Category: bevidsthed

The Alignment Problem: Machine Learning and Human Values

– relationen mellem “Kunstig Intelligens” og “menneske-intelligens”

The Alignment Problem: Machine Learning and Human Values (2020)
by Brian Christian

Finalist for the Los Angeles Times Book Prize

A jaw-dropping exploration of everything that goes wrong when we build AI systems and the movement to fix them.

Today’s “machine-learning” systems, trained by data, are so effective that we’ve invited them to see and hear for us―and to make decisions on our behalf. But alarm bells are ringing. Recent years have seen an eruption of concern as the field of machine learning advances. When the systems we attempt to teach will not, in the end, do what we want or what we expect, ethical and potentially existential risks emerge. Researchers call this the alignment problem.

Systems cull résumés until, years later, we discover that they have inherent gender biases. Algorithms decide bail and parole―and appear to assess Black and White defendants differently. We can no longer assume that our mortgage application, or even our medical tests, will be seen by human eyes. And as autonomous vehicles share our streets, we are increasingly putting our lives in their hands.

The mathematical and computational models driving these changes range in complexity from something that can fit on a spreadsheet to a complex system that might credibly be called “artificial intelligence.” They are steadily replacing both human judgment and explicitly programmed software.

In best-selling author Brian Christian’s riveting account, we meet the alignment problem’s “first-responders,” and learn their ambitious plan to solve it before our hands are completely off the wheel. In a masterful blend of history and on-the ground reporting, Christian traces the explosive growth in the field of machine learning and surveys its current, sprawling frontier. Readers encounter a discipline finding its legs amid exhilarating and sometimes terrifying progress. Whether they―and we―succeed or fail in solving the alignment problem will be a defining human story.

The Alignment Problem offers an unflinching reckoning with humanity’s biases and blind spots, our own unstated assumptions and often contradictory goals. A dazzlingly interdisciplinary work, it takes a hard look not only at our technology but at our culture―and finds a story by turns harrowing and hopeful.

 

Hands on a wall

– Cueva de las Manos, Patagonia

 

A young doctor friend of mine is cycling around Patagonia. 2 months of cycling in diverse landscapes, from difficult to very challenging.

He came across the Caves of Hands, where people over several periods 7000  years ago depicted their hands on cave walls. He was quite struck by it, as he recorded his experience in his tent out in the Patagonian wilderness.

What is this thing with the hand, he speculated.

The following flowed out of me:

Most people don’t understand the abstract, and will satisfy themselves with the concrete level of acknowledgement of their existence: The people around them, family and those recognised as tribe and under the same rules as themselves. Before what we call consciousness some people left their marks on the world around them on a symbolic level. People, whom this made sense to, were probably the tribes’ shamans.

A handprint on a cave wall could have a spontaneous mark, but the subsequent marks both acknowledged those gone before, “the forbearers”, and the value in making their own mark to those following in the same position as themselves: Belonging to a very small group of people, who understand the value of time – in a conscious or instinctive way. Who sees a handprint as both real and symbolic. Who sees the world both as real and symbolic. That everything can MEAN something – everything can have a value – and not just the value ascribed to it through need or use – but an inherent value. An immanent value. Something inside the thing, the event, the ritual, the process.

From this approach to reality language grows, and from language, greater consciousness and wider naming of things. And with this, the abstract world of ideas growing INTO normal human beings, filling them with a larger world, a world where ideas can come alive, where ideas can exist as entities, become matter and action. And begin to grow into justice and catering to individuals, which in turn begin to regard themselves as valuable and their personal ideas worth fighting for, growing into systems of belief – politics, forbearer worship, deities, aesthetics, respect of life, humility as value, violence as value, purity as value.

The hand (on the grotto wall) is probably the most powerful sign of humanity – the finding of roots, the drying, the grinding, the mixing, the painting of a colour, most often red, the hand on the wall, I am here, I was here, here were humans, history is a memory of humans, humans die, but leave traces of themselves. Humans matter.

Erfaring er algoritmisk ukomprimérbar

– Forståelse = eksistenspræmis + fare

Citatet tilhører science fiction-forfatteren Ted Chiang.

https://www.salon.com/2021/04/30/why-artificial-intelligence-research-might-be-going-down-a-dead-end

Sammenhængen er forestillingen om AI’er, kunstige intelligenser, som er ude af stand til at udvikle selvbevidsthed alene baseret på tanke – de er nødt til at have en mobilitet, en krop, at støde sammen med verden i for at kunne gøre sig sine egne erfaringer. Noget, der siger nej og ja, som der skal uddrages mening og forståelse af.

Erfaring.

—
“Erfaring er algoritmisk ukomprimérbar.”

Citatet er en Y-vej: Sende AI’en ud på vejene i en krop for at gøre sig erfaringer blandt mennesker, eller acceptere, at en Kunstig Intelligens “kreativt set” altid kun vil kunne løse opgaver – opgaver, der kan defineres inden for en (snæver) ramme. Underforstået:

At SKABE en reel AI (give en maskine en krop og evolutionær mulighed for at tage ved lære blandt Jordens andre livsformer) er dog fortsat sindssvagt risikabelt. Tænk Skynet. Men intelligens er et produkt af en virkelighed i egen højde – menneskeintelligens, egernintelligens, delfinintelligens, chimpanseintelligens, træintelligens – i mødet med de andre intelligenser/evolutionsmaksimeringer. Og som hjernen fungerer er det meste (95%) ukendt rum og kun 5% kan kaldes for “hidrørende fra bevidsthed”.

Det står ikke i artiklen, men med den absolutte amatørs henkastede selvfølgelighed vil jeg sige, at fraregnet den fysiske styring og regulering af kroppen er de 95% (den kaotiske elektriske aktivitet i hjernen) en sorteringsmekanisme, der “summer omkring data” – dvs. data fra kroppen og data igennem kroppens sanseapparat. En art si – der skaber og understøtter organismens virkelighedsopfattelse – og på dén måde et direkte udtryk for organismens evne til at flexe i forhold til ukendte indtryk inde- og udefra.

Det understøttes delvis af forskningen selv:

I modsætningen til den digitale intelligens’ tændt eller slukket-opfattelse af virkeligheden (ér det eller ér det ikke (noget kendt)?, befinder menneskehjernen sig et sted, hvor det er hverken eller. En NU-gråzone, som svarer til at hoppe fra alligatorhoved til alligatorhoved hen over en uendelig bred flod: Evolutionen tillader ikke hjernen “at blive stående”. Og selvom AI kan processere data 1 million gange hurtigere end menneskehjernen, kan den aldrig nå frem til hverken eller-tilstanden, før den selv kommer i fare for at ophøre med at eksistere.

Så, skal man indprogrammere sådan en forudsætning? Og skal man gøre AI’en mobil og autonom og dermed give den mulighed for at erfare selv? Og hvem skriver dens algoritmer for statistisk behandling af indsamlede data?

—
Asimovs “Robotikkens love” var et forsøg på at eksemplificere denne problematik ved at kaste dramatisk grus i samme maskineri og tilsætte lidt dramatisk menneskelighed og mulighed for (læseren for) at føle med robottens udvikling og overlevelse.

“Erfaring er algoritmisk ukomprimérbar” siger så, at det ikke er muligt at skrive en algoritme, der springer evolutionen over. Det lader sig altså ikke gøre at lave en fabrikation af robotter med Robotregler ift. mennesker, fordi regler/algoritmer ikke er lig den sameksistensmulighed, vi kalder empati eller forståelse.

Forståelse = eksistenspræmis + fare.

Menneskeheden har ikke råd til at lade AI tage ved lære. Menneskeheden på individplan kan ikke udtrykke kærlighed på et niveau, hvor en AI indprogrammeret med overlevelsesinstinkt IKKE automatisk ville lære at forsvare sig med VOLD.

Det er dét.
–

Umiddelbart vil jeg sige, at ikke AI og hjerneforskere ikke konsulterer kreativt skabende nok eller tager tilstrækkeligt med hallucinogene stoffer. I forskellen mellem det accepterede og det uforståelige kaotiske ligger et kæmpe potentiale, som kun kan afdækkes ved at lægge måleredskaberne væk og lade kaos komme til én.

 

Quora Q/A: Are the belief, faith, and believe the same?

– No, not even close

No, they are not. But you are missing some:

faith, belief, conviction, trust, confidence, hope.

Continue reading “Quora Q/A: Are the belief, faith, and believe the same?”

Kan du sige “antropomorfisere”?

Hvis du kan, er du blandt et voksende antal.

Det betyder “at menneskeliggøre”.

Har du nogen sinde set naturprogrammer, hvor dyrene eller planterne eller naturfænomenerne tillægges menneskelige egenskaber, følelser, intentioner, forståelse, tilbliven osv.? Det er meget almindeligt at høre speakeren sige: “Og når den så har spist, vender aben tilbage til sin hyggelige hule”.

Et dyr er ikke et menneske, og selvom vi kan observere, at visse dyr f.eks. synes at underholde sig selv og drille artsfæller på en måde, der minder om menneskelig adfærd, betyder det ikke, at dyr kan have samme ansvar for deres overordnede eksistens som vi for vores. Allerhøjst kan vi udlede nogle ting omkring forløsning af energi gennem “leg”, som vi og visse dyr måske har til fælles.

Slappe narrativer, hvor observationer af naturen antropomorfiseres uden grund i videnskaben, har den uheldige virkning at tilskuere – snarere end at komme tæt på menneskers afhængighedsforhold til Naturen – skubbes “den anden vej”, tilbage til at bekræfte en meget gammel fordom om mennesker som mere værd end klodens andre arter og naturen som helhed. Continue reading “Kan du sige “antropomorfisere”?”

One Voice Less

A very very loong list of scifi and fantasy in development (2018).

(Almost) Every Sci-Fi/Fantasy and Comic Book Adaptation in the Works

For some, probably the dream of imaginings – for others, maybe part of the problem, but here is a 2018 list of  planned scifi and fantasy series and films dreamed of. And WHAT a list!
—

Sure, there’s a market and investors and actors and an industry needing the work – and many series and movies are becoming less sexist and open to identification on a human level, which is a relief – but it’s hard not to see this collective drive as one long Creative Scream for another world. A manageable world. With comprehensible endings and right beginnings. With coherent reality. With identifiable villains. With heroes of human strengths and weaknesses. Just something other than this mess of an Earth, which many are seeing at present.

Continue reading “One Voice Less”

Rationality – From AI to Zombies by Eliezer Yudkowsky

Rationality From AI to Zombies
by Eliezer Yudkowsky

Buy for free or more!

Released under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported license
⃝cc BY-NC-SA 3.0
http://creativecommons.org/licenses/by-nc-sa/3.0/

Contents Continue reading “Rationality – From AI to Zombies by Eliezer Yudkowsky”