Here is my annotated and augmented summary:
|
|
| ©NZZ |
Jacob Coxon, an AI researcher at Anthropic, resigned from his position criticizing
both Anthropic and OpenAI for "gambling with our lives"
When Jacob proclaimed his Badspel, CNN anchor Anderson Cooper looked worried, well Anderson always looks worried.
As the discussions turn toward doomsday scenarios, Jason Abaluck,
economics professor at Yale University, posted:
Jason is mostly afraid of RSI, the Recursive Self-Improvement, of one AI
system dominating not only all the others but the world?
One of the best-known doomers is AI developer
Geoffrey Hinton, who asserts, “The only thing that can force these companies to do more
safety research is legislation.”
“Never has a less intelligent thing controlled a more intelligent thing. Developing a powerful AI is like getting a tiger cub. If you can’t be very sure that it won’t try to kill you once it grows up, you should be worried.”
Dario Amodei left OpenAI after sharply criticizing the irresponsibility behind developing the ChatGPT agent. He founded Anthropic and he now drives the development of the Claude agent. He is a doomer, yet he sees both the dangers and the potential of AI. Amodei advocates for slowing down AI research to buy time for more safety reviews.
Elon Musk who had been sounding the alarm on AI for a long time admitted recently in an interview with The Economist, “I don’t see any way to really stop this incredible surge in AI and robots.”
Then Elon went on to explain his shift from an AI doomer to an AI
advocate, saying that within the framework of techno-libertarianism, he
arrived at the ”philosophical” conclusion that “even if there were a stop
button, we probably shouldn’t press it, because the most likely outcome is
incredible abundance for everyone.”
“Future humans will no longer have to work. Their lives will be funded by an ‘unconditional basic income.’”
Doyan Marc Andreessen, co-author of the first web browser Mosaic, is even more optimistic than Musk. In his essay “The Techno-Optimist Manifesto,” he writes: “We believe that artificial intelligence is our alchemy, our philosopher’s stone — we are literally making sand think.”
As a result of the “Law of Accelerating Return,” according to which technological progress generally reinforces itself, “more and more becomes possible with less and less. Prices fall, purchasing power is unleashed, and demand for new goods and services rises.” This results in exponential growth. Andreessen’s view is in line with the radical current of techno-libertarianism known as “e/acc,” or effective accelerationism.
Sam Altman, the CEO of OpenAI, sounds far more utopian in his essay “The Intelligence Age.” He is convinced that AI can tackle climate change, explore the entire field of physics, and establish a space colony.
After all, Sam, too, believes in exponential growth according to Moore’s
Law: “Semiconductors roughly double their performance every two years
while prices remain constant. Computers and smartphones get better with
every new model, and their processors take up less space, while the price
of a new device stays roughly the same*.
*Roughly the same? The iPhone 18 Pro Max is 150 euros more expensive than last year’s model.
He concludes his remarks, “A defining feature of the Age of Intelligence will be massive prosperity.”
*Roughly the same? The iPhone 18 Pro Max is 150 euros more expensive than last year’s model.
He concludes his remarks, “A defining feature of the Age of Intelligence will be massive prosperity.”
Alongside the doomers and utopians,
Gary Marcus, professor emeritus of psychology and neural science at New York
University, considers the current discourse to be exaggerated: “The
idea that AI will wipe us all out in five years is completely absurd.”
Timnit Gebru, founder of the Distributed Artificial Intelligence Research Institute, wrote in Wired magazine: “The industry has become a collection of cults and die-hard believers. Like a preacher proclaiming that the end of the world is nigh.”
How true.
AI researcher Melanie Mitchell demonstrated that the AI agents developed by OpenAI and Anthropic do not exhibit malicious behavior by chance, but rather as a result of their training. These AI companies didn't lose control of their systems; they trained them incorrectly. Subsequently, they failed to monitor their inherently risky systems properly.
Is her finding any consolation and shouldn’t we steer the discourse around
AI away from the hype and back toward the real problems of AI?
|
|
| ©AP |
This has mutated into a war of attrition, although unlike in World War I,
the dwindling numbers of soldiers on both sides are not keeping each other
in check with heavy artillery fire.
Like in 1917, there is calm on the front lines: “All Quiet on the Eastern Front”. Now, AI-controlled drones are bringing death and destruction deep into the enemy’s rear, and the ones who suffer are the civilian population, women, and children.
And with this, war criminal Putin is conducting sham negotiations for a temporary ceasefire, leading the world by the nose, and holding it hostage. Even India’s Prime Minister Modi, who greatly benefits from Russian oil at a low price, is pushing for serious negotiations.
Sorry for the digression. I just had to get that off my chest.
**
**











