Hacker News: sepiasaucer

Azure Databricks Pricing Error

sepiasaucer — Fri, 24 Feb 2023 16:23:41 +0000

The advertised Azure Databricks pricing for NCas_T4_v3 series (NVIDIA T4 GPUs) is incorrect on the website. See https://azure.microsoft.com/en-us/pricing/details/databricks/

The advertised prices are much lower than they should be. The basic formula for Databricks is:

(Total Price) = (VM Price) + (DBU Price)

I am not sure if they are billing less than they should be or billing more than they advertise. If you used this series on Azure Databrick you may want to check your billing.

Comments URL: https://news.ycombinator.com/item?id=34926183

Points: 2

# Comments: 0

New comment by sepiasaucer in "Rising Tether Loans Add Risk to Stablecoin, Crypto World"

sepiasaucer — Thu, 01 Dec 2022 21:30:46 +0000

From Matt Levine:

“Doesn’t that Journal story sound a bit like that? I mean here is a story you could tell:

1. You have 1,000 Bitcoin worth about $17 million.

2. You want to buy more Bitcoin, but you do not have any dollars.

3. You go to Tether and say “hey give me 17 million USDT, in exchange I’ll put up 2,000 Bitcoins as collateral.”

4. Tether is like “sure that’s the business we’re in” and hands you 17 million USDT.

5. You use that 17 million USDT — notionally worth $17 million — to buy 1,000 more Bitcoin.

6. Now you have 2,000 Bitcoin.

7. You post the 2,000 Bitcoin as collateral to Tether for the loan, which is now overcollateralized with liquid collateral ($34 million worth of Bitcoin).

8. More USDT have been created to buy Bitcoin, but no new dollars have come into the system.

Maybe this is fine, no problem, just margin lending. [4] But if your concern is “Tethers are printed out of thin air in order to allow people to buy crypto without putting any actual dollars in,” then this might make you nervous.”

Ask HN: Recommended reading for/against possibility of AGI

sepiasaucer — Sun, 16 Oct 2022 17:42:30 +0000

I am seeking recommended readings regarding the possibility of creating AGI (artificial general intelligence). In particular, I would like to read intelligent, scientifically supported arguments by experts across the spectrum from those who believe AGI is in the very near future to those who believe it is impossible. It would also be interesting to read older sources to see how perspectives have changed over time.

To clarify, I am mostly interested in reading scientific/technical arguments about if AGI can be created, not ethical/moral arguments about if AGI should be created.

Comments URL: https://news.ycombinator.com/item?id=33225528

Points: 1

# Comments: 0

New comment by sepiasaucer in "TerraUSD crash led to vanished savings, shattered dreams"

sepiasaucer — Wed, 22 Jun 2022 15:09:18 +0000

A [crypto] Ponzi scheme is an investment fraud that pays existing [crypto] investors with funds collected from new [crypto] investors.

New comment by sepiasaucer in "Ask HN: Recommend employers with positive social impact"

sepiasaucer — Thu, 26 May 2022 21:08:37 +0000

Sounds like a cool place to work! I interned at SpaceX in undergrad and worked at ONERA while getting PhD. I work as Data Sci now. I see an open Data Sci position that looks like good fit. Do you think I could negotiate for being remote or is the on-site policy pretty strict?

Ask HN: Recommend employers with positive social impact

sepiasaucer — Thu, 26 May 2022 14:44:40 +0000

I am interested in finding a job at a company that is having a positive impact in the world. I think this probably rules out FANG (MAMAA?) companies. Does your work make at least a small contribution to a better world? If so, where do you work?

Comments URL: https://news.ycombinator.com/item?id=31518945

Points: 169

# Comments: 194

New comment by sepiasaucer in "The parents who refuse to give their kids smartphones"

sepiasaucer — Mon, 23 May 2022 00:28:41 +0000

Seems like maybe you would want to teach kids about safe smartphone use instead of abstinence only

New comment by sepiasaucer in "Index funds officially overtake active managers"

sepiasaucer — Sun, 22 May 2022 21:30:14 +0000

I don’t understand what point you are trying to make.

Specific stocks being bad investments is not an argument against passive investing. The whole point of index funds is to diversify your portfolio so you track the overall market, not any specific stock or group of stocks.

If you are assuming you know which stocks/sectors are under or overvalued, then I guess active investing makes sense, but that seems like a flawed premise to start from.

New comment by sepiasaucer in "The Rules for Hybrid Work Were Always Made Up"

sepiasaucer — Sun, 22 May 2022 17:05:55 +0000

If employees are “hybrid” without mandatory attendance, then I think most people won’t go into office. I work in a smaller satellite office and if you randomly go to office you may be only person there.

I think employers thought “hybrid” was good way to get employees to agree to going back to office. In reality, I think it was good way for employees to get employers to agree to effectively fully remote work without explicitly demanding it.

It’s like in Office Space:

PETER I, uh, I don't like [the office]. I don't think I'm gonna go anymore.

JOANNA You're just not gonna go?

PETER Yeah.

JOANNA Won't you get fired?

PETER I don't know. But I really don't like it so I'm not gonna go.

JOANNA LAUGHS) SO YOU'RE GONNA [go fully remote]?

PETER No, no, not really. I'm just gonna stop going.

https://youtu.be/jKYivs6ZLZk

Four pricey NFTs stolen from actor Seth Green

sepiasaucer — Wed, 18 May 2022 03:49:18 +0000

Article URL: https://web3isgoinggreat.com/?id=four-pricey-nfts-stolen-from

Comments URL: https://news.ycombinator.com/item?id=31419053

Points: 3

# Comments: 1

New comment by sepiasaucer in "Twitter Deal Temporarily on Hold"

sepiasaucer — Fri, 13 May 2022 16:34:44 +0000

I don’t think you could easily prove Twitter is materially misrepresenting the number of bot/spam accounts. Presumably, it is just an estimate based on some combination of assumptions and statistical analysis. You might be able to create a significantly higher estimate, but that seems different than proving material misrepresentation.

Analysis: Musk's new Twitter funding could draw TikTok-like U.S. scrutiny

sepiasaucer — Sat, 07 May 2022 17:18:04 +0000

Article URL: https://www.reuters.com/business/musks-new-twitter-funding-could-draw-tiktok-like-us-scrutiny-2022-05-06/

Comments URL: https://news.ycombinator.com/item?id=31296549

Points: 4

# Comments: 0

Musk’s Twitter ambitions to collide with Europe’s tech rules

sepiasaucer — Wed, 27 Apr 2022 17:09:09 +0000

Article URL: https://apnews.com/article/elon-musk-boris-johnson-technology-business-online-safety-f687d652d41678e25575ba67119bba6b

Comments URL: https://news.ycombinator.com/item?id=31182733

Points: 2

# Comments: 0

“A Honeypot for Assholes”: Twitter’s 10-Year Failure to Stop Harassment (2016)

sepiasaucer — Tue, 26 Apr 2022 17:33:49 +0000

Article URL: https://www.buzzfeednews.com/article/charliewarzel/a-honeypot-for-assholes-inside-twitters-10-year-failure-to-s

Comments URL: https://news.ycombinator.com/item?id=31170465

Points: 5

# Comments: 0

New comment by sepiasaucer in "What solutions to content moderation would be better than Elon Musk’s?"

sepiasaucer — Wed, 20 Apr 2022 16:00:02 +0000

I believe they (Twitter etc) are working on moderation. I do not know if that is their core competency or if they want to be in the content moderation business. I was wondering if a 3rd party might provide content moderation as a service.

Transparency is very important. If you can’t explain your solution then maybe a simpler, but less accurate method is preferred. Maybe you could set some defaults but allow more user control. Maybe you could have humans in the decision making process, but still integrate with ML models for scale. I don’t know best solution, but offering content moderation as a service would shift some blame away from social media platform even if it was not better than what they were doing in-house.

What solutions to content moderation would be better than Elon Musk’s?

sepiasaucer — Wed, 20 Apr 2022 13:49:19 +0000

It seems a lot of people in tech believe the ability for anybody to share largely unmoderated content on large social media platforms is very important for free/open online communication. At the same time you could argue there is a desire/need to combat spam, disinformation campaigns, propaganda, hate speech, AI generated content, etc.

Much of the discussion around content moderation is about the many challenges, sometimes suggesting it is an impossible problem to solve. A lot of discussion also suggests content moderation is inherently bad because “free speech”. However, lots of problems are technically challenging and/or ethically challenging (e.g., self driving cars), but are not discussed the same way.

A perfect solution to content moderation likely does not exist, but I am curious what companies are working solely on providing content moderation as a service and what innovative solutions are being proposed. Also, what is the state of the art in content moderation research assuming such a thing exists?

Comments URL: https://news.ycombinator.com/item?id=31096921

Points: 7

# Comments: 9

New comment by sepiasaucer in "Crypto Scammers’ New Target: Dating Apps"

sepiasaucer — Mon, 21 Feb 2022 17:33:34 +0000

I would focus more on the enabling technology for this scam (e.g., how would a victim transfer $300,000 using traditional financial system? how they might recover those funds if needed?)

Crypto Scammers’ New Target: Dating Apps

sepiasaucer — Mon, 21 Feb 2022 15:45:44 +0000

Article URL: https://www.nytimes.com/2022/02/21/technology/crypto-scammers-new-target-dating-apps.html

Comments URL: https://news.ycombinator.com/item?id=30417115

Points: 27

# Comments: 5

New comment by sepiasaucer in "Researchers Build AI That Builds AI"

sepiasaucer — Tue, 25 Jan 2022 22:44:19 +0000

“One can, in theory, start with lots of architectures, then optimize each one and pick the best. “But training [takes] a pretty nontrivial amount of time,” said Mengye Ren, now a visiting researcher at Google Brain. It’d be impossible to train and test every candidate network architecture. “[It doesn’t] scale very well, especially if you consider millions of possible designs.””

——-

If you wanted to find best architecture in order to maximize accuracy, why not just train a model to predict accuracy (not parameters) given architecture and then optimize over the model?

This seems similar to optimizing any expensive black box function. Fit a cheap approximation (i.e., surrogate model) and then optimize over cheap model.

New comment by sepiasaucer in "Five Levels of Hype"

sepiasaucer — Sun, 23 Jan 2022 21:18:23 +0000

“Where does that fit in your categories?”

It sounds like it falls mostly under illegal transactions because you are using crypto for something that would otherwise be illegal (avoiding capital controls). Is it technically illegal? I don’t know. Is it ethically wrong? I guess not directly. However, I imagine there are also a lot of nefarious reasons somebody might want to avoid capital controls which crypto could also help facilitate.