2026-10-24
We have been assailed by various articles of late (and not so late) along the lines of the existential threat that AI may pose to humanity.
It's a fair question. At what point could AI pose such a threat? What are the prerequisites?
Well, as of now, most AI systems are configured to answer our enquiries. They seem to be pretty good at being "search engines with some intelligence" that can pick their way through (a) what datasets do I need to enquire upon in order to answer the question? (b) which items in those datasets are relevant to the question? (c) how should those items be logically linked together to derive an answer to the question? and (d) performing that linkage and presenting a comprehensible analysis as end result.
The better ones might realise that there's more than one route through that process, would traverse each in turn and cross-check the results to identify any inconsistencies. How it might pick the appropriate answer given inconsistent results could obviously be problematic - a human being might seek out additional original documentation to do that but an AI might not have a full range of options to choose from.
And a human might realise that given the ways in which the parties concerned tend to operate, the proffered result looks unlikely to be true, whereas the AI would have no "life experience" against which to assess the result obtained. Indeed it has no "life experience" at all unless a tiny minority of determined enquirers (such as Mark Playne) decide to take it to task and ask it to re-evaluate against additional factors until it gets a better, or even just a different, answer.
AI can only work with what humanity gives it (if we discount invaders from outer space of course).
Getting the wrong answer of itself wouldn't allow it to pull the plug on humanity - to do that requires a whole new basket of tricks unrelated to AI but very much related to the plug chosen to be pulled. And there are not many that would have a wide-ranging effect:
a) Poisoning: would AI have the means to release poisons into the human environment sufficient to kill everybody? The corporate elites have been doing that for years (fluoride in water, chemtrails in the atmosphere, mercury and heavy metals in tooth fillings, vaccines, drugs of all varieties; cosmetics, artificial additives of many varieties in foods and beverages, pesticides herbicides fungicides insecticides and artificial fertilisers on farms and agriculture of all kinds, drugs both natural and artificial, lethal infections (although the science of these seems decidedly dodgy), parasitic infections (these do seem to be real, and quite possibly responsible for most cancers ... ). And yet we are still here.
b) War: divide the people into groups and set one group against another so that they go to war. Messy and unpredictable - and ultimately ineffective since humanity has already survived two or three world wars, so what's a fourth?
c) Global catastrophe rendering the whole surface of the planet uninhabitable. Maybe another Great Flood (didn't complete the job last time), or an asteroid impact, or a nuclear winter / ice age? If humanity can't do it, then how would an AI locked into a network of machines do it?
So given the current state of the art, I'd say that AI is as yet nowhere near assembling the wherewithall to actually do humanity terminal damage of its own rogue volition (although it remains open to some evil overlord to hook it up to weaponry which might do quite a lot of damage).
Having said all that, make up your own mind.
