Good article wrt question of forecasts and accuracy - charts 30 past years of forecasts and matching observations.
It might be easy to dismiss the Pennsylvania forecast. In some corners of the Internet, Mann is viewed as a master of climate doomerism for his outspoken views on the perils of a warming world. But if anything, since its first issuance in 2007, Mann's forecast has proven to be conservative.
The forecast accuracy could use a dedicated article.
It tracked very well from ~1997-2002, 2008-2012, and 2014-2015. It was off, from 1994-1996, 2003-2004, 2006-2007 2015-2020. In 3 of the 4 cases there the fore cast under-predicted major storms, outside the std of the forecast. I wonder why.
I don't think a ring of galaxies is going to look very different from anyplace within the solar system. Anyway I think moralestapia's point is that the circle might not be centered on us, so the redshift of the galaxies would not be the same. We could still determine that a circle exists by plotting the galaxies in 3D.
No, I mean, a 2D circle could appear as a line from a certain perspective in 3D space.
Spin up your mental model of a circle in 3D space, look at it from a vector perpendicular from its diameter, rotate it 90 degrees in any other axis but the one you're looking at it; on that 2D projection, it will be a line.
>No, I mean, a 2D circle could appear as a line from a certain perspective in 3D space.
Right, and as a matter of fact that's exactly what we DO see with the Milky Way galaxy. It can be conceived of as a circular disc, more or less, but in our sky we see it from the side, as a streak or a band rather than a disc.
I was browsing test kits where you send a sample back to a lab. They're $79 - $300 and advertise 1 or 2 parts per trillion. Looks like getting to 0.02 ppt requires some very specialized equipment, and would probably be optimized for continuous monitoring of a water supply.
Fine tuning can be useful if you need to generate lots of output in a particular format. You can fine-tune on formatted messages, and then the model will generate that automatically. That could save a bunch of tokens explaining the output format in every prompt.
Does this Python package control the LLMs using something other than text? Or is the end result still that that Python package wraps your prompt with additional text containing additional instructions that become part of the prompt itself?
Looks like it actually changes how you do token generation to conform to a given context-free grammar. It's a way to structure how you sample from the model rather than a tweak to the prompt, so it's more efficient and guarantees that the output matches the formal grammar.
The output of the LLM is not just one token, but a statistical distribution across all possible output tokens. The tool you use to generate output will sample from this distribution with various techniques, and you can put constraints on it like not being too repetitive. Some of them support getting very specific about the allowed output format, e.g. https://github.com/ggerganov/llama.cpp/blob/master/grammars/... So even if the LLM says that an invalid token is the most likely next token, the tool will never select it for output. It will only sample from valid tokens.
