Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
echen
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
8 ms
·
31.
▲
by
echen
4y ago
Completely agree on the need for serious commitment and attention! Funnily enough, though, many ML engineers and data scientists I know (even those at Google, etc., who depend on human-annotated datasest) aren't familiar with these kin
32.
▲
30% of Google's Emotions Dataset Is Mislabeled
(surgehq.ai)
334 points
by
echen
4y ago
|
144 comments
33.
▲
To solve the water crisis, companies are increasingly turning to AI
(fortune.com)
2 points
by
echen
4y ago
|
1 comments
34.
▲
Building Better Developer Search: How Neeva Measures Search Quality
(surgehq.ai)
5 points
by
echen
4y ago
|
0 comments
35.
▲
by
echen
4y ago
I've experienced the same. Here's a larger collection of obvious spam failing to get flagged by Gmail's spam filters that we gathered: https://www.surgehq.ai/blog/are-the-spammers-winning-failure... For
36.
▲
by
echen
4y ago
Interestingly, we've actually helped other AI companies build robust violence detectors precisely to prevent these kinds of violent, gore language model generations. You can read about it here https://www.surgehq.ai/blo
37.
▲
by
echen
4y ago
This is DALL-E 2. The original DALL-E isn’t released by OpenAI.
38.
▲
Generating Children’s Stories Using GPT-3 and DALL·E
(surgehq.ai)
138 points
by
echen
4y ago
|
145 comments
39.
▲
Google moves to keep campaign messages out of spam
(axios.com)
1 points
by
echen
4y ago
|
0 comments
40.
▲
Sustainable AI: Environmental Implications, Challenges and Opportunities
(research.facebook.com)
1 points
by
echen
4y ago
|
0 comments
41.
▲
Inverse scaling prize for finding a task where larger language models do worse
(github.com)
3 points
by
echen
4y ago
|
0 comments
42.
▲
‘Anti-hunger’ molecule forms after exercise, scientists discover
(med.stanford.edu)
289 points
by
echen
4y ago
|
268 comments
43.
▲
Book Review: San Fransicko
(astralcodexten.substack.com)
4 points
by
echen
4y ago
|
0 comments
44.
▲
The delivery apps reshaping life in India’s megacities
(technologyreview.com)
2 points
by
echen
4y ago
|
0 comments
45.
▲
Meta Made Millions from Ads That Spread Disinformation
(wired.com)
7 points
by
echen
4y ago
|
1 comments
46.
▲
Humans vs. Gary Marcus: The Complexity of Measuring Machine Intelligence
(surgehq.ai)
3 points
by
echen
4y ago
|
0 comments
47.
▲
Meta has built a new language AI–and it’s giving it away for free
(technologyreview.com)
3 points
by
echen
4y ago
|
2 comments
48.
▲
Senate bill would ban data brokers from selling location and health data
(arstechnica.com)
3 points
by
echen
4y ago
|
0 comments
49.
▲
On the Advance of Making Language Models Better Reasoners
(arxiv.org)
2 points
by
echen
4y ago
|
0 comments
50.
▲
How We Built It: OpenAI's GSM8K Dataset of 8,500 Math Problems
(surgehq.ai)
4 points
by
echen
4y ago
|
0 comments
51.
▲
A founder’s guide for picking the right startup law firm
(carta.com)
2 points
by
echen
4y ago
|
0 comments
52.
▲
A Face Search Engine Anyone Can Use Is Alarmingly Accurate
(nytimes.com)
7 points
by
echen
4y ago
|
6 comments
53.
▲
Sentiment Analysis Dataset of Social Media Stock Conversations
(surgehq.ai)
2 points
by
echen
4y ago
|
0 comments
54.
▲
Are the Spammers Winning? Failures in Gmail Spam Detection
(surgehq.ai)
10 points
by
echen
4y ago
|
0 comments
55.
▲
Restoring and attributing ancient texts using deep neural networks
(nature.com)
1 points
by
echen
4y ago
|
0 comments
56.
▲
by
echen
4y ago
Yeah, the human-drawn scientists all have an evil gleam or element of craziness to them... I missed that myself until you said it!
57.
▲
We asked 100 humans to draw the DALL·E prompts
(surgehq.ai)
138 points
by
echen
4y ago
|
73 comments
58.
▲
Use NLP to tackle online toxicity now
(txt.cohere.ai)
3 points
by
echen
4y ago
|
0 comments
59.
▲
by
echen
4y ago
Quality control is often indeed by volume. When I was at Google / Facebook, search evaluations would typically have 5 raters each. One of the difficulties, though, is that search evaluation can be quite subjective. And honestly, most o
60.
▲
by
echen
4y ago
It can be suprisingly difficult. A lot of the difficulty is in understanding the user intent behind a search query (since you're usually rating other people's queries, not your own). For example, here are some queries in my histor
More ›