Sorry if I missed it, but how is a single token output from an LLM comparable to a search result from an engine? The author here compares 1k tokens (as an estimate for an average LLM single query response) to 1k web search queries. How is this not a factor of 1000 error?
> To compare a midrange pair on quality, the Bing Search vs. a Gemini 2.5 Flash comparison shows the LLM being 1/25th the price.
That is, 40x the price _per query_ on average (which is the unit of user interaction). LLMs with web-search will only multiply this value, as several queries are made behind the scenes for each user-query.
EDIT: thanks, zahlman, he does quote LLM prices in 1M tokens, or 1k user-queries, so the above concern is mistaken!
> The author here compares 1k tokens (as an estimate for an average LLM single query response) to 1k web search queries. How is this not a factor of 1000 error?
The author compares 1k uses of the LLM - resulting in an estimated 1M output tokens, and the prices are quoted per 1M tokens - to 1k uses of the search engine (the prices for which are directly quoted per 1k uses).
Gemini 2.0 Flash is listed at 0.4 USD / 1M tokens. Bing search API is 15 USD / 1k queries. So the LLM is indeed 37 times cheaper for a 1000 token query.
> By uploading any User Content you hereby grant and will grant Y Combinator and its affiliated companies a nonexclusive, worldwide, royalty free, fully paid up, transferable, sublicensable, perpetual, irrevocable license to copy, display, upload, perform, distribute, store, modify and otherwise use your User Content for any Y Combinator-related purpose in any form, medium or technology now known or later developed.
You may have issued such a license...
Though without an explicit sublicense from Y Combinator, they may have issues with this application:
> Except as expressly authorized by Y Combinator, you agree not to modify, copy, frame, scrape, rent, lease, loan, sell, distribute or create derivative works based on the Site or the Site Content, in whole or in part, except that the foregoing does not apply to your own User Content (as defined below) that you legally upload to the Site.
This is simply calling the official YC API[0] from the end user's browser, so any user is basically doing the equivalent of clicking the Text-to-Speech accessibility button in their browser, really (albeit with better voices).
"Writers disagree on how best to define and classify atheism, contesting what supernatural entities are considered gods, whether atheism is a philosophical position or merely the absence of one, and whether it requires a conscious, explicit rejection; however, the norm is to define atheism in terms of an explicit stance against theism." (emphasis mine)
But the polls weren't wrong [1]. This was a normal polling error. If we expect polls to give us pinpoint accuracy, it's our expectations that are wrong.
Normal polling errors would lead you to believe that the actual results would fall equally and normally on both sides of the polling forecast. But they don't. The polling error is nearly always in one direction. Polling consistently underreports support for Trump in every case where we have actual results. We don't have this problem with other candidates, or even with other types of polls. There are theories about "low propensity voters" and maybe they don't show up in polls, but I suspect the real reason is more complex than that and involves some amount of self-reported filtering because of the demonization of Trump.
Perhaps Trump -- and Republican -- voters avoid taking polls or some are afraid to go on the record and admit how they're going to vote. For some there is a non-zero risk certain family members may cut off contact.
> ...some amount of self-reported filtering because of the demonization of Trump.
Indeed. He demonizes himself. I was ashamed to admit I voted for him in 2016.
I know plenty of people who have said they have permanently cut off friends and family for having conservative views. I don't know one single conservative who has said the same.
> I don't know one single conservative who has said the same.
Could it be that progressives haven't behaved badly enough that supporting them warrants cutting off contact?
Like say taking away rights from an enter gender, separating kids from their parents and locking them in cages, granting broad immunity to corrupt ex-Presidents, attempting a coup, openly calling for violence, threatening to use the military against political opponents, etc.
It can probably be explained as an indication that Democratic candidates are "normal" ones, and a win of one isn't a reason for drastic measures, while Republican candidates are "not normal", and a win of a Republican is enough grounds to cut off friends...
I'm ashamed to admit I didn't vote for him this time around because I haven't been in my home state and traveling - couldn't really get access to a ballot or mail in ballot due to my traveling.
Note that the prompt is AI-generated as well, though. From the linked blog post:
> OpenAI has another AI, GPT-3, that I used to generate many of the ideas for DALL·E prompts. I wanted to explore DALL·E using a wide variety of styles and artists, and I have limitations and biases when it comes to my knowledge of art history. GPT-3 cast a wider net of styles and artists than I would’ve come up with on my own.... The GPT-3 prompts I used evolved over time, but this one is emblematic:
> Suggest 5 unique concept ideas for a work of visual art inspired by Luke 14:7-11 (do not pick the place of honor) in the Bible. Include art direction and a specific medium and artist to emulate. Include artists from a variety of eras, styles, and media. Try for an unusual perspective. Title, year, medium. Description.
> The ʻokina has historically been represented in computer publications by the grave accent (`), the left single quotation mark (‘), or the apostrophe ('), especially when the correct typographical mark (ʻ) is not available.
While this does kinda suck, note that Marianne Williamson is excluded from many (if not all) of these same examples, while having announced her candidacy well before (and even making it into the televised debates as well). Politics aside, my guess is this is more than anything else a symptom of not knowing what to do with candidates who don't already hold an elected position.
Marianne Williamson seems like a lovely person, but she's not even close to being a serious candidate. I can't blame people for not being interested in her campaign.
I’ve seen the two talk in long form on YouTube and Yang definitely had a more comprehensive platform and policies. What Marianne Williamson was onto was not trivial either, I think there is a lot of of truth to what she’s said about healthcare “sickcare”, Loss of meaning, and unity. But Yang had a lot of very well though out positions that never really came out in 30 second soundbites.
Usually I'd see his supporters point out polling averages and/or fundraising which seems reasonable. He'd often be omitted from groups that included people with much worse metrics on both.
I don't think mass media is justified to some degree basing their coverage off polling until the primaries start.
There's always a bunch of candidates who seem credible but can't get enough support, and others who do.
Compare Yang, Williamson, Klobuchar and Buttigieg and a bunch of other candidates. Buttigieg started polling well, which justified coverage. Klobuchar didn't get any coverage until her outstanding debate performance and then her polls picked up.
The others just never polled high enough for anyone to consider them credibly able to be nominated.
Buttigieg's positive coverage started when he was polling below 2% (even when he was polling below 1%). A lot of it had to do with having the right establishment connections[1]. His success in the polls seems to be the result of the media deciding to give him a lot of positive coverage, not vice versa. The same seems to be true of other candidates this cycle (like Klobuchar and Warren).
> Klobuchar didn't get any coverage until her outstanding debate performance and then her polls picked up.
Klobuchar was getting lots of coverage, including major media endorsements (like the NY Times, which did a weird split endorsement of her and Warren), before the debate performance, and even after she's still barely registering in the polls nationally, polling at around 5% on average
President Trump's been tweeting crazy conspiracy nut things for years, like having 'extremely credible sources' saying President Obama is from Africa. Clearly tweeting crazy things doesn't disqualify one from becoming President.
But it should. Trump got through a filtering processes that should have prevented him from becoming president. This is not evidence that we should not filter crazy people out of the electoral process. If anything it’s evidence that the filter needs some work.
Voter support, as measured by contributions and surveys.
And why do voters feel as they do? Because his weirdness is possibly the result of underlying genius, while for her weirdness it’s more likely to be mental illness.
Same was said for trump and yet he managed to cancel tpp, renegotiate NAFTA, and a new trade deal with Chinese. Three of Sanders's biggest talking points.
He also made enough political changes so that Sanders went from "open borders are a Koch brothers idea , bad for the nation", to "free healthcare for illegals"
Looking at LibriVox specifically (linked to by many of the entries in OP), each audiobook has an RSS feed, so I'll just be listening to them in a podcast app. If you have a decent one that keeps track of which "episodes" you've listened to and allows downloading for offline listening (ideally automatically), seems like that should cover the audiobook usecase pretty well.
Mammals are tetrapods, which are indeed under Osteichthyes (unless you consider Ostreichtyes to be paraphyletic).
If you look at the cladogram on the wikipedia page I linked, you see a picture of a salamander near the top representing the category Tetrapodomorpha. We (and whales) are also on that branch. Sharks are outside of that whole cladogram, as they are cartilaginous fish. ( * )
In other words, any clade which includes all bony fish also includes mammals.
[*] which means that the most recent common ancestor of sharks and goldfish lived longer ago than the most recent common ancestor of whales and goldfish
The way I read it, the Bible makes it a point to communicate that Jesus was/is a real person (and that his importance actually hinges on that fact). AFAIK, this is and has always been the Church's position.
> To compare a midrange pair on quality, the Bing Search vs. a Gemini 2.5 Flash comparison shows the LLM being 1/25th the price.
That is, 40x the price _per query_ on average (which is the unit of user interaction). LLMs with web-search will only multiply this value, as several queries are made behind the scenes for each user-query.
EDIT: thanks, zahlman, he does quote LLM prices in 1M tokens, or 1k user-queries, so the above concern is mistaken!