So, let's unpack the whole mess of human hubris surrounding everything #AI and why this is so dangerous. The hubris demonstrated by GenAI companies is of course well documented. But it's also very evident in the criticism, and in a way that is distorting the real issues here. 1/7
📌
The biggest danger usually is described as AI "surpassing us" or mankind "losing control" of ever-powerful AI which will then annihilate or enslave us and "destroy the planet". But there is a problematic subtext here regarding our place in the world and in the technosphere we have built. 2/7
The idea behind this is that we humans are a superintelligence controlling the planet by our technology. And now a new super-superintelligence will wrest this control from us and causes runaway destruction on "our" planet while we are relegated to the sidelines. But this is rather misleading. 3/7
The reality is that we already have very little control about our technology and its consequences. We have no control over the planet and understand precious little of it. Because of that we have caused very dangerous developments that already make us little more than bystanders. 4/7
AI doesn't need to "surpass us" or escape our control to cause untold harm. There are already more than enough examples how a mere statistical text generator can mess up people and societies. Because we really don't have much control about technology at all. And that's really the point. 5/7
All this talk about dangerous superintelligence, control and the planet obscures this very important insight. Humanity is floundering within a technosphere out of our control on a planet we barely understand. There is a lot of stuff that can really hurt us, and I doubt we know all of it. 6/7
This is the context of generative AI. It's the next example of humanity rolling out technology at scale without understanding the consequences. It might harm us directly or by societal and planetary changes. This is not because we lose control of either. It's because we never had control. 7/7
Good point.
Mir scheint als würden viele Technologien erst im Nachhinein als "unter Kontrolle" wahrgenommen. Als sie aufkamen, hatten die einen Angst, die anderen warnten davor und wieder andere verfielen in jubelnde Begeisterung. Mir macht Ai keine Angst. Aber es gibt durchaus Menschen, die mir Angst machen.
I'm still not sure what you mean by “statistical text generators.” Are you referring to LLMs or some prominent representatives of our species?
🤭 Why not both?
I disagree quite fundamentally with that. We have perfect control over the planet and technology. What we don't have is control over other people, because we'd like to retain some freedom, while also preferring to have control over other people. That contradiction is the fundamental problem here.
Earthquakes, hurricanes, heat waves, snow storms, tornadoes, microbursts, tsunamis, volcanoes, climate change, sea level rising, mass extinction events to name a few:
We don't control physics. But we do control the circumstances in which physics happens. And we have much greater possibilities than most people realize, because we don't coordinate on a planetary scale. Just look at what ancient people did BY HAND and compare with our puny efforts with what we have.
Scientific fact: Most of the world is not volcanoes. Don't go there. You can control that. You can even build barriers to decrease the area at risk from lava. Most of the world is not threatened by floods. Neither 100-year floods nor 1000-year floods or any other. Don't go to there, or build dams.
That's rather the point, isn't it? And sorry, totalitarian regimes actually do worse, so dictators won't create that "perfect control" either. And the idea that we have any control over the planet is rather optimistic given that even the fastest processes span several human generations.
Dictatorships are governments that often have the least (actual) control over the population. Most work serves no actual purpose, other than pleasing the dictator or his cronies and the dictatorship couldn't do otherwise, even if it tried. That's why dictatorships are so miserable most of the time.
The problem of control over technology is a coordination problem among people and indistinguishable from any other coordination problem not related to technologies - new or otherwise.
Another is that "having freedom" and "having control" can refer to orthogonally different domains. It's perfectly conceivable to want to have freedom in one domain, while being fine with relinquishing control in another domain. The problem is, you can't usually pick what happens in which domains.
Because a lot of domains are interconnected. Some for fundamental reasons, others for socially constructed reasons (you work for me, therefore I control your sleeping schedule etc.).
Might I recommend our book? The questions you bring up, like the control of technology, the power-related notion of intelligence, etc are exactly the starting points we take, too