cross-posted from: https://lemmy.today/post/48410284

Article: https://www.newscientist.com/article/2516885-ais-cant-stop-recommending-nuclear-strikes-in-war-game-simulations/

cross-posted from: https://lemmy.today/post/48472755

Humans have always made tools - it's why we have opposable thumbs along with the intelligence and dexterity to utilize them. Spiders are likewise built and programmed to make webs, and beavers to make dams. However, tools were always supposed to be a means to an end. A human end, not inhuman end. An end that is beneficial to human wellbeing, not simply generating more money while relationships break down, happiness declines, physical and mental health deteriorate, and governments tighten their control over our lives.

Short-sighted thinking and human vices have caused technology to no longer serve human ends. It has instead become an overwhelming net negative to humanity for over a century. Time and time again, a technology has become dominant because it provides short-term convenience, efficiency, pleasure or money. But it always has a strong negative for society once widely adopted. What good is endless entertainment when you are less productive, less satisfied with life and far more likely to be depressed? What good is instant long-distance communication when you have fewer close friends and family? What good is easy access to all the written works of history when your reading level and attention span are shot from addiction to social media and nobody else can discuss them with you? What good is modern medicine when it can't fix the problems caused by modern food, microplastics and drugs in the water and ever-present radiation? And what good are cheaper products when the actual things you need for a fulfilling life can't be bought?

Despite all these problems arising from apparently wholesome technologies, new technologies continue to be promoted that have much more obvious dystopian overtones. These include self-spreading vaccines, genetically modified insects, VR headsets, sex robots, lab-grown babies and brain chips. Yet there is one threat that is greater than all of these combined - one that could end all human life completely. Generally accessible weapons of mass destruction.

The threat of extinction

You see, we know from experience that technological progress enables things to be done more efficiently, easily and cheaply. This has been the case with weapons too - killing large numbers of people has only become more efficient, easy and cheap. Instead of relying on spears to kill, we developed guns, then canons, then bombs, then nuclear weapons, each one requiring less cost and effort for each person killed. Defenses against these weapons haven't advanced even a fraction as quickly, as it is much harder to protect than destroy. Nuclear weapons have also become more destructive and easier to produce than they were originally.

The average person too now has more ways than ever to kill others cheaply, using a gun, a car, or even a cheap drone with weapons attached. Individuals can even design, share and build their own weapons and weapon modifications at home using 3D printers. It therefore seems that if technological progress were to continue indefinitely, and humans continue to exist and have a small measure of freedom, a weapon capable of ending all human life on the planet would eventually become easily accessible to the average person. Then all it would take is one particularly angry, evil, inebriated or mentally ill person to put such a weapon to use and humans are no more.

That prospect might seem like a long time away, but it almost certainly isn't. You see, AI is now able to form coherent sentences and images. Fairly soon it will likely be forming coherent virus genomes and nuclear blueprints. It has already become better than humans at specific scientific tasks like predicting protein folding. AI doesn't need to achieve super intelligence, general intelligence, sentience or the singularity. It only needs to get close to human intelligence in some areas of science or engineering and then anyone with money to provide it materials may be able to accomplish decades of progress in a single year.

Some fields may require expensive physical or biological experiments to arrive at a generally accessible weapon of mass destruction, but others likely would not. For example, the creation of self-replicating robots would not require any exotic materials or scientific experiments, just clever design. If these robots use common materials that occur in nature or human settlements then they could quickly outnumber and exterminate all humans. To give another example - we have already modified harmful viruses to make them more infectious to humans, and some pathogens are 100% fatal to humans. Therefore, we are probably not far from being able to design a pathogen that would be capable of infecting and killing every human on the planet.

In conclusion, if ordinary people are free to develop AIs, open source AIs can (and will) be developed without alignment to any particular ethics, and anyone wishing to end humanity can attempt to fulfil their wish. Consequently, the attempts will continue until they succeed in extinguishing humanity or humans are so decimated worldwide that they're no longer able to run such powerful technologies.

The totalitarian trap

As technology gets more advanced it's going to be increasingly obvious how dangerous it could be in the hands of a bad actor. Therefore, governments will no doubt introduce restrictions on the public's access to technology - e.g., by criminalizing development or use of an AI without government certification and attempting to monitor all computer activity, even offline, to prevent the illicit activities. This will advance the surveillance state while enforcing an oligopoly over AI and other powerful technologies, centralizing power into the hands of a few who run the governments and big corporations.

No government or small fraction of the population can be trusted with such great control over technology, which could easily (and definitely would) be used for totalitarian subjugation. Technology is the ultimate power in today's world, and those without control over the technology would have no possibility of overthrowing the few who could effortlessly use AI to direct a vast army of robots, personalized propaganda regime, individual brain wave monitoring and constant video surveillance analyzed in real time. It is simply unrealistic to imagine the most powerful technologies being limited to the hands of a few and not being abused for mass domination.

Eventually, this course of events also leads to a near extinction event as over time the few with power are replaced by their offspring or there are internal battles for dominance. With changing hands of power and high stakes conflict it's only a matter of time until one group decides to end it all or something goes wrong and power falls into less judicious hands.

So what's the solution?

It is evident there must be restrictions on technology if humanity is to exist in 100+ years from now. But these restrictions should not be enforced from the top down by governments or any other group of a few. Not only would this lead to a huge centralization of power and near (if not total) extinction of mankind, but the public would clamor for the technology they are denied and see exploited by the few.

Having rejected centralised restrictions on technology then, the alternative we are left with is decentralised restriction. This could include boycotts, agreements, social stigma, parallel economies, civil disobedience and more, with the goal of limiting the development, distribution or adoption of anti-human technologies. For this strategy to be effective at stopping the development of AI and other dangerous technologies, it would likely require a majority of the population in each of the most significant countries to be convinced they are a serious existential threat to humanity.

The number of people to be of this opinion has been growing in recent years as technology has become more advanced and dystopian, so this goal may in fact become feasible as things get worse. However, most of those people currently do not see this solution to the problem, so do not have strong incentives to take action like boycotting AI or developing parallel systems. Many think that Pandora's box has been opened and cannot be shut. But that's not the case. The future of humanity is for humans to decide - there's nothing that can't be undone if enough people want to undo it.

"There's no way this could ever work"

Nobody thought it would be possible to end slavery either until it happened, or end the Roman Empire, or end Catholic dominance in Europe. The cult of technological progress at all costs is just one more thing that is dominant today, but it didn't use to be, nor is it our inevitable future. It may seem like a long shot, but we have to fight it by growing our numbers before it's too late - there is no better option. Rather than giving up or pretending everything will be fine, there is in fact something we can actually do that will at least push humanity in the direction away from disaster. Namely raising awareness of the problem and being part of the decentralized solution. Doing this may actually be rewarding and personally beneficial, as you will learn to be more independent, form new communities, and save yourself from the exploitation and mental deterioration that comes with much of today's technology.

As a final note, remember this is a battle for the survival of the human race - as many people as possible need to be brought on board. Therefore, we cannot risk to be divided over other issues - as important as they may be right now, they won't matter if mankind isn't around anymore. So whoever you are - wherever you may be - you have been placed in this important time for a reason. We hope you will join us in saving the world!

~(Image~ ~source)~