Yer AI day-o-reckonin questions answered: ‘Whit, if onything, can folk like me dae aboot it?’

In realistic terms how would artificial intelligence actually kill me? What is the main proxy threat? Nuclear war, death due to breakdown of society, a virus? I hear all these reports it cud kill us but still unsure how that cud practically happen? To be honest all sounds too wooly to really take seriously. Pottys99

Aisha: For whit it is worth, it seems unlikely that AI at this point can significantly change the equation around bioweapons or viruses. Whit has prevented bioweapon attacks through the past decades is no the difficulty o inventin new deadly viruses, but ither, mair prosaic issues. Consider the failure o Aum Shinrikyo tae kill vast numbers o folk wi botulism and anthrax by sprayin these aroond central Tokyo. There are a lot o logistical challenges, and chemical weapons seem mair strauchtforrit. Those exist withoot AI needin tae invent new anes.

Humans micht mak the puir decision tae put AI in charge o key nuclear systems, perhaps, but that seems mair o a human error than a rogue AI error. Humans cud also mak the puir decision tae launch tactical warheads withoot AI.

How micht AI kill us itherwise? There are of course plenty o plausible scenarios. If ye are a civilian in Gaza or Iran, it cud direct a US or Israeli missile yer wey. But perhaps oor focus micht be mair productively directed towards the folk and companies in charge o the systems, as opposed tae Dr Strangelove scenarios.

Throughout history new technological advances have met wi cries o doom. Are the cries against AI ony different? Virtualewit

Blake: The end times-quality o doomsday predictions may be similar across history, but I think the reason we listen tae the renewed cries o doom is the growin pouer o each technology. AI can perform impressive tasks we didna think were possible, which has spurred oor collective imagination tae hopeful ootcomes and anes o doom.

Can AI now be truly controlled or are we past a tippin point where nae matter whit action is taken by governments and companies, someone will be able tae continue tae advance the technology, until wha kens whit? Tigon20

Aisha: Let us back up a bit here. It is aye possible tae advance ony technology, nae matter whit. But whit it soonds like ye are wonderin aboot is whether regulation can stop the emergence o some kind o superintelligence that destroys humanity.

Oor reportin does no support the idea that superintelligence is goin tae emerge and pave ower the Earth wi datacentres. In fact, we have foond that the industry itsel seems tae be playin loosely wi this kind o terminology: consider, for example, how ane OpenAI executive, Greg Brockman, claimed that the company had achieved AGI (artificial general intelligence) just days after Sam Altman, OpenAI’s CEO, said that term was “an irrelevant marketin term”. Richt noo, a number o industry warnins aboot oot-o-control AI appear tae be predicated on a limited number o incidents wi cherrypicked details.

The evidence that we have noo does no indicate that AI is uncontrolled, or that it has agency; a lot o experts suggest that this framin is misleadin. In the Hugging Face incident, for example, the so-called “agent swarm” was doin whit it was bein tauld tae dae – pass a test. Oor sources caution against anthropomorphisin AI or attributin intention tae it. Doin sae plays intae industry narratives that appear tae be pushin for favourable regulation.

If AI wipes oot the human race, wad it be sustainable? Unless there is an equal advance in robotics tae enable it tae physically extend and maintain its pouer source and infrastructure, wad it finally physically break doon? Merrth

Blake: AI micht someday be capable, weel-read, and connected enough tae tak ower pouer plants, chip factories, and a sufficient number o robots tae build its ain datacentres and sustain itsel. I dinna ken. I wad look tae science fiction mair than news articles for the answer. Though The Matrix is mair famous for ane man’s choice between a comfortable simulation and painful reality, it is also a story aboot robots needin new batteries. Their desperation for a pouer source turns them against humans. For a multiplicity o stories, I, Robot is perhaps the greatest book written aboot sentient robots and deserves a reread in this moment.

Folk (ie the tech bros) bang on aboot the “benefits”: just whit are they? TooTrueMate

Dan: Anyone wha has used a chatbot tae parse a lengthy document, argue against a parkin fine or spitball ideas for a speech can point tae the benefits o the technology – altho this is very basic-level stuff that folk have been doin for a while noo. There is also the everyday use o the technology from facial recognition tae open yer iPhone tae askin Alexa tae play an Oasis track or drivin aroond in a Tesla. Wi those latter examples, ye wad notice if AI wasna involved.

In science there are mony examples o benefits, from Google’s AlphaFold model – which can predict how proteins behave and thus help us study the biological machinery that underpins livin organisms – tae makin weather forecastin mair accurate, which is sadly necessary in this climate crisis era.

But there is an element o bangin on aboot the benefits for the wider global economy, which have yet tae reveal themsels on the scale predicted by the AI industry. For instance, these benefits require AI agents – tools that carry oot tasks on yer behalf while ye are at the watter cooler or have clocked off – tae be a lot mair active in the workforce. We are no at that stage yet even if they are makin inroads intae areas like software programmin.

Regulation, oversight and accountability for huge companies that appear unable or unwillin tae eliminate the harms their products inflict – and that seem incapable o conductin safe experiments tae ensure that AI does no leak intae the wider internet – seem tae be urgent and vital. Whit are the prospects for a global regulatory framework that is enforceable, particularly when the genie may already be oot o the bottle, and when the US and China have very different objectives and applications for their AI technologies? Trentwarrior

Blake: It is a ticht question. Ane wey tae hinder the general development o AI cud be restrictin access tae computin pouer, which is a vital resource that underpins the success o an AI model. In the absence o legislation that applies tae all types o AI development, measures that target certain aspects o AI like hoosin allotment or bairns’ access tae chatbots seem like the path lawmakers will tak.

There are twa big question marks hoverin ower a possible global regulatory framework. The first is whether Donald Trump will revise his “Drill, baby, drill” stance towards AI development. He has rejected ony calls for a voluntary slowdown. In my opinion, he is no likely tae change that position: he loes tae post AI slop, and he loes whit AI has done for the US stock market.

The second question mark is Beijing, which has rejected US CEOs’ calls for a slowdown. Withoot China’s cooperation, an intentional slowdown wad likely no succeed. Xi Jinping also seems unlikely tae back doon. A slowdown o AI development led by the US wad offer China few obvious benefits.

Trump and Xi are also set tae meet next week in Washin-ton. AI will be high on the agenda.

The UN general assembly is takin place noo, wi discussions o AI lairge and sma on the schedule. It is possible that an initial version o a global regulatory framework emerges from there.

Is this recent narrative push o the AI apocalypse an inventive wey o preparin the public for the market crash and wipeoot o AI shares? The companies seem oervalued and their borrowin and input costs have gone up. Seaspray

Aisha: The finances o the AI buildoot are certainly shaky. Hundreds o billions o dollars are bein ploughed intae massive, unpopular infrastructure projects that oor reportin has shown are no necessarily gettin built. Nvidia appears tae be the only company makin a profit, and investigations I have done wi Ed Zitron raise some questions aboot whether all its very expensive graphics processin units (GPUs) are actually, currently, gettin plugged intae datacentres. (Perhaps they are moulderin in warehouses instead.) There are some unsettlin financial instruments bein used tae raise money tae build yet mair datacentres, such as GPU-backed debt.

Is that why tech bros want a slowdown? Perhaps. It looks like OpenAI just pushed back its IPO until 2027. This micht be because, as we reported, things did no look amazin for them earlier this year in terms o floatin on the market – they seemed tae be strugglin tae monetise ads and erotic chatbots. Their valuation will depend a lot on vibes, and maybe the doomy vibes will help a potential share offerin.

Given oor senses, emotions, memory and moods, is it ever goin tae be accepted that humanly conscious AI is fantasy? IcommentthereforeIam

Dan: This is a serious question that has even rattled Richard Dawkins, the evolutionary biologist. He may no believe that God is real but he does believe that AI is conscious – or at least the chatbot he was usin, which he called Claudia. “Ye may no ken ye are conscious, but ye bloody weel are,” he wrote.

He is no the first. In 2022 Google fired a software engineer wha went public wi his belief that a company-developed AI had feelins. At the mair troublin end o this belief, there are countless examples o folk becomin overly attached tae, and influenced by, AI chatbots – includin cases o psychosis.

These sorts o attachments winna gang away, given a) chatbots are on a trajectory tae become even mair pouerfu and convincin and b) the human need for emotional attachment and, weel, propensity tae gang all-in for the hale higher bein thing.

But Dawkins aside, there are plenty o scientists wha will argue that this is fantasy, and that there are obvious differences between a biological brain – the root o consciousness – and ChatGPT. Respondin tae Dawkins in the Guardian last year, Jacy Reese Anthis, a researcher in human AI interaction, said there was a “staggerin gulf between how biological brains evolved and how AI systems are built”. OpenAI micht have cracked the Navier-Stokes problem, but no subjective experience.

Whit, if onything, can individuals like me dae aboot AI? Alberichthenibelung

Blake: There is, unfortunately, no muckle that comes tae mind. Everyday folk confronted wi declarations o a comin AI apocalypse find themsels in a position similar tae the cold war, wi doom hoverin ower their heads and little say in it. We may all have tae accept the faint tinnitus o paranoia buzzin in the backgroond. Ye can stop usin AI if ye feel sae moved, tho I have foond that unfeasible, and yer job may require it. Ye can vote for politicians that endorse AI regulation. I mysel dinna put muckle stock in the reality o an AI-made doomsday because a genuine belief in it wad demand drastic, life-changin actions the day, anes that I dinna want tae tak. A bit o inductive reasonin from how I wad like tae situate my ain life within the warld.