Innovation

Por una inteligencia artificial con valores humanos

Jan 16, 2023 Ana Palacio

MADRID – 2023 puede resultar año de referencia porque la inteligencia artificial transforme la vida diaria. Así se expresó Brad Smith, presidente de Microsoft en un evento sobre la IA que organizó el Vaticano la semana pasada. Pero la declaración de Smith no fue tanto una predicción como una llamada a la acción: el evento (al que asistieron líderes de la industria y representantes de las tres religiones abrahámicas) se inscribe en el objetivo de promover un enfoque ético y antropocéntrico para el desarrollo de la IA.

Es indudable que la IA plantea tremendos desafíos operativos, éticos y regulatorios. Y darles respuesta no será en absoluto sencillo. Aunque el desarrollo de la IA comenzó en los años cincuenta, aún no se han definido el concepto general ni su impacto probable.

Algunas facetas del inmenso potencial de la IA se han conocido por significativos avances recientes en el campo, que van de los textos escalofriantemente «humanos» de la herramienta ChatGPT (creada por OpenAI) a aplicaciones que pueden acortar -por años- el proceso de descubrir nuevos fármacos. Pero todavía es imposible prever todos los modos en los que la IA transformará nuestras vidas como humanidad -y, más ampliamente, la civilización-.

To continue reading, register now.

Subscribe now for unlimited access to everything PS has to offer.

As a registered user, you can enjoy more PS content every month – for free.

Already have an account? Log in

Subscribe to PS Digital

Access every new PS commentary, our entire On Point suite of subscriber-exclusive content – including Longer Reads, Insider Interviews, Big Picture/Big Question, and Say More – and the full PS archive.

Subscribe Now

new comment has been posted. new comments have been posted.

0 Comments on this paragraph, 2 Comments on this article

Robert Johnston Jan 22, 2023

Humans have never been very successful in regulating "human values" among humans. No reason to think they will any more successful in regulating "human values" in AI.
A new reply to this comment has been posted.
Prashant Kotak Jan 17, 2023

The idea of regulation of *any* type in the context the author suggests is outright nonsense. Let me try and explain why.

In 1905, Einstein published his series of papers that ignited the nuclear age. From that standing start (there were no incremental developments prior to put even the possibility of nukes in the imagination of anyone before that), in under four decades, the US had a working A-bomb. There are a number of implied lines about the nature of future developments from this story, too complicated to discuss in detail here, that people are simply ignoring, even if they have an inkling about them. The shocking thing is that Einstein, prior to his fame, was essentially a lone individual, no super-stellar academic background, no high tech labs, no huge teams of high quality colleagues, no access to the endless corporate money-well - who produced his masterworks as a pure construction of symbols, working in his spare time. Einstein in effect created a new reality. And he did it in the context of science and mathematics than were more complicated and less open than the technologies AI researchers nowaday work with - many many AI projects are open source and smart people who can understand what the technologies do abound around the world.

People outside tech don't fully understand this, but programming creates *precisely* the same opportunity, to create new realities in abstract domains (in fact game developers do literally that) but this time open to many many more people. Of course the same opportunity is available to national military establishments and to corporations and so on - in spades. But the opportunity exists for individuals and small collaborations as much, regardless of gender, nation or creed. Big corporate science, the politics, the culture stuff, morals, values, rules, regulations, etc can *all* be completely bypassed if the right individual or group of individuals or small obscure company comes along who are smart enough or simply stumble upon breakthroughs.

The point I'm very awkwardly trying to make is, nuclear technologies required high-end precision mechanics, device physics and electronics. They required teams of physicists and rocket scientists. Very few countries had the necessary scientific heft, so the barrier to entry of who could create nukes persisted, right to the early 2000s. That natural check on nuclear tech, and the fact that the US (rather than the nazis or the bolsheviks) got to create nukes first, was pure luck, but people don't see it that way. But these natural checks simply don't exist with cutting code - anyone can do it, all you need is a laptop, the brains, and the will. Virtually everything you need to kick-start is available for free - training resources, compilers, dev tools, the open source projects, the lot.

So what does the author suggests? Send teams of doddering old Brussels bureaucrats to make daily trips to software companies around the world (stay at the Hilton, all expenses paid of course) to stand over the shoulder of millions of programmers while they cut code to make sure they don't do something untoward? Or create AI policing software, to be henceforth compulsorily installed on every PC sold, so governments can keep tabs on millions of coders across the world?
A new reply to this comment has been posted.

https://prosyn.org/mow4pQnes

Por una inteligencia artificial con valores humanos

Subscribe to PS Digital

New Comment

Contact us

Log in/Register

Por una inteligencia artificial con valores humanos

Subscribe to PS Digital

Edit Newsletter Preferences

Set up Notification