Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
jaredtn
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
8 ms
·
31.
▲
by
jaredtn
6y ago
Credit card chargebacks really hurt, and are the appropriate response to a situation where they continue to take your money.
32.
▲
by
jaredtn
6y ago
Native virtualization and the ability to run iPhone and iPad apps directly on the Mac. This could unlock a whole new level of focus and optimizations for Apple.
33.
▲
by
jaredtn
6y ago
Seems like Intel is dying an accelerating death. The unified software stack across Mac, iPhone, and iPad apps might unleash a new level of focus and innovation for Apple.
34.
▲
by
jaredtn
6y ago
Which extensions were compromised? Not seeing a list anywhere.
35.
▲
by
jaredtn
6y ago
"As further proof, features from the model achieve state-of-the-art performance on a number of classification datasets and near state-of-the-art unsupervised accuracy on ImageNet." Impressive stuff! This performs well even without
36.
▲
by
jaredtn
6y ago
Agreed, to rephrase my concern: By default accounts don’t even have the option to enable E2E encryption in a meeting. There’s no button to press that will turn on encryption if nobody’s dialing in. Your user account has to go through a sepa
37.
▲
by
jaredtn
6y ago
It’s still only opt-in. Users have to submit an application (including text message verification and other personal info) to gain access to E2E encryption. Zoom has shown that it does not care about privacy.
38.
▲
by
jaredtn
6y ago
Interesting, don't see that listed on https://hey.com/pricing/ . Where are those policies listed?
39.
▲
by
jaredtn
6y ago
Agree that it adds complexity - nobody wants a repeat of Java enterprise apps from years ago. In the world of scientific computing and recommendation systems, it's a godsend. Writing your own sort function and your own min function get
40.
▲
by
jaredtn
6y ago
$99
41.
▲
by
jaredtn
6y ago
That's the thing - tough to know how long a bubble lasts for. And choosing a time horizon based on the court system is difficult because bankruptcy proceedings can drag on for a long long long time.
42.
▲
by
jaredtn
6y ago
Good point. Are you making that trade?
43.
▲
by
jaredtn
6y ago
A $2 put for Jan 22 is currently trading with a 1.60-1.75 spread, meaning the break-even point is $0.25-0.40. Not much reward to have there. No shares available to short either.
44.
▲
by
jaredtn
6y ago
I would say opaque rather than complex . The details of a CDS aren't complicated, nor is the Eurodollar system. But it can't be inspected from the outside, and even the Fed is reliant on running complex models in an attempt to
45.
▲
by
jaredtn
6y ago
In response to this specific event, or as a general bonus? I would love to hear any more details you can share.
46.
▲
by
jaredtn
6y ago
Interesting. How does that work when you want to rest your forearms while typing?
47.
▲
by
jaredtn
6y ago
What will the continued support for this project be, given that Uber has shuttered their AI Labs?
48.
▲
by
jaredtn
6y ago
CycleGAN: https://github.com/junyanz/CycleGAN
49.
▲
by
jaredtn
6y ago
Not sure why you’re being downvoted. Agree with the message, but “consistently” means 5-10 years, not just a few months of concerted effort.
50.
▲
by
jaredtn
6y ago
Can you explain further?
51.
▲
by
jaredtn
6y ago
Generalized Pretraining. The original GPT paper doesn’t use the acronym, but got rebranded as GPT in retrospect once GPT-2 came out.
52.
▲
by
jaredtn
6y ago
32GB RAM is nowhere near what's necessary. It can barely fit the GPT-2 models on there with a batch size in the single digits. We'll need extensive model parallelism libraries (Zero2 from Microsoft) to run this at all.
53.
▲
by
jaredtn
6y ago
They use extensive model parallelism when training. Even TPUs (64GB) or Tesla V100 GPUs (32GB) don’t have enough memory to fit a model into a single child, so you’ll need activation checkpointing or model parallelism.
54.
▲
Training a Lite Bert on AWS
(aws.amazon.com)
2 points
by
jaredtn
6y ago
|
2 comments
55.
▲
by
jaredtn
6y ago
Yeah, interesting the things people want to punish for. The inverse isn’t quite a grace.
56.
▲
by
jaredtn
6y ago
Ah, sorry for that. I meant E[return on option - cost of option] ~= 0.
57.
▲
by
jaredtn
6y ago
Great point! The "uncle point" is a great phrase I hadn't known before.
58.
▲
by
jaredtn
6y ago
True, but you have to account for the chance of ruin. Once you go broke, you have to stop betting. Many traders will continue betting larger and larger amounts of money when they win, but stop betting if they lose big. Making risky bets is
59.
▲
by
jaredtn
6y ago
If I roll a 100-sided dice that pays me $99 if I roll a 1 and -$1 if I roll 2-100, then my expected value is $0, but my chance of profitability is 1%.
60.
▲
by
jaredtn
6y ago
No, I mean the mathematical mean, not mode. If you take this action infinitely many times, what is your average (mean) return? https://en.wikipedia.org/wiki/Expected_value
More ›