35
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 13 Oct 2023
35 points (100.0% liked)
Solarpunk
5393 readers
57 users here now
The space to discuss Solarpunk itself and Solarpunk related stuff that doesn't fit elsewhere.
Join our chat: Movim or XMPP client.
founded 2 years ago
MODERATORS
I don't think we need AI. Without the need to constantly work the tutor can just be one of the child's parents. This would work better because children naturally respect and want to emulate their parents. The tutor doesn't even need to know everything and just teach how to analyze situations and find knowledge.
But I agree that kids should be included in workspaces to teach them about necessary (or interesting) jobs.
Overall I think the best way is to allow kids to find their own best ways to learn.
I do think it is valuable to separate knowledge-seeking (where it is good to have access to a knowledge not limited by the parents) and emotional support. Being able to learn without fearing being judged and evaluated, that's valuable. Also as a parent, I do know that my reserve of patience is not infinite and I am happy that my kid finds sources of knowledge that are independent of me and my biases. Then we discuss things.
As someone who finished high school at the beginning of the internet, I can guarantee that access to free information unencumbered by the limitations of adults around me (including my loving knowledgeable parents) was essential. To me a virtual AI tutor is just a mean to make this accessible earlier. My kid still has trouble reading long and complicated text, and I rather have a smart tutor proposing him audio content than random youtubers.
knowledge will obviously come from other sources too. When kids socialize with others they will learn things naturally, and discussion should absolutely be encouraged. However AI produces a lot of problems. AIs have bias based on the information they learn, they require resources to build and maintain and cannot discuss information accurately. I just don't see what AI adds over just interacting with other people.
Solarpunk societies, like all post-capitalist societies, are build on strong human relations, replacing one of the avenues of creating them with an hallucinating rock (exaggeration I know) just seems weird.
Also, teaching kids is not a dull or hard job that has to be automatized away. A lot of people actually love doing it so much that they even put up with the horrible limitations and shortcomings of the current education system. Just make sure they are adequately equipped and rewarded and I'm sure we need no AI. Just a really good library system.
Oh teaching kids once in a while is fine and fun, but providing all the info they want at the pace they need when they need it is another thing. I love teaching things to my kid but there are times when it is dull, there are times when it is hard. It is doable, but being able to open the floodgates of knowledge when there is a demand is a huge boost to what they can learn effortlessly.
Currently we find enough such dedicated people to do that for ~30 kids at a time. Even for a good teacher that's ridiculously inefficient.
As a knowledge-thirsty kid who grew up before the internet, and who spent a lot of time in libraries and bookshops, having to go back to the bottleneck of books and outdated shelves is something I wish to no one.
My kid has the same problem as I did: he has basic questions (is the sun heavier than earth), that quickly lead to advance subjects (oh so black holes are the heaviest objects out there?) then into research subjects (wasn't the big bang densest than black holes? How did it get apart?). Whatever subject you are into, you quickly outgrow your resources if you go deep enough. AIs (today, I dare not imagine in 5-10 years) are able to take an advanced research paper and make an ELI5.
Me? When I was in junior high, like many geeky kids I was into space conquest. I had weird looks from the school librarian when I was asking about a book that could tell me about the harnessable energy sources on other planets. "Just try the encyclopedia". At high school a friend had internet access, I was quickly reading NASA studies from programs that were not even mentioned in a single book at my school.
Solarpunk utopias are not energy-poor utopias. Quite the contrary. They are what happens once we have decorrelated CO2 emissions and energy use.
When I was a kid, my parents brought me weekly at the town's library, and about monthly at the city's bookstore. There is some margin before a computer usage comes close in terms of CO2 emissions.
Most AI companies "offset" their carbon footprint. I guess a part of that accounting is greenwashing but some are doing that directly with solar panels. I would argue that in such a case, their energy usage is irrelevant. And I trust that they probably do what they claim because it does save them a lot of money to do so.
Also "a lot of energy" is really debatable. Even if they used power directly from the US grid, the Llama 2 models (which fuel a democratization of LLMs like none before) have emitted about the same as one international flight for their training, that needs to be done only once and that is now free for everyone to use. There are not a lot of fields that have such an impact for such a low footprint. One international conference bringing people from many countries would have 10x that footprint already.
I guess you can decorrelate CO2 emissions, but in turn will have to put up with similar disadvantages caused by mining pollution, solar panel production, etc. Sometimes it seems that the proponents of an energy rich future still dream of having a free lunch and eating their cake too because 'renewables', but these technologies need resources and infrastructure as well, which an energy hungry population might not be able to provide in a sustainable way (can't burn the forest faster than it grows back, can't cover the entire surface of the earth with solar panels).
Sometimes it feels that people living in the fossil fuels world have a hard time understanding that the fossil fuel maths is not universal. A fossil fuel world requires extraction per kWh (energy) used, a post-fossil world requires extraction per kW (power) installed. Once your solar panel installed, whether they produce or not does not change the environmental impact. Actually one could argue that not producing energy is what causes waste and environmental impact. Also, panels are highly intermittent. At noon you will likely have a spike of free energy. Yep, that's a free lunch with a cake and cherry at the top.
A portion of the Sahara or of any ocean would be enough for several times the current world consumption. And if we start deploying in space (we have the tech for btw, yes including microwave transmission, tested over long enough distance) there is basically no limit before Kardashev II.
What are your solar panels built out off? They never need maintenance? We can just plaster them over the bit of remaining wildlife we have on the planet? And I find future solutions in space for very present problems too optimistic, sorry. I do not agree with your username, I think we shouldn't keep the pace, but seriously slow down. Because a lot of the tech we have doesn't really add life quality but rather reduces it, at the stage we're at. 'Shooting solar panels into space' sounds very much like another tech-heavy idea when we first need to relearn to live and coexist with the life we haven't destroyed yet. A lot of energy can be saved that we are wasting. Are we still cooling data centers with water that is used for nothing else? Why doesn't that heat get used where it's needed? Solar panels without shade-loving crops planted under them? Wasted space. Things not recycled? Unrecycleable things still produced?
Not saying that solar panels in space at some point might not be a useful idea if our tech evolves a lot - but before that we absolutely have to learn to not shit everywhere we walk, so to speak, and contain our production and energy cycle processes. But I guess this just might be a discrepancy in how far we want to look into the future here. I'm just afraid if we go too fast we miss the first step in our current affair of mess we made: cleanup, containment, co-existence.
I am not saying that space-based solar power is our go-to solution or that we need to bet on that to solve all other problems. I am saying that room to put solar panels is not the limiting factor of the tech. Cover deserts and oceans, use them as shades as you propose, and we have enough room to produce several times our consumption. I am only mentioning space only to express how limitless this factor is.
For the essential part mostly silicium, which is abundant on earth. Usually some structural metal, like aluminium or steel as well but many other materials work. You need small amounts of rare earths as impurities, many different tech exist. None has a problem of sourcing the minerals. We have enough proven reserves of these to switch to renewables. At this point the conversation usually switch to the environmental impact of resources extractions, which is not a tech problem, but a political one: we can make clean extraction. We make shitty extraction because it is legal to pay mineral from countries with no environmental protection and no labor rights but that's like saying farming can't be done sustainably because in some countries it is done by burning down forests to install ever-growing farms.
Not a lot. Cleaning in some places, though a well designed system will have the rain for that. Weeding once in a while in luxuriant places. In some places panels are put flat, not at an angle, so that a dumb automated cleaning system is easy to put into place. It is less efficient per area of solar panel, but apparently as efficient per area of ground occupied. In space no, no maintenance (though here again, not advocating that directly)
The biggest problem about Saharan solar farm is how are you going to get the power to where it needs to be?
It has been studied, partially funded, called DESERTEC, put on hold since the Arab Spring.
Basically: big DC power lines. The tech is known and exists.
I don't know if the various hydrogen production or methanation processes are advanced enough to consider these as an alternative though.
Things my kid learnt through "socializing and learning things naturally":
Nonwhithstanding the fact that we are talking about future tech and that the pace at which AI is advancing right now is crazy, even today, on these metrics, I think they do better than we do as a society without them.
The examples you provide are negatively biased. You don't know all of the normal and useful things they learn because they don't stand out. Also two of those examples (Church and school busses) come from current cultural biases, something a solarpunk society would hopefully mitigate.
I think AI is not suited for discussion. It might be good at conversation but discussion isn't just conversation. Discussion requires understanding of others to a degree I don't think AI can achieve.
I concede my point about resources, but will add that the model will get outdated and will need retraining every once in a while.
Textbooks are bad. I agree. I just think they should be replaced with a human that knows what they are talking about and the topics that are learnt are things that the kid actually wants to know instead of what people think they should know.
Also I can't help but notice you ignored one of my core arguments: that solarpunk societies are about strong human connections and replacing one of the main sources of these connections is a bad idea.
I also think that the process of finding information is as important as the actual information. If all of your questions are answered just by typing it into the computer then you never learn the importance of checking information accuracy, accounting for bias and other very useful skills.
AI allows you to shortcut to the information you seek which means you never learn how to actually think for yourself.
I'll address that one first then
I have had people tell me that we should not automate cashiers because it removes human connections. I believe that humans want human connections and that if you remove the obligation to get something out of these connections, they will become richer and more meaningful. I sometime feel I am tricking my kid to trigger their interest in "useful" things. I'd rather play with him than force his to rewrite the same word with good pretty round letters a hundred times.
When you remove painful obligations from a human connection, you may sever some that you were forcing yourself into, but you also give room to much more of them. If teaching was accounted for, I would spend more time showing my kid what I really love in life, places that make me feel at peace, events that make me feel alive, techniques that I find interesting despite their practical uselessness.
Human are social animals. We make social connections and may even die for it. Don't worry, removing an obligation will not remove the need for more meaningful ones.
Now for the rest:
Why do you assume that a society that manages to mitigate biases that are extremely central to our current problems in social relations would have a hard time mitigating bias in AIs training dataset?
Look at what AI mentors do today, right now. Ask them follow up questions, ask them what you don't understand. Several conversations with GPT-4 (which is the best right now but trailed by more open models) convinced me otherwise. Even if you may argue that such models currently are not as good as a human tutor, I find it hard to argue that it does not beat the "conversation" of a class of 30 with a lone teacher.
True in all day and age and on every media. Had to teach it to several kids in my family as school seems to do a pretty poor job at it. Interestingly enough, something that LLMs can teach well as well despite being (currently) pretty poor at it.
Okay, I agree. LLMs might be useful for education. But they should not replace practical experience.
I just am skeptical to everyone that says we should replace something with AI, because most of the time these decisions seem to be motivated by profit and aren't actually better.
I think 99% of the companies that hope to replace something with AI for profit are going to fail, as these models become lighter and lighter and have open source equivalents and a VERY motivated community behind them to prevent corporate lock-in.