|
Sam Havens
@
sam_havens
|
|
"'putin' on them 'puters"... mostly NLP and bad jokes
|
|
|
5.283
Tweetovi
|
2.215
Pratim
|
638
Osobe koje vas prate
|
| Tweetovi |
|
Sam Havens
@sam_havens
|
2 h |
|
Soda water, bitters, and lime
|
||
|
|
||
|
Sam Havens
@sam_havens
|
3 h |
|
Blinding rage
|
||
|
|
||
|
Sam Havens
@sam_havens
|
5. velj |
|
Yes! I thought a philosophy class to middle schoolers. Two of them went on to study it in undergrad; both of them emailed me and made me cry.
|
||
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
Stephen "😭” Woods
@ysaw
|
4. velj |
|
I've worked professionally in software for 18 years and I can say with certainty that you should not use software for anything
|
||
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
Rob Napier
@cocoaphony
|
3. velj |
|
Periodic Reminder: When debugging, you must first accept that something you believe is true is not true. If everything you believed about this system were true, it would work. It doesn't, so you're wrong about something.
This is a surprisingly common stumbling block for devs.
|
||
|
|
||
|
Sam Havens
@sam_havens
|
2. velj |
|
“Until they’re gone”, what do you think the end game is if not permanent power?
|
||
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
Hilary Parker
@hspter
|
31. sij |
|
The way we talk about data science and focus so much on methods, we actually incentivize working with *bad* data, rather than spending the time to collect good data and then use easy methods with it
|
||
|
|
||
|
Sam Havens
@sam_havens
|
1. velj |
|
“I’m super excited about the paper but have him blocked on Twitter “ could that be anyone BUT Taleb
|
||
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
Phil 🕴️
@Duke_of_neural
|
31. sij |
|
Problem: Joint first co-author never works in citations.
Solution: create a corporate entity whose legal name is "(name one)_AND_(name two)" and give that corporation first author credit.
|
||
|
|
||
|
Sam Havens
@sam_havens
|
31. sij |
|
“Doing”...? Do you mean like, which drugs? If so, all of them
|
||
|
|
||
|
Sam Havens
@sam_havens
|
30. sij |
|
Credit to instagram.com/nodobird?igshi…
Not sure if they’re on twitter
|
||
|
|
||
|
Sam Havens
@sam_havens
|
30. sij |
|
We wrote a library to make it really easy to use transformer LM to rank lists of words given a context. The target audiences are researchers doing Bert-syntax experiments, and SWE without DL experience who want “smart string interpolation”
medium.com/@samhavens/int…
|
||
|
|
||
|
Sam Havens
@sam_havens
|
29. sij |
|
Looks like the operator overloading happens with the scope of a context manager (`with` block) so no worries there
|
||
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
Explosion 💥
@explosion_ai
|
28. sij |
|
Introducing the new Thinc, a refreshing functional take on deep learning!
🔮 Static type checking
🔥 Mix @PyTorch, @TensorFlow & @ApacheMXNet
⛓️ Integrated config system
🧮 Extensible backends incl. JAX (experimental)
🧬 Variable-length sequences & more
thinc.ai
|
||
|
|
||
|
Sam Havens
@sam_havens
|
29. sij |
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
Georgina Carson 👩🔬 🏳️🌈
@GeorginaRCarson
|
29. sij |
|
"Slap a Gaussian on it" - ancient data science proverb
|
||
|
|
||
| Sam Havens proslijedio/la je tweet | ||
|
inconvergent
@inconvergent
|
27. sij |
|
i hate computers.
|
||
|
|
||
|
Sam Havens
@sam_havens
|
26. sij |
|
Came here to say this. This technique works shockingly well
|
||
|
|
||
|
Sam Havens
@sam_havens
|
25. sij |
|
I generally agree that it’s a hard business. But there are some great success stories, like Sidekiq sidekiq.org
|
||
|
|
||
|
Sam Havens
@sam_havens
|
24. sij |
|
Disagree. As someone who’s always made a career of “I can do a bunch of things!”, I don’t think every job must involve specialization and certification
|
||
|
|
||