Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
losteric
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
20 ms
·
241.
▲
by
losteric
3y ago
I think the transitions from punch cards to CLIs, to GUI/mouse interfaces, then to touch interfaces, are more apt metaphors when considering the impact of VR. Touch found it's place as the interface for small-form devices. I'
242.
▲
by
losteric
3y ago
Honestly this sounds like a shower-thought post. With even basic research, Internet Archive and The Eye have Reddit historical data freely available. My desktop PC has all comments and posts from 2007-early 2023, in a convenient jsonl zst.
243.
▲
by
losteric
3y ago
I presume you are blaming the American federal government? Why? "The supply chain" is a global emergent phenomena across a variety of capitalist nations. Personally, I would blame corporate greed that took advantage of short-term
244.
▲
by
losteric
3y ago
My parents are technical sheep. They'll do what a site tells them to, even if it bogs down their phone, adds notifications, and inserts yet another advertising tentacle into their life. They won't be mad because they don't un
245.
▲
by
losteric
3y ago
Are you saying that rent seeking and anti-competitiveness are "not capitalist"? How do you figure?
246.
▲
by
losteric
3y ago
Isn't explicit subreddit-level engagement and information organization incredibly valuable for targeting/monetization? imo, it like they're disorganizing the site to undermine and eventually cut out independent moderators, to
247.
▲
by
losteric
3y ago
Also a great way to catch commercial use of non-commercial weights, I would be surprised if secret phrases were not already trained in.
248.
▲
by
losteric
3y ago
Do you use any tools to keep current on all these sites?
249.
▲
by
losteric
3y ago
It's a self-endorsement without any contribution towards the discuss topic. I didn't want to flag the poster as they seem new, but this was explicitly against the community guidelines. Imagine if everyone that had a blog, newslett
250.
▲
by
losteric
3y ago
I think there's at least two pieces here 1. Concern about loss off creativity in creative exercises: this is a fair UX critique. Personally, I find that using ChatGPT as a rubber duck is extremely valuable, even for complex "softs
251.
▲
by
losteric
3y ago
Disclaimer: hoerzu made this. Kinda uncouth promoting by replying to an unrelated high visibility post.
252.
▲
by
losteric
3y ago
Can you elaborate? I don't follow how the amount of money spent "makes it effectively illegal" to audit.
253.
▲
by
losteric
3y ago
that's likely a major motive behind RTO, yes - voluntary attrition and termination with cause (ie no severance, unlike layoffs and pip). A large walk out would probably boost share pricing.
254.
▲
by
losteric
3y ago
The vast majority of "bootcamp types" just add overhead to the time/cost of finding a good hire, I doubt they meaningful contribute to any "market saturation".
255.
▲
by
losteric
3y ago
only in the chat app. Per TOS, data sent to APIs is not used for training. All thesee services undoubtedly use the API
256.
▲
by
losteric
3y ago
Yes. All the FAANGs have such roles. Sure, most PEs are "applied" engineering or management right-hands, but the same companies have PEs up to VP/Distinguished Engineers deep in technology... React came from Facebook, Google
257.
▲
by
losteric
3y ago
Generally, the models seem to struggle with fine details that correlate with broader features. Each finger typically looks ok, it's the hand that's off... remote controls might have appropriately detailed buttons with a nonsense l
258.
▲
by
losteric
3y ago
> I'd also speculate that our current LLMs can say whether a piece of text looks like a fact. I tested it with davinci-003 ``` Label these statements as fact, lie, or unknown. 1. There are 10 days in a 10 day period 2. Michael Jack
259.
▲
by
losteric
3y ago
It will be interesting to see how training horizons impact development of new software languages, frameworks, libraries. I have already found that I am much more productive using older patterns that Copilot "knows", compared to mo
260.
▲
by
losteric
4y ago
Framing LLMs as primitives is marketing-speak. These are high-level construction for specific runtimes, which are difficult to test and subject to change at anytime.
261.
▲
by
losteric
4y ago
yeah, I believe some readers are misinterpreting the report. The OS manages mmap, it won't show up as "regular" memory utilization because it's lazy-loaded and automatically managed. If the OS can keep the whole file in
262.
▲
by
losteric
4y ago
Are the weights on NVME? Old SSD? HDD?
263.
▲
by
losteric
4y ago
I've wondered, what's the right way to pronounce this library? "face"?
264.
▲
by
losteric
4y ago
What are your critiques of the OP article? This bill seems indisputably bad. One could just as easily say vague American "powers that be" were stopped rushing the law through congressional lobbying propaganda.
265.
▲
by
losteric
4y ago
I'm all for socialism, communism. That's what you're proposing, right? Otherwise you're saying "can't trust the status quo, can't trust the alternative" - noise, at best.
266.
▲
by
losteric
4y ago
> I assume anything Apple is cooking is using Nvidia in the server room already For training, sure. For inference, Apple has been in a solid competitive position since M1. LLaMa, Stable Diffusion, etc, can all run on consumer devices tha
267.
▲
by
losteric
4y ago
Maybe that is the disease? greed, sociopathy... money worship
268.
▲
by
losteric
4y ago
There's the OIG dataset ( https://laion.ai/blog/oig-dataset/ ) which was used to train a NeoX 20B ChatBot ( https://huggingface.co/togethercomputer/GPT-NeoXT-Chat-Base-... ). The dataset is
269.
▲
by
losteric
4y ago
Is it possible to teach GPT3 to do something like this through prompts? Like, responding with an API call to get information when it doesn't have enough? The closest I've gotten is having the model output both the be API call and
270.
▲
by
losteric
4y ago
This link does not cover pgvector
More ›