Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
geraltofrivia
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
9 ms
·
31.
▲
by
geraltofrivia
3y ago
If you go through the links, specifically [3], you will find pretty objective definitions of multiple perspective of fairness. This is a mathematical concept.
32.
▲
by
geraltofrivia
3y ago
The point I was trying to make was that certain NLP models (maybe these LLMs as well) might give better results if YOU speak to them as a middle aged white man. To be very clear, I want ALL queries (I presume you mean LLM prompts) to be ava
33.
▲
by
geraltofrivia
3y ago
These are what debiasing tasks are concerned with, more often than not. RLHF tuning depends greatly on the H part of it, and that data is probably proprietary. So, I guess time will tell. But if I were to hazard a guess based on the content
34.
▲
by
geraltofrivia
3y ago
Yes. In my case, my particular worldview involves (i) indifferent to gender markers in text (I don't want my LLM to convert me to a female because I'm asking it to write a cover letter for a hairdresser/secretary/nurse r
35.
▲
by
geraltofrivia
3y ago
I may be wrong, and if so, correct me. But do you use ChatGPT as a recreation thing, or as a part of a real world solution? In my current job, we've started offloading a lot of real work to GPT4 (it just works ). Sometimes this is solv
36.
▲
by
geraltofrivia
3y ago
That's not _completely_ true. There are a bunch of datasets (specific to some cultural/lingual contexts) like CrowsPairs[1], StereoSet[2]. There is a lot of work you can do to make sure that the model's predictions are fair a
37.
▲
by
geraltofrivia
3y ago
I like how they brush off VERY REAL concerns of bias, mis-representation as having "spicy takes" and "a rebellious streak". More power to anyone financing the training of an LLM but if you're too lazy to red-team it
38.
▲
by
geraltofrivia
3y ago
I'm not super technical but if someone knows, I would love to know how does it pull the video from Youtube/Twitch in the first place? Like wouldn't Youtube try and block sending data to this application somehow? Wouldn't
39.
▲
by
geraltofrivia
3y ago
Thats absolutely not a conspiracy theory. It’s a well known, very real phenomenon. To the point where FTC has warnings about similar things [1] [1] https://consumer.ftc.gov/consumer-alerts/2023/05/scammers-ar.
40.
▲
by
geraltofrivia
3y ago
HuggingFace or HF was a incorporated entity writing some well engineered and easy to use Deep Learning libraries. Notably, around the time when the Transformer paper was released, and when the NLP community started using them en masse, HF’s
41.
▲
by
geraltofrivia
3y ago
Yes indeed. Like would you like the heated pan or the actual fire? /s
42.
▲
by
geraltofrivia
3y ago
It really sounds like: if you care about Privacy, why do you care about anything else? I want both, privacy and convenience of a Mac OS. I shouldn’t have to compromise.
43.
▲
by
geraltofrivia
3y ago
It’s not too far off the actual response: “I must admit, I have rarely come across someone with a more unique perspective on things. It’s truly intriguing to witness such different ways of thinking.”
44.
▲
by
geraltofrivia
3y ago
Obsidian is great. My notetaking solutions went from - raw text files - raw markdown files - + homemade server rendering the markdown live - Obsidian - Notion I stick with Notion because it really is infinitely configurable. I rarely use 99
45.
▲
by
geraltofrivia
3y ago
It's not the most intuitive export (since there are backlinks, complex hierarchy, databases and all that jazz in Notion pages) but they do return Markdown for every page, CSV for databases (with some information loss), maintain folder
46.
▲
by
geraltofrivia
4y ago
Explosion's core contribution (not moneymakers) is the great SpaCy library they first released in 2015. It was an excellent work, far better designed (IMO) than NLTK and other offerings at the time. Of course the library isn't mon
47.
▲
by
geraltofrivia
4y ago
HuggingFace is also hosting models (CLI/code/API), have an extensive community driven dataset store, and has some great positive feedback loops. Great Library -> Large Community -> Large Coverage of Models & Datasets -&g
48.
▲
by
geraltofrivia
4y ago
Was available on Linux, yes. Performed much worse though, AFAIK.
49.
▲
by
geraltofrivia
4y ago
If foreshadowing is one of their intention with the teaser, I'm afraid they'll push raytracing very hard. Their trailer had a lot of 'reflect in water', 'reflect in glass panes' moments.
50.
▲
by
geraltofrivia
4y ago
Right? Like mixed-use housing Or even pedestrian only streets ! Who can imagine! For a Finnish studio, I would have hoped Colossal Order focused a bit less on American cities, American urbanism.
51.
▲
by
geraltofrivia
4y ago
We need to get off of our high horses for a hot minute, I think. When big tech offers increasingly personalized results, we can’t afford being bothered by IP based location specific results. The thing is, I want local results. I use DDG as
52.
▲
by
geraltofrivia
4y ago
In hindsight this make a lot of sense. Growing up, it didn't. I did my schooling in a small Indian city where from Grade 3-5 (that's approx 8-10 year old) we were given access to Logo in the computer room and instructed to make th
53.
▲
by
geraltofrivia
4y ago
Some sell technology, some sell vision. The latter involves trusting them/submitting to their vision which would put off a lot of developers, I feel, but still create intrigue in managerial circles for them to get traction. I can’t rem
54.
▲
by
geraltofrivia
4y ago
I wouldn't comment on the absence of French vis-a-vis other languages. It's just slightly surprising because English <-> French is honestly a very widely studied translation sub-task, with an enormous amount of parallel corp
55.
▲
by
geraltofrivia
4y ago
Probably because its easy and common to Unionise in Germany, they don’t have to go through hell and high water to form a collective (as silly as it sounds in the first place!)
56.
▲
by
geraltofrivia
4y ago
They do because they can go to the executives and deny unreasonable goals outright, something to the effect of ‘how do you think four people with 39h of work weeks can do this’, implying that these hours are not going to increase so let’s h
57.
▲
by
geraltofrivia
4y ago
In my previous job in Germany. Sector: Tech and Research. Excellent managers, amazing WLB (worked till 8PM once and the manager apologised next day), well paid: earning the median income straight out of college. Had an excellent union and
58.
▲
by
geraltofrivia
5y ago
My parents are tech illiterate, and very scared of being scammed. When I mention like “you can listen to any song you like, amongst millions” because I set up family sharing on their iphones, and ask them to use Apple Music they ask what do
59.
▲
by
geraltofrivia
5y ago
> Temporary distractions might cause me to zone out and miss a part of the story, and "rewinding" to listen to the same part again isn't as easy as just re-reading a paragraph. I agree. I think it also explains why I find
60.
▲
by
geraltofrivia
5y ago
Apologising in advance to derail from the point of this comment. Lately, I've been giving audiobooks a go. It really does make "consuming" a book much, much easier. Listening during commute/chores is a huge plus. But als
More ›