Alec Radford ✓
@AlecRad · San Francisco, CA · joined 23 Oct 2012
72 819Followers
305Following
565Posts total
15.3KViews on collected posts
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
1.6K19 Dec
9.4K24 Feb
4.3K29 Mar
Last 3 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.62%19 Dec
0.28%24 Feb
0.52%29 Mar
Reactions — likes, reposts, replies and quotes — divided by views.
What the audience does
Likes91.9%57 in total
Replies1.6%1 in total
Bookmarks6.5%4 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Latest posts
@AlecRad narrator: they did in fact train a 1b parameter model on 10,000x the data to get it to generalize to solve penn treebank
4.3K views · 21 likes
· 0 reposts · 1 replies
29 Mar 2026
· Open on X →
@AlecRad why did leave OpenAI bro？
1.6K views · 10 likes
· 0 reposts · 0 replies
19 Dec 2024
· Open on X →
@AlecRad @Scriptonite_ https://t.co/N4RmqGcwVr
0 views · 3 likes
· 0 reposts · 0 replies
27 May 2019
· Open on X →
@AlecRad Twitch chat is... interesting. Hopefully we can get a histogram of PogU / minute - could be a nice secondary training signal :)
0 views · 9 likes
· 0 reposts · 1 replies
25 Apr 2019
· Open on X →
This is a really fun live experiment with twitch chat predictably oscillating between love and hate based on the sample. https://t.co/5XPCPcEcrx
0 views · 214 likes
· 14 reposts · 19 replies
25 Apr 2019
· Open on X →
@AlecRad This is actually legit, I think, as PTB is very specific news from quite a long time ago. Very tight domain.
0 views · 23 likes
· 0 reposts · 0 replies
17 Feb 2019
· Open on X →
By the way - I think a valid (if extreme) take on GPT-2 is "lol you need 10,000x the data, 1 billion parameters, and a supercomputer to get current DL models to generalize to Penn Treebank."
0 views · 590 likes
· 58 reposts · 16 replies
17 Feb 2019
· Open on X →
The DL CV community is having a "oh wait, bags of local features are a really strong baseline for classification" moment with the BagNet paper.
This has always been clear for text classification due to n-gram baselines. It took an embarrassingly long time for nets to beat them.
0 views · 411 likes
· 72 reposts · 5 replies
11 Feb 2019
· Open on X →
Nice discussion of the progress in NLU that's happening with BERT, OpenAI GPT, ULMFiT, ELMo, and more covered by @CadeMetz in the @nytimes I'm super excited to see how far this line of research will be able to get in the next few years!
https://t.co/v0Raiygv0p
0 views · 167 likes
· 45 reposts · 0 replies
19 Nov 2018
· Open on X →
What I've been working on for the past year! https://t.co/CAQMYS1rR7
Inspired by CoVE, ELMo, and ULMFiT we show that a single transformer language model can be finetuned to a wide variety of NLP tasks and performs very well with little tuning/tweaking.
0 views · 1.9K likes
· 449 reposts · 50 replies
11 Jun 2018
· Open on X →
Similar accounts
Dustin Grage ✓
@GrageDustin
145.6K followers
· 74.4K posts
Tato Young
@eltatoyoung
145.6K followers
· 10.9K posts
KEVIN USA 💯 ✓
@kevinUSA100
145.6K followers
· 1M posts
dogwifcoin ✓
@dogwifcoin
145.6K followers
· 8.7K posts
Ms. Grace Kuhlenschmidt
@GKuhlenschmidt
145.6K followers
· 4.2K posts
Acer ✓
@Acer
145.5K followers
· 41.4K posts
Spectrum ✓
@GetSpectrum
145.5K followers
· 38K posts
Sampson the Service Dog ✓
@sampson_dog
145.5K followers
· 22.9K posts