I've been using #ChatGPT since December 2022 and #Codex since late 2025. They've become my primary tools for product work and side projects. They don't just help me write—they help me think faster, explore more ideas, and build more.
@jasonhickel This explanation is misleading. e.g. arg1 is applicable to Ancient Greece and Roman Republic — they existed much earlier than capitalism. Also, arg1 is applicable to all socialistic/communistic orders. Regarding arg3 — modern democracy is a sub product of capitalism
@sashachapin Ultimately, the experience you describe is likely to be subjective and may vary from person to person. It may be more useful to try to describe the specific qualities or characteristics of this experience in order to convey what you are trying to express
@sashachapin It's difficult to pinpoint a specific word that encapsulates the experience you describe. However, some words that may come close to expressing the concept you are trying to describe include: Bliss, Ecstasy, Nirvana, Serenity
@MattNiessner In addition to working on a specific research project, pursuing a PhD also involves taking coursework, participating in seminars and workshops, and possibly teaching or assisting with undergraduate courses
@tunguz It is important to carefully consider your fundraising goals, how you will promote the campaign, and how you will use the funds that are raised. It is also important to be transparent about your campaign and to use the funds in a responsible and ethical manner
@olivercameron Overall, AI has the potential to transform many different industries by automating tasks, improving efficiency, and providing insights that were previously not possible
Totally agree on this. Would add that trees are better explainable which allows you to tune them much easier + manage performance params (i.e. RAM consumption)
Most GBT implementations can deal with them right out of the box. Some also have inbuilt mechanism for dealing with categorical data. For most other ML algos you need to do a LOT of judicious preprocessing before you unleash the ML on them. 2/3
What matters most when training a neural network is how well it generalizes to unseen data.
For neural networks, it turns out there's a simple principle that can allow you to understand model generalization. (1/18)
A thread ↓
@Thom_Wolf However it’s quite difficult a) to track new techniques, b) compare their impact, c) combine into one algorithm, and d) test the contradictions. So, it would be good to have an API that takes care of all those problems and allows me just to assemble an algorithm like lego
@Thom_Wolf There are frequent enhancements in RL, e.g. on ICLR’22 I’ve found two useful papers, one on policies learning (https://t.co/MmfVmyb9Mk) and another one on env resetting (https://t.co/HsfUplkpce)