97% is great (though that is just for vegas) but it is still a long way from enough. Its a truism of availability that each 9 of uptime is more difficult to get to than the last, i.e. 99.9% is significantly more difficult/expensive than 99%
Then get it from the sources that already exist.
The problem here is that you cant simultaneously say "Solar is so much better than everything else we should just build it" and "we'll just use other sources to cover the gaps". Either you calculate the costs needed to get solar up to very high availability or you advocate for mixed generation.
None of which is to say that solar shouldnt be deployed at scale, it should. We should be aware of its limitations howver and not fall prey to hype.
One chat request to an LLM produces about as much CO2 as burning one droplet of gasoline (if it was from coal fired power, less if it comes from cleaner sources). It makes far less CO2 to talk to a chatbot for hours upon hours than a ten minute drive to see a therapist once a week.
The problem isnt anthropic get to use that defense, its that others dont. The fact the the world is in a place where people can be fined 5+ years of a western European average salary for making a copy of one (1) book that does not materially effect the copyright holder in any way is insane and it is good to point that out no matter who does it.
FWIW, a short query to a typical sized LLM takes about 1Wh of energy, there lots of variance on how big the model you are using and how long the input and outputs are but thats the correct order of magnitude. 1Wh is the amount of energy consumed by a 1kW electric kettle in 3.6 seconds or a 2kW hairdryer in 1.8 seconds.
if you assume that energy was produced in a coal power plant (the worst for co2 emissions) then it makes around 0.3g of co2 emissions, which is the equivalent of burning about one droplet of gasoline.
I agree with your first part, but I dont think I've ever used a windows, osx or linux computer that hasnt had issues connecting to printers, the problem there isnt with the computer.
Statistics:Length of war: forty-eight years, one month.Total casualties, including machines (reckoned on logarithmic sentience scale), medjel and non-combatants:
851.4 billion (\B1 .3%). Losses: ships (all classes above interplanetary) -91,215,660 (\B1 200); Orbitals - 14,334; planets and major moons - 53; Rings - 1;Spheres - 3; stars (undergoing significant induced mass-loss orsequence-position alteration) - 6.
Historical perspectiveA small, short war that rarely extended throughout more than .02% of the galaxy by volume and .01% by stellar population. rumours persist of far more impressive conflicts, stretching through vastly greater amounts of time and space... Nevertheless, the chronicles of the galaxy's elder civilisations rate the Idiran-Culture war as the most significant conflict of the past fifty thousand years, and one of those singularly interesting Events they see so rarely these days.
It's more that it is mind bogglingly huge by our standards, yet still a blip on the scale of the galaxy.
I'm pretty sure that in 100 years time people will look back at the current age of social media with the same kind of horror as we get looking back at doctors recommending cigarettes for weight loss.
In this case, they are being forced into it because the NYT got a judge to agree that AI might be used to access their content and that blanket surveillance is an appropriate response. The IP mafia are the guilty parties in this instance not openAI.
All electrons have spin 1/2, that's a property of it being an electron. They have a spin vector (the arrow shown) and whether it is in the same direction or opposite direction to the magnetic field it's in determines where it is plus or minus.
Now you might think "but what if it is not entirely aligned with the field, then it wouldn't be 1/2", which is true, on aggregate for large numbers of electrons, but if you ever look at a single electron its spin will either be "up" or "down" never any other orientation.
This is the kind of thing people are referring to when they say "no one understands QM", we know it is the case, we can measure it and predict it, but it makes no fucking sense.
Because he's speaking to a British newspaper about British policies. I'm assuming the second part as I don't subscribe to the times so cant read the article, but there are currently plans in place in the UK to introduce an opt-out framework for people to remove permission for training on their work, with pushback from big names that want to charge rent on their old works, so I assume that is the subject.
Even if he wasn't talking about the UK at all (which I think it is clear he is from context) my larger point still stands, the choice isn't between stopping AI and allowing AI, its between allowing AI companies to operate in your jurisdiction or AI being trained elsewhere that is out of your control. There is no option for "stop this entirely", unless you can persuade the USA and China at the very least to sign up to it. Which they wont.
The bit you're missing is that the choice isnt between killing AI and killing the music industry, its between killing AI in the UK or pissing off IP holders somewhat. Do you think China give a fuck who's IP they use in training models, or that they will stop if the UK passes a law making artists default out of using their work as training data?
97% is great (though that is just for vegas) but it is still a long way from enough. Its a truism of availability that each 9 of uptime is more difficult to get to than the last, i.e. 99.9% is significantly more difficult/expensive than 99%
The problem here is that you cant simultaneously say "Solar is so much better than everything else we should just build it" and "we'll just use other sources to cover the gaps". Either you calculate the costs needed to get solar up to very high availability or you advocate for mixed generation.
None of which is to say that solar shouldnt be deployed at scale, it should. We should be aware of its limitations howver and not fall prey to hype.