Not necessarily. There were some tests last year-ish from hf that showed that simply alternating (randomly) between claude and gpt (whatever their versions were at the time) on a task produced better results than either of them individually. So during a task, the first call was sent to one, then the other and so on.
There's also the concept of "smart routing" requests based on some heuristics / embeddings. You'd get "simple" tasks handled by smaller (cheaper) models and use a bigger model to curate / sort / merge the results.
There's a lot of things to try here. I wouldn't personally pay for this service, but I don't think it's "a joke"...
Yet there is a group that reliably loses in the United States: average citizens. From "Testing Theories of American Politics" (Gilens & Page 2014)[1]:
"These results suggest that reality is best captured by
mixed theories in which both individual economic elites
and organized interest groups (including corporations,
largely owned and controlled by wealthy elites) play
a substantial part in affecting public policy, but the
general public has little or no independent influence."
I guess this result doesn't seem surprising to me? A majority of the general public can't even identify how long Senate terms are (https://www.pewresearch.org/short-reads/2023/11/07/what-amer...); they simply don't have the required knowledge to meaningfully influence politics without mediation through interest groups. (Anecdotally, it’s a truism in political circles that you will drive yourself insane trying to understand the median voter’s theories of politics.)
If they are "running the world" they are certainly not doing so for your benefit.
(There is also a great deal of distance between "running the world" and "influencing some events for the benefit of a select few, no matter what the costs to the rest of the world". Personally, I find the latter far more likely, but also undesireable.)
Quantity has a quality (or lack thereof) all its own.
How many shit books can or will you wade through to find a good one? Particularly when some percentage of the shit ones are "good enough" that you won't necessarily know it was shit until after you've read some or all of it? What do you do when the ratio of shit to good is 100:1 or 10,000:1? In the past, you could find a trusted publisher or reviewer without too much trouble. Soon, if not already, the publisher and the reviewers will have the same dross ratio as the books do.
I think that picking the correct (or most correct, which is trickier) use of the word in context (out of, as you say, many options) might be a good way to test for receptive vocabulary.
It is not. Vocabulary is far from the binary of "you know this word or you don't". At a minimum, it is usually split into passive and active vocabulary, with passive being the words you understand when encountered, and active being the words you can use effectively. Wikipedia's entry is a pretty good overview.
My understanding has been that an additional reason breeder reactors have been discouraged is due to nuclear proliferation fears; the neccessary reprocessing of the fuel also produces weapons-grade plutonium.
proliferation argument doesn't hold considering you can do it much cheaper with distributed enrichment of uranium. Breeders can open a pathway for plutonium warheads but it's more about optimization. If a country wants such warheads, it'll start with easy enrichment path
In the sense that the arsonist who has set the building on fire has been stopped from also lighting a wooden bucket on fire right now, sure, positive news.
"Now this is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning."
The building is not burned down. Not making everything worst all the time has to be the first step.
Second derivatives, and all that.
No reason to pop the Champaign on time square for everyone ; but all the scientists working on this can take a good, deep breath of relief. And the republican senators who broke rank need to be nagged in breaking ranks on other topics. (Don't get me wrong, they do it purely for electoral purpose - but that's a feature of representative democracy, not a bug.)
reply