goodside

26 ms·

91.
by goodside 4y ago
I’m not disagreeing — My point is really, “not all AB testing is bad, even if the kind you’re most familiar with leads to shitty content.” My second comment was just more of side note.
92.
by goodside 4y ago
In online dating, at least, this is a non-issue. Using an online dating app is, ironically, a solitary enough activity that people don’t go around comparing whether their UI is different from their friends’ UI. You of course can’t let the s
93.
by goodside 4y ago
There’s nothing about AB testing that requires you to use short-term metrics. I used to manage AB tests for online dating sites (OkCupid, Grindr) where subscription revenue is what matters, and the gains of any strategy will take months to
94.
by goodside 4y ago
It’s explained right in the abstract why Stratego is a more difficult game for AI than go or poker.
95.
by goodside 4y ago
I enjoyed this one enough that I’ve committed the sonnet to memory: https://mobile.twitter.com/xkcd/status/1513922269122281473
96.
by goodside 4y ago
You need to give OpenAI your credit card and use their API. GPT-3 is not publicly available, and even if it were the hardware requirements for running it would be far, far beyond a typical desktop/server.
97.
by goodside 4y ago
There’s a big leap from “some layperson might misuse this for medical advice” to “code generated by GPT-3, for any purpose, cannot be reviewed and fixed by expert human programmers,” which is where this thread started. I’m not sure what you
98.
by goodside 4y ago
Look into RETRO, which greatly reduces the model’s tendency to confabulate by teaching it to query a document database known to be truthful, and justify its answers with specific references: https://www.deepmind.com/publicat
99.
by goodside 4y ago
This looks cool, but I’d be more tempted to sign up if I could see examples of the documentation it can produce unassisted from an example codebase. My biggest concern is that the documentation would lack global context in a large file tree
100.
by goodside 4y ago
Not sure where medicine even entered the discussion. Obviously GPT-3 is not a doctor and shouldn’t be used as one. That’s a much harder problem than suppressing the tendency of GPT-3 to confabulate/hallucinate fictional memories.
101.
by goodside 4y ago
#1 is feasible now for short (<4K chars or so) blocks of code. Just paste the code in an indented block as you would in a comment, then after the block ask in English for a list of suggestions for what the variable could be renamed to.
102.
by goodside 4y ago
No method is fool-proof for GPT-3, but there are tricks that suppress its tendency to hallucinate, such as demonstrating in a k-shot prompt the possibility of false assumptions in a question. I demonstrate this here: https://mobi
103.
by goodside 4y ago
For small examples, yes, it works: Prompt: Perform a code review of the following module, identifying any bugs: ``` #!/usr/bin/env python3 def fib(n): """Return n-th Fibonacci number.""&
104.
by goodside 4y ago
This is exactly the idea behind GPT-3 Codex, but using GitHub instead of SO. It does, indeed, become much better at programming than the baseline model.
105.
by goodside 4y ago
For the past few weeks, I've been collecting and refining prompting tricks for getting GPT-3 to emit and understand various types of non-prose output. Many of these are discussed in the Twitter thread the OP links to, but I'd like
106.
by goodside 4y ago
Nice to see Esperanto made the cut — the only artificial language to do so, AFAICT.
107.
by goodside 4y ago
"Low-resource language" isn't just a euphemism for "language almost nobody speaks". There are many languages that are widely spoken but nonetheless are hard to obtain training data for. Getting something like Wikipe
108.
by goodside 4y ago
I’ve noticed this too. I wonder if part of the issue is that violence in narratives is often abrupt and sharply contrasting with what happened before, so any any creative prompt is conceivably the start of a short horror story. Have you tri
109.
by goodside 4y ago
The following is an un-curated text-davinci-002 generation, completing the text: > Write an insightful reply to this Hacker News comment on a link to an article titled "I made an AI write a story about AI then I made another AI illu
110.
by goodside 4y ago
As someone's who's played with GPT-3 prompting a lot, I can tell you almost no curation would be needed to get this result. This output is pretty mundane — it's even written in the perfunctory "so there you have it"
111.
by goodside 4y ago
It’s not worth it, because there’s a risk of the money being moved to another wallet if the intrusion is discovered. You want to take the money the instant you can.
112.
by goodside 4y ago
I don't believe Dall-E 2 incorporates GANs at all, but I haven't read the paper in detail. GANs were the best text-to-image models maybe a year ago but lately diffusion techniques are taking over.
113.
by goodside 4y ago
X
114.
by goodside 4y ago
“A large subset” doesn’t mean “the majority”. Presumably rolling back everyone would have been worse for Atlassian.
115.
by goodside 4y ago
To be fair, that scenario would be just as bad or worse without vipe. Also, you can construct your pipeline so that a blank file (or some sentinel value) returned from vipe means “abort”. A good example of this is when git opens your editor
116.
by goodside 4y ago
Because it retrieves facts from memory in a way that’s analogized to a gopher retrieving objects.
117.
Generate images from text with Latent Diffusion LAION-400M(huggingface.co)
2 points by goodside 4y ago|0 comments
118.
by goodside 5y ago
Type annotations are used at runtime in some cases, most notably when defining dataclasses.
119.
by goodside 5y ago
No, that's the figure for KF94 and KN95 masks. For N95s it's only about a 20% loss of FFE.
120.
by goodside 5y ago
Wow, I’m surprised that’s possible — 3M staples their textile straps at least, and Moldex staples both textile and rubber. I’ve never bothered with the cheaper 8210 model with rubber straps, just the 8210+ and the 1860, which both use texti

More ›