Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Role of AI in Accelerating Skill Development

saulcosta.com

Post reply on HN
Print view View original

21–30 of 53 posts

Re: The Role of AI in Accelerating Skill Development

#21

Post by capableweb »

waboremo wrote: post #19
capableweb wrote: post #7

Earlier quoted context omitted.

> but you need to actually know something about the topic to evaluate ChatGPTs ability. I don't necessarily think you have to, but you have to be ready to look up further, more trustworthy sources after your discussion with ChatGPT.

People already don't do that with our current tools, what makes anyone believe it will happen with ChatGPT? Especially when they now have to go back to their favorite search engine to find those sources and validate them. It's as if everyone has forgotten how humans behave, they don't even read or go to actual wikipedia sources that are properly laid out! Nevermind having to go find trustworthy sources.

I don't really care about how others use it. I myself find it enough useful and productive to use it, and what I wrote before is how I "protect" myself. As you say, people who don't check sources won't do so with or without GPT4.

Re: The Role of AI in Accelerating Skill Development

#22

Post by itissid »

Most of the times I have found chatgpt useful is to explain something I am not fully clear on, but have some idea about. Random example:

I read from this[1] source that CLO transactions(and hence cashflows) are "arbitrage". And I wanted to understand what is the exact arbitrage.

------------------------------ Prompt: Do you know what is Arbitrage CLOs in this context?

ChatGPT: Yes, I can explain what Arbitrage CLOs are in the context of Collateralized Loan Obligations (CLOs).

Arbitrage CLOs are a type of CLO that is structured to take advantage of differences in the yields of the loans included in the CLO's underlying asset pool. The goal of an Arbitrage CLO is to generate a higher yield than the cost of financing the CLO by purchasing the underlying loans at a lower cost and then securitizing them into higher-yielding CLO securities.

In an Arbitrage CLO, the issuer will purchase a pool of loans at a discount and then securitize them into several tranches of securities with different levels of risk and return. The securities are then sold to investors at a higher price than the cost of the underlying loans, generating a profit for the issuer.

The difference between the yield on the underlying loans and the yield on the CLO securities is known as the "spread," and it represents the profit that the issuer earns from the CLO. In an Arbitrage CLO, the spread is typically larger for the more junior tranches of securities, which offer higher yields but also carry higher risk.

Arbitrage CLOs can be an attractive investment for investors who are looking for high-yield, high-risk investments. However, they can also be volatile and carry significant risk, particularly in periods of economic stress or market volatility. As with any investment, it is important to carefully consider the risks and potential rewards before investing in an Arbitrage CLO.

-----------------------------------------

So, I learn't(or confirmed my intuition) that the arbitrage occurs because of discount on loans.

[1] https://content.naic.org/sites/default/files/capital-markets...

Re: The Role of AI in Accelerating Skill Development

#23

Post by nice_byte »

yieldcrv wrote: post #10
JackFr wrote: post #2

This is terrible. The idea of ChatGPT being able to fill in gaps of not knowing what you don’t know is appealing, but you need to actually know something about the topic to evaluate ChatGPTs ability. This is just spewing pure nonsense, and the author thinks he has learned something from it.

this morning I needed to populate an SQL database with dummy values, only a few dozen though so I told ChatGPT the model with types and asked it for random values in csv format. I’m sure the API could spit out far more if I used that I imported into the database and moved on. 30 seconds? mostly waiting for it to output maybe we should make a site where people share what they used LLMs for

but... this has nothing to do with what parent is talking about...

Re: The Role of AI in Accelerating Skill Development

#24

Post by non- »

qingdao99 wrote: post #20
Jeff_Brown wrote: post #11

Earlier quoted context omitted.

Just a quibble, but "hundreds of millions of people" seems high by an order of magnitude or two. That would be around 1 in every 20 people in the world. But this is known. Has OpenAI released usage figures?

> ChatGPT has more than 100 million users within the first two months of its launch and has more than 13 million daily visitors as of 2023. https://www.reuters.com/technology/chatgpt-sets-record-faste...

Wow that's way more than I expected even being aware that they set new records for fastest to 5 million.

I was kind of hoping that we were still early on the trend, despite the massive popularity within tech.

I guess we are, assuming this is going to grow to Billion+ territory.

Re: The Role of AI in Accelerating Skill Development

#25

Post by non- »

capableweb wrote: post #21
waboremo wrote: post #19

Earlier quoted context omitted.

People already don't do that with our current tools, what makes anyone believe it will happen with ChatGPT? Especially when they now have to go back to their favorite search engine to find those sources and validate them. It's as if everyone has forgotten how humans behave, they don't even read or go to actual wikipedia sources that are properly laid out! Nevermind having to go find trustworthy sources.

I don't really care about how others use it. I myself find it enough useful and productive to use it, and what I wrote before is how I "protect" myself. As you say, people who don't check sources won't do so with or without GPT4.

It's useful to learn new tips and tricks. e.g. the GPT4 demo showed me how to get much better code output.

Re: The Role of AI in Accelerating Skill Development

#26

Post by Jeff_Brown »

qingdao99 wrote: post #20
Jeff_Brown wrote: post #11

Earlier quoted context omitted.

Just a quibble, but "hundreds of millions of people" seems high by an order of magnitude or two. That would be around 1 in every 20 people in the world. But this is known. Has OpenAI released usage figures?

> ChatGPT has more than 100 million users within the first two months of its launch and has more than 13 million daily visitors as of 2023. https://www.reuters.com/technology/chatgpt-sets-record-faste...

Wow. 13 million per day = 1 in every 400 people on Earth with internet access, if Google is correct that world population is now 8e9 and 64.4% of the world has internet access.

Re: The Role of AI in Accelerating Skill Development

#27

Post by photochemsyn »

One thing ChatGPT probably won't tell you about macroeconomic theory unless specifically asked about it is that a lot of the foundational assumptions are highly idealized and aren't that useful for making real-world predictions:

https://www.scientificamerican.com/article/the-economist-has...

> "But what is not widely known is that these now legendary economists—William Stanley Jevons, Léon Walras, Maria Edgeworth and Vilfredo Pareto—developed their theories by adapting equations from 19th-century physics that eventually became obsolete."

I'd probably trust ChatGPT more on elucidating the history of the development of macroeconomic theory then on applications of the theory to real-world problems, but that goes for the working professionals in the field as well, with all their rosy predictions of the benefits of NAFTA to the working public in the 1990s, or their predictions of the benefits of financial deregulation in the early 2000s prior to the 2008 Recession, and so on.

E.g. ChatGPT claims American academic institutions adopted macroeconomic theory with Milton Friedman at Chicago and Alvin Hansen at Harvard, and this was preceded by development in the 1920s at the National Bureau of Economic Research and in the 1930s with the Cowles Commission. If you ask it what role the above four individuals in the quote played in the development of macroeconomic theory, it gives illuminating answers on utility, general equilibrium, indifference curves, and Pareto efficiency (which, ChatGPT claims strongly, is not zero-sum game thinking). This all seems fairly accurate based on other sources, except that the relationship between game theory and Pareto efficiency seems fairly contentious.

Also, here's a good question to ask it in the context of the desirability of optimizing Pareto efficiency:

"Has macroeconomic theory ever been criticized on teleological grounds?"

Generally, this is probably not the best subject to give to ChatGPT as a quantitative test. Maybe try it out with something more concrete, like how to use gravitational assist (aka 'slingshot') to send a spacecraft to Jupiter or Saturn, and see how explicit it can get in terms of a launch trajectory?

Re: The Role of AI in Accelerating Skill Development

#28

Post by rqtwteye »

onemoresoop wrote: post #4

Tired of Crypto hype? Well, it passed. Now they hype is big on AI. Useful things will surely come out of it but before lots of money will be made on hype alone.

It's hype for sure but for AI I can think of a ton of real-world use cases whereas all crypto use cases sounded pretty contrived. I think AI will be a revolution of the scale of the internet or even bigger. Maybe the current version of ChatGPT is flawed but the possibilities are there. I am thinking analyzing long contracts, writing letters, verifying system requirements and tests and so on. The potential seems endless.

Re: The Role of AI in Accelerating Skill Development

#29

Post by sosodev »

worrycue wrote: post #17
sosodev wrote: post #9

Earlier quoted context omitted.

Do you think that other sources are infallible? Almost every source is wrong sometimes. ChatGPT is as good as my professors were in college. They would regularly say things that were just completely incorrect but were knowledgeable and correct often enough to be helpful.

Your professors generally don’t make stuff up if they don’t know something. Your professors are also more likely to catch on if something doesn’t make sense - unlike ChatGPT which can’t seem to reason effectively because most of the time it doesn’t know what’s it’s talking about.* * Maybe it has the ability to reason, maybe not. Maybe it just needs more information - after all it’s trying to model reality with just t…

Some of them most certainly did make stuff up. Ego prevents many people from admitting that they don't know something.

I also interacted with some who wouldn't even admit to being wrong when they were confronted with facts. So I don't know that I would say they catch on either.

Re: The Role of AI in Accelerating Skill Development

#30

Post by waboremo »

capableweb wrote: post #21
waboremo wrote: post #19

Earlier quoted context omitted.

People already don't do that with our current tools, what makes anyone believe it will happen with ChatGPT? Especially when they now have to go back to their favorite search engine to find those sources and validate them. It's as if everyone has forgotten how humans behave, they don't even read or go to actual wikipedia sources that are properly laid out! Nevermind having to go find trustworthy sources.

I don't really care about how others use it. I myself find it enough useful and productive to use it, and what I wrote before is how I "protect" myself. As you say, people who don't check sources won't do so with or without GPT4.

Maybe you'll eventually care about how others use it when your trustworthy sources are flooded with irrelevant, outdated, and downright false information due to the constant generation of "kind of true" content.
Post reply on HN