flat-pluto
I found this[1] to be a good resource. Another thing that has helped me is adding a prefix to my review. Classifying it as: suggestion, nitpick, props, etc. More about this here[2].
[1] - https://stackoverflow.blog/2019/09/30/how-to-make-good-code-...
Another example of this is how homeowners are treated vs renters. Specifically, in terms of tax breaks for mortgage interest payments, expenses incurred during major home improvements, etc. Also you can withdraw a decent chunk of money from your IRA early without the hefty penalty if the withdrawal goes toward buying a home. On the other hand, renters are not eligible for any such benefits.
Teachers have a difficult job as it is, not to mention that they are already mistreated - they have almost no authority over students since they never have backing from parents or school administrators.
I recently came across a post on Reddit[1] where a student pepper sprayed her teacher because he confiscated her phone during an exam. One of the commenters claimed that the same teacher was previously punched in the face by a different student for taking their phone away after catching them cheating on a test. It is disconcerting to say the least.
[1] https://www.reddit.com/r/facepalm/comments/1397n4l/teacher_t...
Anyway, this[1] is a really good introduction to Prompt Engineering and how one can tailor prompts for their use-case.
Interesting podcast episode that touches on this stuff - "In an age that favors the formulaic and generic to the ambiguous, complex, and unexpected, it's no wonder that computers can sound eerily lifelike. Leslie tells EconTalk host Russ Roberts that we should worry less about the lifelike nature of AI and worry more that human beings are being more robotic and predictable."[2]
[1] - https://www.nocode.ai/introduction-to-prompt-engineering/
[2] - https://www.econtalk.org/ian-leslie-on-being-human-in-the-ag...
Earlier today there was a Show HN post[1] which showed how to visualize a Pandas dataframe (can come from CSV, JSON whatever). I tried it for basic tasks and it is pretty good. It's minimal code (<5 lines) - just reading the json and calling pygwalker in a Google Colab environment[2] or something. Something like this:
import pandas as pd
import pygwalker as pyg
df = pd.read_json('{filename}.json')
gwalker = pyg.walk(df)
Groupthink, herd mentality, conformity > Meritocracy
This should be a fun couple of years, sigh!
As outlined in the article, common sense solutions do exist so the problem isn't technical per se. Is it just that not enough people are using these datasets for it to be an important enough problem to solve? Or is it more of a bureaucratic/political issue where its difficult to come to a consensus among these departments?
[1] - https://www.kalzumeus.com/2010/03/20/running-a-software-busi... [2] - https://www.kalzumeus.com/2012/01/23/salary-negotiation/
Email is in my bio.
Sidenote - if you can find the time, you should write more often. I just went through your articles and there is a lot of useful advice to be found. The projects are pretty interesting too. Cheers!
I'd say this is just the first stage of creating something, an MVP of sorts. After that you do need to get some feedback, iterate and improve it step-by-step to get the finished product.
1. Government - it'll prolly be hidden from the public under the garb of national security, and used for god knows what
2. For-profit companies - Google, Meta, MS, etc who have shown themselves to be untrustworthy and purely motivated by profits
3. Non-profit companies - OpenAI (and others?) who have explicitly made the commitment to optimize for long-term benefit, safety, and cooperation [1]
I'm not saying there is a right answer, but the non-profit route seems like the least worst option.
> Non-profit or not, AGIs shouldn’t be controlled by anyone.
I don't see that ever playing out in the real world. It is much too powerful a tool to be made accessible to everyone.
The creator states - "They were interpreted by AI. We've aimed to be as accurate as possible, but we can't guarantee accuracy or validity."
I know of several Android phones that have this feature (it is off by default though).
Just want to stress that training an ML model is going to be really difficult for this scenario, and false positives are to be expected. It isn't like this was unexpected either given that "the phone flew off its handlebar mount".
Sidenote - Only a few years ago, I thought it'd be damn near impossible to compete with FB + IG in the social media space OR with Google + FB in the digital ads industry. On both counts though, I've had to rethink my position given the rise of TikTok, Amazon[1] and Apple[2].
[1] - https://www.statista.com/statistics/1305698/amazon-advertisi... [2] - https://www.statista.com/statistics/1330127/apple-ad-revenue...
4) I've also received discounts which helped us save on query and storage costs (20% off above a certain threshold).
As a hobbyist, I have used Google Colab which has been pretty good as well. (Don't know if it officially comes under GCP).
This is known but is worth repeating - GCP is under-pricing its products as a way to gain market share, and once they have a larger piece of the pie, they are likely to increase prices.
In the near future, I imagine it'd be easy enough to take a snippet of audio and transfer a celebrity's voice to it, similar to what we can already do to images using StyleGAN or something similar.
Exciting times ahead!
I guess its another source of feedback that can validate findings from other data. Like getting the data points explicitly from the user as opposed to inferring it based on their behavior.
Or sometimes they may not have other data. So for example - I'm based in Chicago and while watching YouTube on my TV, I get asked something to the effect of "Which store do you visit most often? a.Foxtrot b.Whole Foods c.Jewel Osco ..." presumably because there is no way for YT to have access to this kinda info.
"At the peak of our 'great reshuffle' we saw a year-on-year increase of 50% of LinkedIn members changing jobs. Gen Z was at 90%".
90% seems absolutely bonkers.
As for the new avatar, it's just the first draft so to speak. Given the talent at Reality Labs, I'm certain that things will look much better a few years from now. People often have this vision of what VR/AR should look like and will be disappointed that v1.0 does not live up to those expectations. As long as incremental progress is made, it will compound over time and eventually we'll get there.
The most roundabout way to determine one's caste that I have come across is inviting someone to swim. Brahmins, the upper-caste Hindus, traditionally wear a thread[1] which is on display when you swim. The ones who don't wear a thread are assumed to be non-Brahmin.
Anyway, once the caste is known subtle discrimination may follow (depending on the people of course).
1. You'll get to learn something new by building it yourself. 2. It will instill a sense of discipline in you which is invaluable.
Think of it like working out, even though it may not be exciting it is almost always a net positive.
It has been a part of Indian society for a very long time, going back millennia, but it was codified in its modern form by the British. For instance:
"Nicholas Dirks has argued that Indian caste as we know it today is a modern phenomenon as caste was fundamentally transformed by British colonial rule. According to Dirks, before colonial rule caste affiliation was quite loose and fluid, but colonial rule enforced caste affiliation rigorously, and constructed a much more strict hierarchy than existed previously, with some castes being criminalized and others being given preferential treatment." [1]
Preferential treatment in this case means educational opportunity, administrative jobs, etc - essentially avenues of upward socio-economic mobility. Since then, it has almost become ingrained in Indian culture. So much so, that people think that one caste is "better" than the other, marrying into another caste would bring disgrace to the family and community, and so forth.
Another factor to consider is affirmative action. There are few good public universities and jobs in India, and a certain percentage of them are reserved for groups who have been historically discriminated against. This kinda rubs upper-caste people the wrong way since some of them believe casteism no longer exists and that even if reservations are required, they should be based on the socio-economic status as opposed to one's caste.
Overall, as you can tell by now, it's a pretty tense matter. Things are improving but by no means is it a non-thing.
I didn't hear of it either until I noticed large groups of Indian-origin devs and managers (mostly at older companies such as IBM, Oracle, etc), and somehow only the 'upper caste' folks would get promoted and be invited to happy hours and such. How much that can be attributed to inter-personal relationships and professional competence as opposed to outright discrimination, one will never know but a pattern does emerge. It's never in your face discrimination, just subtle favoritism towards one's group.
Mind you, this happens with people from the same Indian state as well. They'll often speak in their regional language during meetings even though not everyone speaks that language, which simply goes on to alienate that team member. (This has happened to friends of mine at T-Mobile and Intel)
"marriage sites have done more to break cast barriers in india"
Not sure if I agree with you there. It has brought to light just how prevalent that kind of thinking is. It's a common trope on Indian matrimonial websites how users have a list of requirements (Looking for fair skinned, Hindu upper caste, etc). Maybe that has helped break cast barriers by bringing it to light, I don't know, but it was really surprising when I came upon it for the first time.