Julio Amador
@julioadl
Followers
360
Following
3K
Media
119
Statuses
3K
Lead engineer @ SiftyML
London, England
Joined May 2012
Got my first session with GPT-4, amazing! Though it failed its first causal understanding test. Me: "Is it possible that smoking causes grade increase on the average and, simultaneously, smoking causes grade decrease in every age group?" GPT-4: It is theoretically possible 1/4
18
48
340
Your periodic reminder about how big space really is: https://t.co/mNrmAoyQQ0
2
10
57
For people wondering why, as a "vision person", I am interested in language models: 1) the distinctions of different areas of AI are blurring very fast, see my earlier tweet thread: https://t.co/cJPYotUl3Z 2) language models are engines of generalization:
evjang.com
In my essay "Just ask for Generalization", I argued that some optimization capabilities, such as reinforcement learning from sub-optimal trajectories, might be better implemented by generalization...
The ongoing consolidation in AI is incredible. Thread: ➡️ When I started ~decade ago vision, speech, natural language, reinforcement learning, etc. were completely separate; You couldn't read papers across areas - the approaches were completely different, often not even ML based.
24
100
836
I will be hiring several pre-docs for 22/23 at the London School of Economics. If you are interested in development, trade and/or political economy apply here! https://t.co/rsdG9jRqPO, deadline: June 6 #EconTwitter
17
234
560
In honor of BTC getting crushed, we described the following to Dall-E AI: “Crypto bro working at McDonald's in the impressionist style" This is what it came back with…
292
320
4K
The new Machine Learning Specialization by @DeepLearningAI_ & @StanfordOnline is now available on @Coursera! You’ll learn the fundamentals of ML & gain practical experience with Python. Please help me spread the word, and encourage others to take ML! https://t.co/qHZoEqmWm2
150
1K
6K
Ok. Deep breath. I think we may look back on this as the first Great Information War. Except we're already 8 years in. The first Great Information War began in 2014. The invasion of Ukraine is the latest front. And the idea it doesn't already involve us is fiction, a lie. 1/
2K
24K
61K
At this link you can find ways to support Ukraine via donations. Let's put our money where our mouth is. My family lives in Galati (Romania), about 15 km from the SW border of Ukraine. Reach out if someone you know is entering Romania at Galati. https://t.co/eSXGepUsLP
linktr.ee
Volodumyr Bogomolov
I had a📞talk with President @ZelenskyyUa to assure him of 🇷🇴RO's full support&solidarity in face of the irresponsible military aggression by RU, and for UA sovereignty& territorial integrity. We coordinate with our partners to rapidly support UA🇺🇦 through humanitarian assistance
0
8
37
The cyber attack that has rendered the largest international logistics company in the U.S. totally non-operational since last Sunday seems under-reported.
392
4K
19K
If you don't remember retweeting this, it means that you have leaked your Twitter Access Token in a public GitHub repository. Not the best practice, right? For details, read our latest article: https://t.co/6WBC6DRNDS
#InfoSec #CyberSecurity #GitHub
76
5K
470
Teachers of courses in Microeconomics: My book with Martin Osborne "Models of Microeconomics" (for undergraduates or master students) is available for free downloading from my homepage
12
247
992
Our latest Twitter engineering blogpost, coauthored with @mmbronstein is now available: Graph Neural Networks as Neural Diffusion PDEs
1
44
180
Join us on Tue @icmlconf to see how to go beyond traditional message passing in graph neural networks:
1
23
175
Can you guess what happens when you provide homeless people with housing? ... ... ... You reduce crime, increase employment, improve health, and do not increase reliance on social benefits. 80% of costs are offset by the benefits in the first 18mos. https://t.co/kNt5PxLWYQ
31
950
3K
Making this lousy experience shorter is a very valuable feature for founders. But most VCs don't realize it's a feature at all. To most VCs, speed is merely something that they have to do in order not to lose deals.
9
20
571
The reason speed matters is that fundraising is, as a rule, an unpleasant experience. It's painful in the wrong way. It's the pain of a nail in your foot rather than soreness after a long run. And it consumes the one thing a startup can least afford to waste: attention.
8
63
778
The main reason late-stage funds like Tiger and Coatue are eating traditional VCs' lunch is that they're faster. Speed is probably the number one difference between what founders think is important in VCs, and what VCs think is important in VCs.
59
292
2K