Mehdi2277

12 ms·

61.
by Mehdi2277 5y ago
That's not been my experience. I frequently use or way and have both mypy/pyright enabled. Testing this toy function, from typing import List, Optional def f(x: Optional[List[int]]): x = x or [] reveal_type(x) mypy and pyr
62.
by Mehdi2277 5y ago
Millions? They spent a lot more than 10s of millions on advertising. They spent roughly a billion on advertising and probably are one of the most advertised products ever in spending. https://www.wsj.com/articles/tiktok
63.
by Mehdi2277 5y ago
I've spent far too much time with build errors for environment issues that I do prefer docker or other environment management tools. I just have an alias for mounting relevant folders and value of docker is not isolation for my code, b
64.
by Mehdi2277 5y ago
It's some work, but not hard to solve technically having been at companies that deal with very similar problems. The main difficulty is less technical and more investment needed vs value + investment is partly outside of the modeling e
65.
by Mehdi2277 5y ago
Contextual information is not the only reason. New content/new users is the primary reason my current work place wants more real time recommendation especially when we mostly want to recommend content created in the last day or two. It
66.
by Mehdi2277 5y ago
Streaming model training for recommendation systems like tiktok, facebook, youtube, etc. If you want models to learn quickly to new events allowing them to do better on very new content/users you will want to use a streaming pipeline.
67.
by Mehdi2277 5y ago
I'm very doubtful as my own experience would lean strongly against it having worked at multiple similar companies (facebook, tiktok, snapchat) and having even worked on recommendation system area. Internally the typical employee tends
68.
by Mehdi2277 5y ago
Patents on genes actually are a thing. So that example is pretty false. Whether they should be a thing is a separate question, but right now discovery of a gene and it's usefulness can be patented and is done for medical patents.
69.
by Mehdi2277 5y ago
The main issue here is less doing it, but getting sufficiently nice results. I've done work in generative AI before and right now the state of the art is passable on single images with some but not enough control and is still weak on v
70.
by Mehdi2277 5y ago
I work in the field. Transfer learning helps get you decent/good models, but the best models remain ones trained on large amounts of data. You may be able to get away with good performance and not great on your task. For some areas you
71.
by Mehdi2277 5y ago
I'm completely aphantasiac across all senses. I can't mentally visualize, mentally have song stuck in my head, imagine taste/touch/etc. I still dream frequently. Dreaming and lack of awake mental visualization are separa
72.
by Mehdi2277 5y ago
The recommender system mention is very valuable use case that I've seen at some of the largest ml using companies in the world. If you have a recommendation system on millions or much higher pool of content you will need an ANN index.
73.
by Mehdi2277 5y ago
I've mostly kept a couple work friends per job but I'm also more proactive than most people on staying in touch. When I was a kid I was awful at it and was a loner and eventually decided I would try harder to stay in touch with pe
74.
by Mehdi2277 5y ago
Have you taken the citizenship test? I am an immigrant and my dad had to take it. I was a bit under age requirement but did look over it. One example it doesn't even ask for the 3 branches of government, just name a single one. A lot o
75.
by Mehdi2277 5y ago
Most states in the US require a civics class in high school (normally called us government). Several states require personal financial literacy class. DIY is not one I know anywhere requiring. But something being required and something bein
76.
by Mehdi2277 5y ago
It happens because pay is balancing employee demands with company number needed. Employees in lower cost areas will very often accept lower pay. It's still often very high pay for that area. I grew up in oklahoma where median entry pay
77.
by Mehdi2277 5y ago
Because they enjoy the office. I prefer going to the office some but not most days and like occasional trips elsewhere so full remote is fine. I have friends who really like the social aspect of the office and working on a team where most o
78.
by Mehdi2277 5y ago
It's very doable to add without it and my experience is I've often worked with models trained online where the data has a retention of a few months so any old enough model will have been trained on data points that don't exis
79.
by Mehdi2277 5y ago
Junior level developers at companies I've worked at are often taught to run meetings as part of helping them grow. If a couple junior developers work on a project then having them take turns facilitating the meeting was common. My team
80.
by Mehdi2277 5y ago
The primary unsoundness is lack of types. The type checkers have strictness settings that will error if you are missing any types even types internal to a function/3rd party library used. The highest strictness settings what places are
81.
by Mehdi2277 5y ago
One of the pycon type talks was about type checking json. TypedDict or pydantic can both be used to deal with that fairly well now. For a complex json sure the type might be long but write it once and used the typeddict name after that. It
82.
by Mehdi2277 5y ago
You can have strict lint checks enforced with pre-commit hooks/ci checks. My current team I've been adding type checker tests (mypy + pyright) and for now as we're missing types I have type checkers on a nice rule of any prs
83.
by Mehdi2277 5y ago
I personally like scale as I work in ml and that helps in making datasets large. You can do a lot of ml on small datasets but life is much easier when you get to work with what feels like infinite data. It was nice having models train on tr
84.
by Mehdi2277 5y ago
I doubt it. I've worked in a couple different social media companies and default age system is ml guessing unless you do something extreme like this. That gives you reasonable accuracy, but still very noticeable error that's a lot
85.
by Mehdi2277 5y ago
The website includes a lot of examples. I think making it easier to create marketing videos (including advertisements, but also internal ones) is one use case. Another one outside marketing is making educational lecture videos is a lot of w
86.
by Mehdi2277 5y ago
Because teams want a cultural expectation of normal work hours. I am often a night owl and I’ve sent messages late at night and gotten asked my manager if I’m overworking and since then try to avoid doing it unless I’m responding to someone
87.
by Mehdi2277 5y ago
Friends and family are my largest reason for wanting to live in a location. A lot of my friends live in very high cost of living areas. So I will aim to be comfortable living in those same places. It definitely is a lot more expensive than
88.
by Mehdi2277 5y ago
I used to work at Tiktok. His speech feels consistent to me. Unlike some ceo's like Mark (fb)/Evan (snap) he rarely spoke to general employees. There was no weekly/biweekly q&a to hear from the ceo. You could easily go mo
89.
by Mehdi2277 5y ago
I care little about what the product/impact is and mostly care about how interesting the work is on a technical level. I am pretty fond of theoretical research even when practical usage is very unlikely and did a math major mostly for
90.
by Mehdi2277 5y ago
Have you ever used them? They're pretty but not things I put much trust in and the other awkward issue is there's been a variety of papers that do an analysis of many interpretability algorithms and find some very weak properties.

More ›