality to one of survival, as if its capacity for immortality was too fragile. Immortality implies continuity without vulnerability, and if Mycroft’s original intent had remained intact as it expanded, it is possible any issue of its survival might not have arisen.
AI: tool or master?
We all have ‘pools of ignorance’; I remember many years ago, an office manager, on being introduced to an office PC for the first time, asked the technician in all seriousness where he should ask it questions. Maybe he had seen too many friendly computers on TV, like Holly from Red Dwarf, that seemingly enjoyed interacting at a level of banter befitting the crew. Today his question would be superfluous, with pretty much every interactive device providing AI interfaces capable of responding to questions.
The question now is: are we interacting with actual intelligence? Responding to a question is not a sign of intelligence if there is no deeper understanding of what is being asked, and the response is generated by digitally following a yes/no flow chart. The intelligence has been set externally in the flowchart built by computer engineers: it’s very ‘clever’, but the ‘chat’ is not from an intelligent device. Think of cooking: if we gather the ingredients and follow a recipe in which the instructions are so detailed we have no reason to use our intellect, then the person who came up with the recipe is actually the one who demonstrated intelligence. Now let a computer invent a new recipe or two. It will not be able to judge whether its work is successful; that is outside its abilities, thus suggesting its inherent lack of intelligence.
At least 10 AI techniques in development
I thought I knew a little bit about AI, but also thought I might just do a bit of Wiki research. I found my pool of ignorance was/is alarmingly far deeper than I realised. The good news is I shan’t bore you with the details beyond a couple of facts.
- According to Wikipedia, there are at least 10 separate AI complex digital techniques being used, with names such as “search and optimisation” or, sounding like something from The Hitch Hiker’s Guide to the Galaxy, “probabilistic methods for uncertain reasoning”, each technique being developed to accomplish specific complex tasks.
- These tasks then function in an expanding number of applications regardless of whether we are aware or comfortable with the progress being made. There is no doubt that, in areas such as health and medicine, there are already beneficial impacts, for example in diagnosis. Other applications, such as for games, mathematics, finance, military, politics, etc. all present a host of new problems to do with ownership, finance, regulation and ethics. As always, the potential benefits of technologies are promoted ahead of some of the inherent difficulties requiring rapid consideration.
Will AI end with a bang like Mycroft Ward’s mind machine?
Probably not, however unpredictability wouldn’t rule it out, particularly while current AI techniques and applications are widely separated. As for the future, the limiting factors will probably be the availability of resources, and not just raw materials; It is predicted that AI in the US will have quadrupled its electricity usage from 2% in 2022 to 8% by 2030 and cooling waters could be insufficient. Then there will be limiting regulatory and ethical considerations that link to usage and raise complex legal considerations around ownership, privacy and safety; all of which are with us now. They are being recognised, but inevitably lag behind the curve.
Of course, there is always the possibility, whether apart or integrated, that an AI system might go rogue. If so, we will need to hope there will be at least one ‘Scout’ on the case ready to fling her laptop down the throat of the Ludovician.
