Advice on putting AIs in a solarpunk setting (by a ML engineer)
Advice on putting AIs in a solarpunk setting (by a ML engineer)
This conversation and the reactions it caused made me think of a few tips to explicitly veer away from AI-aided dystopias in your fictional universe.
Avoid a monolithic centralized statist super-AI
I guess ChatGPT is the model people use, the idea that there is a supercomputer managing all aspects of a community. And people are understandably wary of a single point of control that could too easily lead to totalitarianism
Instead, have a multitude of transparent local agents managing different systems. Each with a different algorithm and "personality".
Talk about open source
The most used AI models today are open source. We have a media that is biased towards thinking that things that do not generate commercial transactions are not important yet I am willing to bet that more tokens are generated by all the free models in the world than by OpenAI and its commercial competitors.
AIs are not to be produced by opaque companies from their ivory towers. They are the result of researchers and engineers who have a passion for designing smart system and --a fact that is too often obscured by the sad state of our society where you often have to join a company to make a living-- they do it with a genuine concern for humanity's well being and a desire that this work is used for the greater good.
It is among AI engineers that you will find the most paranoids about AI safety and safeguards. In a solarpunk future, this is a public debate and a political subject that is an important part of the policy discussion: We make models together, with incentives that are collectively agreed upon.
AIs are personal
You don't need a supercomputer to run an AI. LLMs today run on relatively modest gaming devices, even on raspberry pi! (though slowly at the moment). Energy-efficient chips are currently being designed to make the barrier of entry even lower.
It is a very safe bet to say that in the future, every person will have their own intelligent agent managing their local devices. Or even one agent per device and an orchestrator on their smartphone. And it is important that they are in complete control of these.
AIs should enhance humans control over their own devices, not make them surrender it.
AIs as enablers of democracy
You not only use your pocket AI to control your dishwasher, it is also your personal lawyer and representative. No human has the bandwidth to go through all the current policy debates happening in a typical country or even local community. But a well designed agent that spends time discussing with you will know your preferences and make sure to represent them.
It can engage in discussions with other agents to find compromises, to propose or oppose initiative.
As everyone's opinion is now included in every decision about road planning, public transportation, construction schedules and urban development, the general landscape will organically grow friendlier for everybody.
I agree with this take that the path to implementing AI from a solarpunk perspective is through open source software which promotes collaboration and privacy.
Nice vision! Please keep advocating this viewpoint to non-insider audiences. Itโs a very realistic extrapolation from today to the near-future. It tickles me because it is also hopeful.
We are building these things so shouldnโt we think really hard, all of us, about how they are built? Everyone is a stakeholder in the future.
Everyone is a stakeholder in the future.
This is really cool, I like it. We go from 'sufferers of the future' to 'stakeholders in the future'.
Thank you for your kind words! 2023 has been a weird year for me as lifelong AI enthusiast. I saw people go from "it is impossible" to "we should not do it" very quickly and was sad that the utopian point of view, which has been the motor of most researchers, is almost never represented in the media. I almost feel that it is more important right now to take part of the public discussion than it is to code AI systems.
we should not do it
Because at this point, some people are suffering damage by the use of AI. I might not be hired as a translator in the future, my income is gone. Others have had their artworks, texts, creative output stolen (plus their work isn't needed anymore). We have delegated creative work, which should be humans' pride and joy, to a machine. Why on earth would anyone in their right mind do such thing?
So we as humans have to have a discussion about the responsible use of AI (I think nobody who screams 'down with AI' has any illusions that it will ever disappear again). As with any new tech product, the discussion should have been had before it was unleashed onto the public, then again you can't talk about it if you don't use it. It's also time for the researchers to pick up on that sentiment and explore the ethical uses of the great power they have created.
If those now losing work as text and image workers were retrained as IT security people or AI prompt inventors before such tech was introduced, brilliant! If the numbers about ecological sustainability and advantages for society add up, I'm on board. But I want this to be a slow process and a public discussions (maybe with time some media learn to tone down the techbro/luddite extremes), before we drown in AI-generated shite nobody ever asked for.
Open source and the self sufficiency of self hosting and running software locally are at the core of solarpunk.
The more llms are optimized in the future, the easier it will be to host and run on lower end hardware with less resources using upcycled hardware.
ollama is great for self hosting llms. It comes with several very useful ones that will work out of its box from its library. With a little bit of effort (creating a modelfile and adding a from line to point to the model's path) one can also run any openly available model if its in gguf format or one can take models found on huggingface and convert them to gguf format if they arn't already.
Yep! Using vLLM and text-generation-webui myself.
Absolutely brillant to see my post spark a discussion like this, I'll be taking a ton of this on-board and rethinking it all. Thank you! :D
Cheers! :-D
Well, this is the first time somebody has managed to convince me that AI could actually improve things, so congrats on that. It has to do with your mentioning of AIs being made as a result of public debate; I'm seeing major parallels to scientific research. Research must be done and documented with rigor, and fully exposed to peer review and criticism. If - and I think only if - AI models end up being handled the same way, I think we have a chance. Not sure if it's already too late for that, but it's a chance.
Thanks for the kind words!
I think the process will have to be different: scientific research has a ground truth. It tests its theories against reality. With AI design we veer in the morality side of things, where ground truths do not exist and the process will have to be a bit political, with people with coherent but incompatible opinions debating.
But I do think that it will be easier than we suspect. We actually agree on 90% of what we want to be done. We want a labor free life, we want a free house. All we will have left to argue will be the color we paint it.
nice ! ive been a bit wary of llms cuz of electricity usage n environmental impact. iz there any things u can point me to for running a greener ai myself?
There is IMHO a very counter-productive dynamics arising around the debate about the environmental impact of IT. We mostly hear luddites and techbros argue in bad faith over invented numbers. I would urge everyone involved in this debate to first make sure that the numbers used are correct.
Using a LLM with today tech (which are not yet really optimized for it) is akin to running a 3D videogame with good graphics setting: it uses the GPU quite a bit but only when you do run queries in the model, which may be infrequent. When it runs at full my GPU takes 170W. Add probably 200W for the rest of the computer. I do know that it is really not my primary emission cause, especially living in France where CO2/kWh is pretty low.
A greener AI would be one that frees my time to work on home insulation or in convincing people to switch to heat pumps. At one point I'll probably install solar panels and home batteries. 400W is a relatively easy target to reach. The water heater and cooking devices use more than that.
The debate is more about the cost of training models, which use datacenters at full capacity for days or even months for the biggest ones. Thing is, many people confuse the training with the use. Training has to be done once. Well, once per model, which is why open source models are so crucial: if you have a thousand companies training their own proprietary model, it wastes a lot of energy but if instead they use a shared trained model and maybe just fine tune it a bit for a few hours, it really decreases the amount of energy used.
Also, many datacenters have been greenwashing a lot, claiming to have decreased tenfolds their environmental impact or even offset it totally. This is greenwashing not because it is false, but because the intent is much straightforward: electricity is a big part of their costs, cutting it down is just good business sense.
It has become customary for big models to publish the energy used and an estimate of the CO2 emitted in the process. Llama2, possibly the biggest open model trained so far, emitted about 1000 t of CO2 equivalent. It sounds like a lot but this is equivalent to one international 10h commercial flight and it fed the open source community for more than a year. Any AI conference would emit more. And unlike flights, it does not have to emit CO2: it uses electricity that can be sustainably produced.
I tend to veer a bit on the techbro side: when you look at the actual numbers, and the actual possibilities, the emissions are not problematic, they are useful uses of electricity that are included in the debate over a sustainable electricity grid.
Okay, even as the resident luddite I love good and correct numbers. 370W for a home computer? Not sure I consider that's a low value. For my everyday laptop I've gone back from shitty GPU to no GPU because I think we should spend time regenerating the soil, installing heat pumps and insulating our homes instead of tiddling around in simulated worlds or converse with simulated intelligences. (Hope you don't take offense, it's not meant to be abrasive, I enjoy discussing this and am happy to have my perspective changed.)
I'm not sure I like the current use of AI (putting me and millions of other text and image workers out of work and enshittifying the internet). Unless we figure out really quickly how to make sure this wondrous thinking power is used for good and can safely limit its misuse I just see additional energy-hungry 'miracle' tech hype products piled onto the already existing systems, and creating more trouble future IT personnel will have to solve. Computing itself almost seems like a self-replicating and self-expanding entity. I can see some interesting applications and maybe even have hope in my tiny heart that this time the tech bros are right, but unleashing shitty commercial data-stealing AI products onto the public/internet amounts to crime imo, and the damage I can perceive from where I stand outweighs the usefulness, so far.
On the other hand, I tend to be a silly old luddite sometimes, and will have my kid give me a proper introduction into AI prompts one of these days. He's been annoyingly smartassish and incredibly useful lately by using AI as research tool. Guess I'll find out what it could do for me and calculate the corresponding computing power (do you have a site where I can look up numbers like this?)
thanks for the reply !! i have an old "gaming"" laptop (that i just use for most things tbh) but its 150W amd got a GTX 1060m in it, so i would be curious to see if i could run some sort of local ai on it. i believe the CO_2/kWh is pretty good here we rely mostly on renewables, but still have some coal power stations around -_- ah ! if training the ai is the hard part and running it is easier on energy then it may be better than i thought, though i am wary of how it gets used by companies... i am very aware though going into a engineering field in the next couple years that llms are gonna be a big tool to use, though i dont agree with stuff like chatgpt for privacy reasons, but i was unsure about how bad the environmental impact would be if i tried it, thanks for the info !! :3
Self hosted Kobold Ai with any model you would like on a pc witch is powered by solar panels?
An environmentalist activist at our local fablab once told me "forget about your fancy electronics, insulate your motherfucking water heater!" :-)
When you look at the hard numbers if your AI machine is your main energy use and source of CO2 emission, you are doing extremely well from an environmental point of view. Even running a LLM locally requires probably less than 300W (at peak) which is very easy to get through solar power.
thanks ! ill check it out ! >w<
@keepthepace noted
Distributed compute like this could also be interesting: https://www.technologyreview.com/2023/08/18/1077548/computer-waste-heat/