AmIDev

Can someone explain what this kind of comment is meant to be and what is the history behind it? It looks grammatically incorrect, and I can't reorder the words to make sense either.

I feel there might be intelligent species out there without either the resources to develop technology or the ability to do so. Sharks could become smarter than us one day, but without thumbs they aren't going to succeed in overtaking the planet. Similarly, there could be a humanoid fish species under thick ice sheets of Europa, and they would never guess there is anything more than their "ocean" in the universe.

LLMs are extremely useful for ML classification.

1. If you have a large amount of unlabeled data, labeling it using LLMs(even a costly XXXB parameter model) can be significantly cheaper than using humans.

2. Text classification works far better with LLMs compared to usual techniques like random forests.

3. We are exploring using LLMs on structured data(tables) for tasks like clustering, where it is difficult to tune other unsupervised approaches. Similarly we are exploring using LLM embeddings for similarity on unsupervised datasets.

4. LLMs could explain their decisions to users(e.g. why was your comment removed?) although work needs to be done on verifying it's correctness.

I wouldn't be surprised if LLM takes the spot of Random forest as the default go to for supervised ML.

If scalpers are able to sell a product at higher price, doesn't that mean the company priced the product too low?

Name is not the only thing that gives away one's caste/religion. For example, my ear being pierced gives away that I am a Hindu. The specific dialect I speak gives away where I belong from, which statistically brings me to 2-3 castes. The food I eat(e.g. I eat chicken but not pork) or the festivals I celebrate will give away the rest.

You can do something similar in India as well. The benefit is that you get the flat at cheaper rate, at the risk of losing all your investment. There have been many cases of the developer going bankrupt, but nothing at the scale of evergrande.

When learning classical computing, I have done the following things that gave me a deeper understanding of how things work.

1. Learned logic gates and built(in simulators) small circuits which can do addition/multiplication.

2. Used a 8085 board to write assembly programs for search/sort etc.

3. Learnt C programming and Operating systems(primarily Linux)

4. Learnt higher level programming languages and paradigms(OOP, compilation, etc).

What set of courses/topics would lead to a similar level of understanding in the quantum domain? I have learnt about the quantum gates, but I do not have to context to understand how they fit in the larger picture.

This is something I struggle with. I have a weekly meeting which includes TL, Manager, PM and sometimes Staff engineers, Data analysts, Managers of downstream teams etc.

I try to keep the details to the lowest denominator, which usually means mentioning what the problem is and what the proposed solution is in few lines. But eventually someone will ask for more details, and then the conversation veers off to technical discussion that I am sure the PMs and Analysts don't understand.

Even that should be equivalent to distributing throughout the economy. If you burn 100 units of money, you are decreasing it's supply, increasing it's value by a tiny bit.

I believe CS degrees shouldn't focus too much on industry skills. Sure it is good to have some engineering focused courses that are likely to be useful in long term(C programming, design patterns), but most courses should be theory heavy.

If you take any specific domain of CS, there are only going to be few hundred good professors on that topic at most in entire world. Them spending their time on teaching undergrads industry skills is wasteful IMO.

Profs should focus on creating next generation of researchers. Bootcamps/Corps/Youtubers should focus on next generation of developers.

That is no different than buying a house with a loan, where the loan is backed by the same house.

As a secular Indian, I highly doubt that majority, or even plurality of Indians are secular. If anything, I've seen most of my friends go from being secular to conservative moderates/hardliners in last few years. I know quite a few people who would support Muslim genocides, including my family.

When I was in school(2002-2012), the Maharashtra state board history education was Maharashtra centric, not delhi centric. Most of the history lessons revolved around the maratha empire, their enemies. Few years were dedicated to freedom struggle, world wars and ancient history.

Even then though, there was very little about even other kingdoms who ruled present day Maharashtra, so there is clear need of diversity in history education.

What I have realized after using Bard(Palm2), ChatGPT(3.5) and some other LLMs is that they are good for tasks where an accuracy <100% is acceptable, and the cost of getting wrong answers is not high.

For example, labeling a million text samples with 90% accuracy by using few shot learning is a good use case. Writing a poem is good use case. Trying to learn a new language is not. Generating a small function that you can verify might be ok. Writing entire codebase is not.

So far, I haven't found any use case for personal use of LLMs. For work however, LLMs are going to be very useful with text(and potentially image) based machine learning tasks. Any tasks where having knowledge beyond the labeled training dataset is useful is going to be a good task for LLMs. One example is detecting fraud SMS.