replicanteven

Shortsighted. If they're willing to stoop this low, they should have started by selling multiple kinds of badges for smaller monthly amounts.

$0.99/mo for emojis. $1.99/ to show your support for various causes. $3.99/ for sports teams. $4.99/ for ID verification. $0.49/ of your red cross badge goes to earthquake survivors.

Like virtual hats, but for social media.

It's more like class than caste, because it's possible to move between classes after birth. Modern advertisers are sort of similar to a priestly/censor class, in the same way that serial C-suite/board members are sort of like an aristocratic class.

Discussing that sort of thing is discouraged in the US, because it stinks of nepotism and conflicts with our culture's ostensible ideals of meritocracy and egalitarianism.

It's an easy mistake to make, because in order to obscure our society's important class distinctions, we make a lot of noise about other differences that people are born with, which feels very caste-y.

The problem is that people who are willing to pay $20/mo for privacy are much more valuable to advertisers than those who can't.

If those users go away, the average $/user from ads goes down, because the only people seeing ads are those who are too poor to avoid it.

To an advertiser, poor people who direct their attention towards whatever is put in front of them are worth a lot less than rich people who carefully curate what their attention is directed towards. It's grody, but that's the advertising industry for you.

Homeless people are often unbanked, and it's very difficult to open a new account without a credit score and permanent address.

Recent immigrants have difficulty opening accounts for a similar reason - only US credit counts. Political activists are also often denied accounts out of an abundance of caution on the bank's part, e.g. with the Occupy Wall St protests.

TBF, you need working hands to use hands-on computers.

Plus, the offline part could make a good starting point for a DIY personal assistant.

That said, their "getting started" sounds...esoteric.

>There normally isn't any output but you should be able to type "hey" by saying "hoof eve yank" and transcribe a sentence after saying "scribe". You can terminate it by pressing Ctrl+c or saying "troll cap".

Hmm. It looks like natural gas is a little less than $2.50 / GigaJoule (~29.3kWh) at the moment. One A100 eats about 400W, and apparently OpenAI used about 1023 of those running for 34 days to train ChatGPT[1].

There's also a bunch of supporting equipment, so maybe they used around 500W * 1023 * (34*24=816)hrs = 417,384kWh?

I think that would be around $35k in fossil fuels, assuming none of the datacenter's energy came from renewable sources?

[1]: https://www.tooltester.com/en/blog/chatgpt-statistics/

In the transcripts I saw, it seemed like Sydney repeated itself a lot. It used the same sentence structure a lot. Stephen Wolfram wrote a good primer on LLMs, which was recently discussed here a lot:

https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-...

To my eye, BingGPT's personality looked similar to the early "low-temperature" examples in that article. Like the model got stuck overfitting to a topic that acts like a local maxima of word probabilities.