Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
cs702
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
22 ms
·
331.
▲
by
cs702
1y ago
Maybe. I'm no longer so sure.
332.
▲
by
cs702
1y ago
EDIT: After reading the Matt Levine piece posted by https://news.ycombinator.com/item?id=44497566 , I'm no longer sure this is market manipulation. The devil is in the details, and the details disagree with my initial
333.
▲
by
cs702
1y ago
Thank you for sharing this. Levine is usually pretty accurate about this stuff. I've changed my mind. See my updated response here: https://news.ycombinator.com/item?id=44499330
334.
▲
by
cs702
1y ago
Every provider of AI as a service (AaaS), including OpenAI, has no choice but to spend money like a drunken sailor who recently touched land after being out at sea for a year. The reckless spending is not fueled only by hype and FOMO. Custo
335.
▲
Jane Street to contest Indian regulator's manipulation charges
(ft.com)
4 points
by
cs702
1y ago
|
0 comments
336.
▲
by
cs702
1y ago
Hardly a sophisticated strategy, indeed, but Jane Street was earning like $1B/year of profit on it, according to https://news.ycombinator.com/item?id=44483691
337.
▲
by
cs702
1y ago
According to Indian regulators, every trading day Jane Street would: 1) buy large volumes of stocks and/or stock futures that are part of an index tracking India’s banking sector, early in the day, 2) subsequently place large optio
338.
▲
by
cs702
1y ago
ROFL. Thank you for sharing that link!
339.
▲
by
cs702
1y ago
> If you don’t like something, don’t buy it. The OP's point is that increasingly, we don't have that choice, for example, because AI slop masquerades as if it were authored by human beings (that's, in fact, its purpose
340.
▲
by
cs702
1y ago
Your may agree or disagree with the OP, but this passage is spot-on: "I don’t want AI customer service—but I don’t get a choice. I don’t want AI responses to my Google searches—but I don’t get a choice. I don’t want AI integrated into
341.
▲
by
cs702
1y ago
Not sure I agree with all of the OP's opinions. I prefer a clean, calm, uncluttered user interface over a noisy, busy, cluttered one. In the OP's example with maps, I'd rather see a full-screen map, instead of a map that is a
342.
▲
by
cs702
1y ago
Nice. I'm impressed. Translator jobs are going to go poof! overnight. Just sayin'.
343.
▲
by
cs702
1y ago
Compared to what I think the daily trading volume "should" be (approximately $0.00), $23M sure is a lot!
344.
▲
by
cs702
1y ago
The new token, USELESS, makes Dogecoin look dignified by comparison. USELESS was launched via LetsBONK.fun, has no utility, no roadmap, and evidently no shame, but it has amassed 12K holders, reaching $23M in peak daily trading volume. You
345.
▲
by
cs702
1y ago
I like sama and many other folks at OpenAI, but I have to call things how I see them: "What Meta is doing will, in my opinion, lead to very deep cultural problems. We will have more to share about this soon but it's very importan
346.
▲
by
cs702
1y ago
If you ask the Xfinity managers who came up with this idea whether thieves will be able to buy live information on whether your home is empty from hackers on the dark web, the managers will likely say... nothing. What they will do is look
347.
▲
by
cs702
1y ago
It could also be that switching model behavior from "good" to "bad" internally requires modifying only a few hidden states that control the "bad to good behavior" spectrum. Fine-tuning the models to do somethin
348.
▲
by
cs702
1y ago
TL;DR: Fine-tuning an AI model on the narrow task of writing insecure code induces broad, horrifically bad misalignment. The OP's authors fine-tuned GPT-4o on examples of writing software with security flaws, and asked the fine-tuned
349.
▲
by
cs702
1y ago
For large tech companies, the investment is not about earning a return. It's about not missing out. It's about staying relevant. It's about avoiding failure. Execs don't want to wake up one day and find their products
350.
▲
by
cs702
1y ago
The author of the study has written a blog post with lots of data, analysis, and visualizations: https://len-sherman.medium.com/how-uber-became-a-cash-genera... I found the author's blog post more informative than the
351.
▲
How Uber Became a Cash-Generating Machine
(len-sherman.medium.com)
2 points
by
cs702
1y ago
|
1 comments
352.
▲
by
cs702
1y ago
Is that the only reason why entrance was denied? Is this story real ? I just checked, and nope, it's not April 1 yet. Does anyone here have addition information?
353.
▲
by
cs702
1y ago
Judging by my own experience, this is spot-on : > When [home rental marketplaces] started, it was kind of wonderful to be able to book somebody’s house or apartment. They cared about it. They left you gifts and flowers and long notes a
354.
▲
by
cs702
1y ago
It seems they used the most recent models available as of March 2025.
355.
▲
by
cs702
1y ago
Interesting. Even the most recent models perform relatively poorly when asked to identify which information in a context has been removed, given access to both the original and edited contexts. The authors posit that poor performance is due
356.
▲
by
cs702
1y ago
"Your personal messages, calls and statuses, they will remain end-to-end encrypted," says the Facebook executive in charge of WhatsApp. A company representative adds that it has "no plans to place ads in chats and personal me
357.
▲
Rise and Fall of Social Networks 2003 – 2022 [video]
(youtube.com)
1 points
by
cs702
1y ago
|
0 comments
358.
▲
by
cs702
1y ago
Maybe. AI models have continued to scale up at a rapid rate, and have continued to get better at performing ever more impressive tasks. Sure, yes, the OP is breathless corporate-speak, but given how much impressive progress we've see
359.
▲
by
cs702
1y ago
Thanks. You're right.
360.
▲
by
cs702
1y ago
340 slides. That's a lot , even for Meeker. Many of the slides are packed with details. Who, exactly, has time to read all that stuff? Only AI models, I'd guess.
More ›