25 comments

[ 2.7 ms ] story [ 68.7 ms ] thread
TBH GPT/AI just felt like the most obvious topic for VCs and crypto grifters to shift their combined hype and focus. Obviously, they don’t want to devalue their work by mentioning this.
Crypto was never useful.
bitcoin/cryptocurrency is useful as a store of value under kleptocratic regimes and super inflationary economies sadly the USA has become very inflationary due to fed printing infinite $$s under the guise of pandemic rescue.

The rest alt-coins, tokens yes I agree with you are never useful except for grift

It’s only useful for criminals because it’s allowed to be converted.

The Fed didn’t print infinite money. The government borrows the money it spends and pays interest.

They borrow (a large portion of) the money from themselves a.k.a. monetizing debt.
And so do consumers when they borrow from a bank. It’s not some mysterious black box.
If you bought BTC in 2021 how much value would it be storing now?
it is not a stable coin for certain @astrange, it is not a USD replacement but it is a deflationary store of value, where the value is supposed to accrue over longer term time horizons due to it is fixed number of 21m cap of number of bitcoins. so with the store of value comes quite a bit of volatility
It’s already been demonstrated how most inflation has been caused by corporate profit seeking, much of this confirmed by corporate executives boasting to shareholders.

Please stop amplifying capitalist propaganda.

(comment deleted)
Unrelated to the article, but I had to close three popovers before I could read it.

Do the people who make these websites not realize how unusable they're making them?

And then even bigger question is: In the entire world do they think there is a single person who actually wants to interact with these popovers before reading the article?

If your performance is measured by these signup metrics, guess what a mediocre manager will do
This was exactly the problem at Twitter, as the goal is to increase the number of accounts and daily active logged-in users, they blocked the website for non-logged users using a floating popover modal box (and it was only recently changed).
I call these “KPI overlays”. As in, I need to get in the way of what you’re here for to hit my KPIs.
“Do you want fries with that?” but for out-of-touch tech people that probably don’t realise the similarities between themselves and a fast food line cook.
It seems you don’t use an ad blocker?

I would assume that it makes them enough money to sustain the business, despite what users may think of it, so that’s what they do.

While we are bashing the UX… I experienced a weird issue with scrolling that was incredibly annoying as I tried to read the article. I had to very carefully start each scroll action from a certain area of the screen, or else nothing would happen. This is on Safari on iOS.

Edit: Strange, the issue was gone after I reloaded the page. The article was much more enjoyable after that :)

This is a fun example of how much difference a comma can make. "How ChatGPT was built, from the people who made it" is what this article should have been titled. "How ChatGPT was built from the people who made it" means something very different.
haha true! but my mind mentally added the necessary comma phew!
The following quote is food for thought for people who have been wondering whether AI development needs to be regulated:

"I would love to understand better what’s driving all of this—what’s driving the virality. Like, honestly, we don’t understand. We don’t know."

I was shocked and disappointed at how casual they sounded. It seems like even the "alignment" guy was like "yolo, can't really guess what it's gonna do, so gotta just throw it out there and see".

I previously felt worried but at least glad there were some experts theoretically in control. Now I feel much less confident in OpenAI.

If these guys are as bad at AI safety as they are at PR, we've got a problem.

This article is less about how ChatGPT was built and more about how surprised openAI was by the success and why
I think this article is a missed opportunity. It focuses on the least interesting thing about LLMs—how to stop them saying things they shouldn't. I learned nothing about how it was built and its potential. It seems bizarre that there's such an intense focus and concern about how to limit LLMs instead of how to use them and improve them.