pants2

23 ms·

541.
by pants2 2y ago
My dishwasher has buttons on the top like this, and during the heat dry cycle the steam will activate the buttons and I'll hear lots of random beeps from the kitchen. Ponce in a while it manages to cancel or restart itself, hilariously
542.
by pants2 2y ago
For me the window is laggy and takes a few seconds to resize or move. It's also not reliable when prompting sometimes it just sits there with no output and I have to quit and restart. I've since uninstalled the desktop app and I&#
543.
by pants2 2y ago
On the flip side, the ChatGPT desktop app is native on Mac and it's buggy and feature-incomplete compared to the web version.
544.
by pants2 2y ago
Yes, just as my 10-yr-old old Java Applets and Flash games run perfectly... oh wait.
545.
by pants2 2y ago
Presumably every step has to also read the tokens from the previous steps, so it gets more expensive over time. If you run it on a single task for an hour I would not be surprised if it consumed hundreds of dollars of tokens.
546.
by pants2 2y ago
Any anecdotes about how many $ of API credits this thing costs to run for a simple task like booking a flight?
547.
by pants2 2y ago
Basically, if it means companies can introduce automation without changing anything about the tooling/workflow/programs they already use, it's going to be MASSIVE. Just and install and a prompt and you've already automat
548.
by pants2 2y ago
It's crazy to me how I can be completely stumped on a crossword puzzle, and then after a jog I come home and knock it out. Really kicks your brain into high gear.
549.
by pants2 2y ago
That's really strange. I don't see any planes in gates at LGW either, but spot checked other airports that have planes in gates.
550.
by pants2 2y ago
For an early prototype it's very cool and fast AF - good job!
551.
by pants2 2y ago
Or there's a 0.1% chance it becomes a $160T company if they're the first to develop superintelligence
552.
by pants2 2y ago
I'd much rather talk to ChatGPT than a human support rep, provided they have the same level of ability (tools) to help you.
553.
by pants2 2y ago
It's incredible that people are talking about the downfall of software engineering - now, at many companies, hundreds of call center roles will be replaced by a few engineering roles. With image fine-tuning, now we can replace radiolog
554.
by pants2 2y ago
I could see it being extremely insulting to be grading ChatGPT-written assignments.
555.
by pants2 2y ago
Agreed, it just seems like Nvidia chips are going to be easier to produce at scale. Cerberas will be limited to a few niche use-cases, like HFT where hedge funds are using LLMs to analyze SEC filings as fast as possible.
556.
by pants2 2y ago
Just don't try to center text on the page!
557.
by pants2 2y ago
It's worth mentioning that the simulations and results were done by and judged by LLMs. It makes sense to me that LLMs would come up with responses that other LLMs think is the most logical.
558.
by pants2 2y ago
Transaction volume goes up and down. Just look at ETH gas prices - varying between 1 and 1000 month-to-month. BTC holders aren't going to be OK with a temporary lull in transactions leading to a double-spend attack etc.
559.
by pants2 2y ago
That was from February. Can't seem to find anything more recent.
560.
by pants2 2y ago
Bitcoin can't always be proof-of-work. Given the capped supply, the mining rewards will eventually diminish to the point that the network is no longer secure. It will have to transition to proof-of-stake or some other mechanism long
561.
by pants2 2y ago
I had the pleasure of visiting a town on the Amazon river a few times over the course of a decade. I watched as western culture and civilization creeped in and ruined their society. The first time I went, people were living off the land, fi
562.
by pants2 2y ago
I do find that the rule-based, classic CV "lane assist / adaptive cruise control" of other makes is preferable, because it's extremely predictable. Tesla's full neural model is less predictable.
563.
by pants2 2y ago
There's a pretty big spectrum between perfect driving and a collision. FSD could drive on the wrong side of the road for miles with no accident because it's an empty road, or oncoming traffic swerved to avoid you.
564.
by pants2 2y ago
Here's the ImageNet recognition challenge results from 2011[1]. Bird is actually one of the easier categories they they call out, but still calling it 'solved' is a bit of an exaggeration. The winning model is from Xerox Rese
565.
by pants2 2y ago
Most code is grounded in real-world concepts somehow. Imagine an engineer at Domino's asking it to write an ordering app. Now your model needs to know what goes in to a pizza.
566.
by pants2 2y ago
There are few use-cases where voice-only would be an improvement to screens. Voice-only is extremely limited in its discoverability and the information that it can relay to you. Imagine going to a restaurant and not getting a menu, instead
567.
by pants2 2y ago
Seems like the best self-driving car you can get today is a new-ish Toyota with Comma.ai installed.
568.
by pants2 2y ago
Indeed, in the early days of Bard, the only way to get it to output only JSON was to threaten a human life[1]. 1. https://x.com/goodside/status/1657396491676164096
569.
by pants2 2y ago
Unfortunately since most of our jobs can be done remotely, that would likely drive a ton of software jobs overseas.
570.
by pants2 2y ago
Your best bet in the US is to use Polymarket with a VPN

More ›