Reddit had the benefit of subreddit moderators policing their own. Twitter has no such thing. Maybe if you squint enough, big accounts block/muting bad actors in their replies can sort of count as self-policing but that does not prevent the bad actor from being a troll in someone else's replies.
That's very impressive. Hats off to them! I dont think this is too out of the ordinary either though. I'd guess they started off with a LLM from hugging face, set up some pipeline to ingest code from replit repos to finetune the LLM. The ML aspect of this is not terribly hard given that they probably dont need to train a LLM from scratch. Figuring out how store and serve from replit repos (or publicly available code bases) is not too difficult. From there it's a matter of productionalizing: how to serve the model in real time, figuring out they want the product to look/feel like and I suppose this part of it might take a while. I'd estimate you'd need 1-2 ML engineers, 2 data engineers, 2-3 swes, 1 PM for the team for a minimal viable product.
Right, but humility is crucial in machine learning (and particularly in the area of self driving cars) where you're constantly calibrating uncertainty and risk with lives on the line.
I mean the answer to this is not as simple as "yes trash belongs in trash bins".
How do you organize trash pickup? Where do these receptacles get placed? How large should they be? Is this for business and residential? Do trash routes need to change as well? NYC is massive and what will probably work in a neighborhood in Manhattan may not work for one in Staten Island. It's not at all a simple exercise.
For a brutal education in this, do standup comedy. It's very easy to find open mics in most major cities. You'll quickly learn through audience feedback. For something a bit more gentle, watching tv/movies is also quite good.
How does one actually program on an FPGA? I know for GPU, you can buy an nvidia graphics cardm, rent an aws gpu machine or use google colab. Is there some consumer facing equivalent for FPGA programming?
I'm curious to read more on the 3-4 other deep learning projects you mentioned that have had a larger impact on your fields. Can you share some links to those works?
Consider how laws are set. Politicians answer to their constituents. If the people who live in the city do not want certain things (short term Airbnb rentals here), you can expect them exert pressure on their representatives to make those things go away. That's exactly what has happened in NYC.
The question of "Why do you think that they should have less rights that people that live in the city full-time?" doesnt really matter. Because the reality is that they have no power/sway over the politicians who are responsible for governing the city.
I still have yet to use graphical models (in the traditional sense, not including the new age variational inference style neural networks as graphical models) in real life. Am I just completely missing something? Where do people generally find compelling uses for graphical models?
I've always found it odd that basic arbitrage (ex: buy asset A on some exchange and sell it immediately on a different exchange and earn the delta as profit) and sandwich attacks (place a buy order before a whale places a big order and place a sell order immediately afterwards) were under the umbrella of MEV. These are trades that are possible in tradfi too. I suppose it's possible that arb trades and sandwich attacks are more of a concern here where miners can insert their own transactions into the mempool but I don't know if we actually see miners engaging in this behavior.
There's "knowing C syntax" and then there's knowing C well enough to be able to read this code and understand what's going on. One of these days I'd like to get to the point where I can conceivably read this code and understand it well enough to build my own matrix library using similar optimizations for the rest of the matrix operation primitives. Some day ...
Rec systems that tech giants use looks vastly different than what ML academics study. Industry rec systems are a constantly evolving beast. Why? Because these teams of ML engineers are constantly running A/B tests to tweak some aspect of the model. The resulting model likely doesnt fit nicely into something category of model you can describe as collaborative filtering/lambdamart/etc b/c it's some highly performant glued together mess.
Best model of this I have is math professors from undergrad. I never saw any notes on their desks during office hours. Just pages and pages of scribbles everywhere, crumped on the ground, in the trash bin, etc. I remember asking one how he retained any of the info when reading dense papers, and he said "Oh I never do. I just figure out the details again each time and eventually it sticks."
Hmm, I had a similar situation with an ex-partner contaminating my recommendations. I emailed support and it was pretty easy to reset my recs to a blank slate.
Describing Marc Lore as "Former Walmart executive" is true but selling him short. He founded diapers.com (sold to Amazon), and later Jet (sold to Walmart).