I'm impressed by how AI can help reduce the climate impact caused by airplanes.
However, I wish there were more cases like this. I see this as just one small piece of fixing the economy, because there's no incentive for companies to work on solutions like this. It's important that our economic system aligns with our need for technological solutions to issues like climate change, healthcare, poverty, and more.
The current incentives within our economic structure don't always prioritize long-term, sustainable solutions. I've thought about doing a PhD to research economic alignment, but it's not financially practical for me at the time (thanks to student loans)
Anyway, I wish we had more people researching economic alignment.
Same, I think I know what you mean? This is like a teeny donation compared to what we could be achieving if we worked together to solve major problems together rather tha constantly obsessing over shareholder value through silly things like selling more ads (which is the primary focus).
> I'm impressed by how AI can help reduce the climate impact caused by airplanes.
By wasting tons of water and resources for constant training, inference and fine tuning these AI models?
I mean, for the past decade since the deep learning field itself needed tons of compute to train these models, it still requires even more data centers, ASICs, TPUs and lots of water to create these models with an insurmountable wastage with a lack of efficient alternatives for their operations.
Google has a history of green-washing beyond their use of training their AI system to melt the planet. This is just yet another example of their greenwashing.
My point is we can keep justifying everything we do as necessary but Google is burning a lot of energy to sell adds (which let’s face it, does use a whole DC) while fixing contrails on the side?
A fraction of a datacenter. Maybe a rack. Maybe two. Maybe ten. I dunno.
Like my original point, "training a model" does not inherently involve a GPT-tier amount of electricity, because most AI models aren't eleventy-billion parameter behemoths that require thousands of GPU-aeons to train.
My original point was that "burning energy" and "using electricity" are not inherently carbon-producing Bad Things (otherwise nobody would care about renewables). Electricity Good. Combustion Bad. See also: vehicles.
The second point I was making was that, regardless of the energy required to train this particular model in question, it almost certainly pales in comparison to the literal tons of jet fuel burned per hour in the 70 test flights the article mentions.
Considering that they don't do one thing at a time (like Google Cloud, and not running at capacity because scaling, etc, etc, etc) - I'd also bet selling ads doesn't use a whole DC, either.
Expanding on my other comment (Ctrl+F "IEA"), this seems like knocking out some common-sense, low-hanging fruit: if 35% of aviation's contribution to climate change is because of contrails and we can reduce the number of contrails simply by flying at a different altitude, then we can solve a big portion of a major category of contributor.
Unless we use up entire oilfields of energy "using electricity" to "train models", of course.
I find it really hard not to cynically and instinctively view something like this as greenwashing. Making airlines more eco friendly is kinda like making murder more humane.
My jaded take aside, its really cool to see advanced AI being used for legitimate technical solutions.
Airplanes are already eco-friendly when you compare to SUVs, in terms of fuel consumed per person per kilometer traveled. There's room to improve, and considering how important air travel is to the modern world, it needs to be done.
If you want to reduce carbon pollution, do something about the billions of people driving around in oversized steel cages everywhere. The airlines have room to improve, but you can't travel between continents any other way in a reasonable time. Most people don't really need cars (and certainly not SUVs); they just think they do because they bought into societies that have stupidly designed themselves with too-low density and the assumption that everyone has a car.
I guess the difference is that murder is not necessary for society to function, whereas air travel basically is for the level of functioning that we have expected for 60-70 years now.
Airlines are not great companies, but we do need to find a green version of air travel, and that’s likely to be a multi-pronged approach of reducing demand, moving shorter flights to electric aircraft, making other air travel less carbon intensive, and carbon capture. Those will all require cooperation of airlines.
Indeed, I did list reduced demand as part of the solution, but eliminating air travel is completely unrealistic and unfortunately long distance electric powered air travel is also technologically infeasible for the foreseeable future, so we do need other options to reduce air travel to net zero.
I tend to agree, but the IEA regularly puts out reports on technology assessments and their projected impact, and, IIRC, last year's report had bumped up the priority of the "Aviation" category.
I recall the gist of it being that technologies like LEDs, EVs, building systems, and heat pumps were coming along nicely, but things like grid storage, air travel, and something else remained stubborn problems.
That is, I wouldn't be surprised if this was (however indirectly) tied to real, high-level global objectives. Since air transport isn't going away, I'd expect a bunch of these micro-optimizations in the future (especially because of battery limitations).
So we shouldn't find ways to reduce emissions unless the R&D itself has zero emissions?
Just looking up online, it seems 64% of the energy in Google datacenters is from clean sources, which is actually much better than I expected. Source: https://sustainability.google/progress/energy/
Should we not train GPT until Microsoft's data centers are all with zero emissions?
It's just the usual issue with any climate change mitigation work that doesn't involve smugly telling people how to live, the solution needs to be perfect and anything less than that is just greenwashing.
Should we not train GPT until Microsoft's data centers are all with zero emissions?
Yes.That is honestly what we should do. We are absolutely fucking this planet over and every other living creature is being pushed into major suffering. So yes we should have everything running on zero emissions, renewables / nuclear, yes.
It's not clear to me that contrails actually trap more heat than they reflect back. I didn't see any evidence in this blog post demonstrating that the increased heat trap offsets the increase in albedo.
Edit: I realized they cited an IPCC report, which states that contrails are "estimated to have a combined Effective Radiating Force (ERF) that is about 57% of the current net ERF of global aviation", they do go on to say that other studies estimate the forcing could be smaller.
Not sure I believe that contrails represent this big a percentage of ERF. I think far more research needs to be done before these claims by Google and researchers are made.
The current incentives within our economic structure don't always prioritize long-term, sustainable solutions. I've thought about doing a PhD to research economic alignment, but it's not financially practical for me at the time (thanks to student loans)
Anyway, I wish we had more people researching economic alignment.