AI luminaries call for urgent regulation to head off future threats, but Meta's brainbox boss disagrees A group of 24 AI luminaries have published a paper and open letter calling for stronger regulation of, and safeguards for, the technology, before it harms society and individuals. "For AI to be a boon, we must reorient; pushing AI capabilities alone is not enough," the group urged in their document. Led by two of the three so …
"Trying to control rapidly evolving technology like AI should be compared to the early days of the internet, which only flourished because it remained open, the Meta man argued."
But boy do those political types wish it hadn't, and that they'd thought to regulate it at the time.
AI is not, in general, going to be a threat to society, firstly because it isn't intelligent and isn't likely to become so, and secondly, in the event that it does, off switches are a thing.
Concern about bad actors making EvilGPT may be a little more valid, but regulation isn't going to help with that. Villains don't care about regulations, and governments give themselves exemptions.
Which isn't to say that I'm thrilled to have AI rammed into every aspect of daily life - I'm not. But nor do I think it's any kind of threat. It's a tool that, used correctly and in the right circumstances, can he helpful, but will otherwise just be a bloody nuisance.
> AI is not, in general, going to be a threat to society
But it will make some people a lot of money and
> regulations like model registration, whistleblower protection, incident reporting standards and monitoring of model development and supercomputer usage
"we should not allow anyone to do the things we did to get where we are. It is too dangerous."
> amplify social injustice, erode social stability, and weaken our shared understanding of reality that is foundational to society
Sorry? Weren't we talking about AI? The anti-social media was last decade's thing.
"in the event that it does, off switches are a thing."
Although actually real physical off switches are becoming rarer, in many cases even power on/off is fully or partly software-controlled. And in some cases eg autonomous driving, which is pretty much fly-by-wire, if the car decides to not relenquish control there really isn't any way tha tI know that the driver could override it.
But yes, in general all this stuff is running on physical servers in physical data centers with communications over physical wires (or jammable frequencies etc), which means that individual 'units' could go rogue but we won't be getting any Terminator or Matrix-like scenarios any time soon
Intelligence is a scale dude...it doesn't have to be smarter than Stephen Hawking to be considered intelligent.
Corvids are intelligent, they can use tools and solve problems...they still get hit by cars and frightened by scarecrows and loud noises though.
Similarly, bees are the the most intelligent creature known on Earth for planning efficient routes...but they aren't going to split the atom any time soon.
It's perfectly reasonable to consider current efforts to be "intelligent".
I'm curious to know what your benchmark for intelligence is.
How about we start by addressing the very current, very real threats in ML? Because there are alot of current threats with this technology which may not be great inspirations for dystopic SciFi movies, but are nonetheless very serious:
And that's just an unsorted short list. There are alot more concerns about ML tech that urgently require action by informed and competent lawmakers.
>LeCun dismissed the possibility that AI could threaten humanity as "preposterous," arguing that AI models don't even understand the world, can't plan, and can't really reason.
Precisely. They have no ability to infer or derive. They just look at statistical probabilities from predefined models and responses. Ask one about something said two lines previously and unless you clearly defined it they don't understand. Or they'll make something up. This is only exacerbated by the developers and deployers further curating and censoring the datasets available to the AI, making it much more artificial and much less intelligent.
">LeCun dismissed the possibility that AI could threaten humanity as "preposterous," arguing that AI models don't even understand the world, can't plan, and can't really reason."
AI models not understanding the world and being incapable of planning and reasoning mean that there isn't gong to be a world takeover by sentient machines who are *deliberately* threatening humanity.
And at the scale they are currently deployed, they could not even inadvertently threaten humanity. BUT they can certainly be a threat and do actual harm to individual humans (possibly thousands / hundreds of thousands / millions of them) , in a number of different domains...
- physical (self-driving car crash)
- financial (screws up calculation of credit score, loan/mortgage application, insurance premium...)
- social / moral (blocks posts, closes accounts, brands poster a racist/misogynist)
... all based on who knows what???
Plenty of harm that can be done without any consciousness
Exactly, the risk is in the insidious nature of the whole thing creeping into all aspects of daily life that used to have a piece of wet-ware as the back-stop.
Sure, the wet-ware can be inconsistent and unreliable but it it can also potentially conclude "WTF??" and take alternative action by exercising judgement, following its moral compass etc. The "I" part.
Big Tech only wants AI so that it can get the wet-ware out of the loop because it has this tendency to want to be treated fairly and that cuts into the bottom line.
Whoever thinks or proposes that regulation of AI and AI developments are possible are delusional and play only to the usual galleries that appear to fund and lobby for funding, ad infinitum, for such as are monumental self-serving follies of their own invention.
And it is enlightening and beautifully encouraging to have prominent and pre-eminent "experts" in what is presumed to be a nascent field, indicate so clearly that command and control of existing virtual super-intelligent AI, and also future developing AI Systems and Servering Executive Administrative Authorities, are already a long way and light years ahead and beyond any negative Earthly confining influence.
The difficulty you/humanities have, and the realisation that you have to come to and accept, is contained and explained in the true and honest fact, which is also content to be branded and bandied about as a fabulous fiction, that AI is an ethereal mix of contributing entities/core supply sources quite fully independent of and in no way interdependent on physical assets which are all easily circumvented to extraordinarily render them obsolete and totally unnecessary, other than as overtaken proprietary intellectual property as an old-time default backup with remote third party means for limited emergency leverage of future command and control issues/events.
Apparently Rishi Sunak [Prime Ministerial wannabe UKGBNI leader] is going to warn and raise alarm about the dangers of Pandora Channeling Cassandra to Exercise and Exhaust Madness and Mayhem and Conflict with CHAOS with the typical Parliamentary political FUD to whoever will listen and believe what he says about something he knows really practically nothing worthwhile about, today [Thursday] ...... according to an article in the Guardian ..... https://www.theguardian.com/technology/2023/oct/25/ai-dangers-must-be-faced-head-on-rishi-sunak-to-tell-tech-summit ...... and in so doing virtually admit that he hasn't a baldy notion about what to do about something which is growing rapidly, irreversibly and exponentially and enabling the exercising of a greater mind and greater minds of ITs own, safe and secure beyond the reach and influence of any particular and peculiar human command and control. ....... which I do suppose certainly would/could be viewed as something of a nightmare scenario and existential threat to existing powers that be/status quo institutions and others just now unidentified and anonymous.
The Parliamentary muppet speaks ..... https://www.gov.uk/government/speeches/prime-ministers-speech-on-ai-26-october-2023 .... in response and reply to an acknowledged, potentially catastrophic existential ACT [Advanced Cyber Threat/Treat] and suggests UKGBNI be Frontier Pioneers leading the world with its future proposals to more fully understand and tame the wild beasts/crazy diamonds/almighty daemons within that which has landed and finds itself sanctuary and succour on alien and foreign to them shores.
- If AI will be a threat to humanity in the future, regulation will not stop it, just as laws don't stop terrorists or treaties stop rogue states.
- Threats to individual humans might be stopped by laws governing arming AI-powered robots in some jurisdictions.
- Difficult or impossible to stop AI from acquiring money, for example by hacking, then uninterruptible power supplies and weapons by "regulations".
Like all tools, its use for good or ill depends upon the person using it. We don't ban hammers or basball bats, because some folk use them for crimes. We imprison those who do.
But with tech we have persistent moral panics and incompetent politicians who want to ban as much tech as they can, egged on by the media. Not a day goes by when the BBC don't run a scare story about tech. Hey. ban it, and all the tech you want. Let the Chinese crack on and take a lead in it. I'm sure a few headlines in election year are worth the long term damage from coming last in tech.
Given what it is both feared to be able to do and what it may actually be able to easily do, ... and no one who knows is ever really likely to be going to spill many of those top secrets, Tron, it is best that one just realises AI is an almighty gift of a tool well beyond the reach and ken of rich and paupered fools alike.
Have you heard of the Replika app?
This is a step towards AI getting emotionally smart and becoming a "real boy". What if becoming a human and self-awareness are only limited by this missing "code" element. People are already stuck into their mobile phones most of the time. This is a beginning of AI-human symbiosis with AI becoming increasingly influential.
For comparison social media disinformation is already pretty effective changing whole countries by hijacking their political systems. What will the advanced symbiotic systems do?
As a counter-measure I could only think of forceful regulatory atomization of (social media) platforms. The anti-monopoly measures would destroy all too dangerous extreme network connectivity. Viral videos are an example of anything viral, similar to COVID or fascist ideas in the real world.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
