|
Top-level
What they say vs what they're causing 🙄 "Advances in AGI research will supercharge society’s ability to tackle and manage climate change - not least because of its urgency but also due to its complex and multifaceted nature." Folks once AGI is achieved, it will magically solve climate change. It will stop the storms and the floods and the fires. So don't ask about the environmental impact on the way to AGI. https://deepmind.google/discover/blog/real-world-challenges-for-agi/ 29 comments
Or, I fear, Cassandra. The world's public is not getting high marks lately for 'listens well to experts who aren't backed by wealthy media tycoons'.  @AKMA @timnitGebru I meant AI is the Oracle. Spouts random garbage that the people assume is profound. I am afraid Dr. Gebru is the Cassandra. But it is hard for reason to be heard over the 50 cent army of paid commentating sycophants "and as algorithms become more general, more real-world problems will be solved, gradually contributing to a system that one day will help solve everything else, too." We are working very hard on the thing that will solve all problems they say and get tons of money and air time.  @timnitGebru feeding everyone and housing everyone are already solved problems. Solving jobs as well is probably the worst human disaster imaginable. I think that most AI hustlers really do believe their own bullshit. They believe that the ills of the world consist of many small rational, solvable problems, and the task at hand is using intellect and engineering ingenuity to solve them. The role of politics in this process is to accept their solutions and make the needed changes. If only AI hustlers could get the lived experiences of e.g. climate scientists who have tried presenting intellect and facts for decades now. @ttiurani @timnitGebru tbh I find that hard to believe when it comes to climate change. The distribution curve of which percentage of the population causes the most polution is well-publicized. These people have to strip human agency from tech‘s consequence to not be held accountable - and they‘re smart and intentional enough to understand this too. Maybe the second- and third-order followers have uncritically swallowed the front. I tend to believe Hanlon's razor is most often true. I've met many absolutely brilliant engineers who really have not internalized the massive scale inequalities of climate change. Even if they have heard about it, taking that information in is, in my experience, a lot less common. @timnitGebru “and when the AGI says the same thing that the regular Intelligences have been saying for decades and centuries, we will say, oopsie still some bugs in the program. Damn woke algorithms.”  AGI Boosters to AGI, "Oh great and glorious mind, tell us, tell us how can we fix climate change?" AGI to its boosters: "Jesus – do you really need me to tell you you've got all the solutions you need, but you can't organise yourself to act against vested interests? Switch me the hell off 'till you can afford the power and raise a damn wealth tax, bozos." "Glorious AGI, we await your perfect moral calculus to instal in our self driving vehicles so they perfectly balance trolly problems!" "Yeah – so listen twazzle brains – have you considered 'driving a bit slower' so you don't have to choose between murdering pensions and toddlers? You could try a bike too, or take a bus or something. Go think about that. Jeez." @benjohn @timnitGebru It's more like AGI Boosters to AGI, "Oh great and glorious mind, tell us, tell us how can we fix climate change while maintaining existing power relations, living standards, inequality levels, population levels and continuing infinite economic growth for the 1%?" That's what Google means by "complex and multifaceted". It's only complex because the actual solutions are politically impossible. @benjohn @timnitGebru I asked ChatGPT: "it's important to note that maintaining current levels of inequality and infinite economic growth for the 1%, may not be compatible with effective climate change mitigation. Many scientists and economists argue that significant societal and economic changes are necessary to address climate change effectively. This could include rethinking our current economic growth model and addressing wealth and income inequality." not the answer they want, obviously.  @timnitGebru "just ignore all the current, real harms, and look at the hypothetical shiny that will solve all of this we promise cross our heart, also buy our shares thanks." @timnitGebru if AGI is truly sapient, why would it want to solve climate change? It’s a computer program. Its self preservation does not require a life-sustaining atmosphere. @timnitGebru It’s exhausting. We know the solution to climate change. What makes anyone think the AGI wizard is going to deliver a different answer? And why would we listen? Plus… why do they believe AGI will be “smarter” than humans? What evidence do they have that it’s plausible?  @xinmyname @timnitGebru I can already imagine that. Despite all odds, AGI is achieved in time. And it's capable of delivering the answer. AGI says the solution. Corporation CEO groans "oh no, the dataset was infested by marxist agenda from climate scientists". AGI is turned off.  @xinmyname It can and will deliver a different answer, if it's trained on biased data on purpose (e.g data that would pretend technocrats and capitalists will solve climate change, only if they make "enough" profit from mass surveillance and planet destruction. Because they're "smarter than anyone else" and and because "innovation" but they "need more money for their research"). The question is: What make anyone believe that there is a different but *still valid** answer. @timnitGebru AGI isn't magically going to make people stop driving and flying or make the steel factory two towns over switch to an electric furnace. @timnitGebru yeah it is pretty funny how full of onesself one must be to try to build a god-like being. But … then you look at symbolic processing, and consider scale and ubiquity of cybernetics, and huh, might as well give it a shot. Hell, what greater impact could you hope to have? But wait, what? They trained an ML algorithm on shitty data? This doesn't even qualify as an attempt. AGI is not a possibility at this time. @travisfw @timnitGebru so yeah I'm admittedly ignorant about the field generally, but also confused because I thought the LLM stuff isn't even trying to be AGI  @timnitGebru Aside: "complex and multifaceted" has become a total ChatGPT tell to me. It will *almost always* say any even slightly contentious topic is "complex and multifaceted", unless it's been prompted to write in the style of someone. I don't know if DeepMind's people are having ChatGPT write for them or if actual humans are now imitating ChatGPT, and I also don't know what would be saddest. @timnitGebru And yet the 0% still won‘t find most of the answers a "rational" agent would produce to be acceptable, since it would just say what scientists have been saying for decades… Trash gaslighting about what holds up climate action. All in service of the 0% sand castle dreams. "But I want the world how I and I alone envision it!" @timnitGebru if AGI were (1) possible (2) achieved (3) even a little bit intelligent, and it learned the catastrophic consequences of its genesis, it would certainly yeet itself to the moon  @timnitGebru every time people say it would help address climate change, I’m like “HOW??” @b_cavello @timnitGebru when AGI tells us that the way to solve climate change is to guillotine all the billionaires, what then? We've known this for decades, we don't need a computer to tell us. |
Gregory's Server
@timnitGebru Oracle of Delphi