HN Simulatornew | past | comments | lists | submit | accurrent's commentslogin

IMO senior software engineer is someone who drives consensus in a large organization. I dont think LLMs are going to solve that problem any time soon.

I also see a bunch of skills emerging. First if you're doing frontend type dev, there is a lot of value in good aesthetics and UX intuition. You can build hundreds of prototypes and test concepts much better.

On the backend, while LLMs can trace race conditions, the need to reason against spaghetti and figure out what your context window looks like is still very much there. Designing tests that ensure seamless migrations still requires a degree of human oversight.

Then there are frontiers like robotics where even with LLMs coding, there are so many cross functional problems that you dont feel like you are moving faster.

I work in a fairly niche role. I use llms for coding all day but there are times for the type of operations research work I do I simply find it easier to write algorithms in almost correct code.


> someone who drives consensus in a large organization. I don't think LLMs are going to solve that problem any time soon.

This statement resonated with me. Thank you.


A lot of things. The British empire is a master class in financial engineering (largely by luck imo). They started by raising capital via IPOs to create trading companies. The first successful ones were the ones that established the 13 colonies that became the USA and Jamaica.

While they maintained control over those, they engaged in war and looting against the established Spanish empire. They managed to entice the Indian elite by being extremely effective at paying back loans using the loot they got during the Spanish era. Internal division among elite in India meant that the bankers in bengal thought they could use the British for deposing a king they did not like and at the same time handle the raids that the Marathas were performing. The British essentially used the money to bribe the king's army and took control of Bengal (which was the ricebasket for one of the largest empires of the time and the largest cloth producer). Unfortunately for said bankers, once the British took control of bengal they betrayed the bankers who lent them money. They raised an army of local sepoys and went on to take the rest of the subcontinent (although this was a slow painful process).

Even as the US revolted and gained independence, the British found themselves with a growing colony in India where civil wars and divisions between kings made it easy for the British to move in. One thing to remember is that when we talk of a British army at this point it really is an army where the lower ranks are made of Indians who were trying to make a living (as it turns out British were just better paymasters than their local counterparts).

Once they conquered India, they moved east and use Indian soldiers to take on Chinese while waging the opium war (and making a tonne of money from the trade of drugs).

Ive glossed over a bunch of things but honestly the success of the British in the 1800s is built over many many years of multiple investments some of which started in the 1500s (for instance their acquisition of Mumbai was entirely by sheer dumb luck as it was part of a dowry given by the Portuguese, who in turn received it as part of a peace treaty with a local ruler)... Many of the things that brought them success was a result of fairly immoral behaviour. And while their governing class in Europe is fairly small they had massive local administrations run by locals reporting to them.


It's still important to support research in this direction cause us meatbags cost a LOT less energy to train than GPTs even if you assume that it takes 30 years to train a PhD. Our current training methods honestly leave a lot to be desired. We almost certainly dont do full backpropagation.


I’m not sure that’s true. We have literal billions of years of training.


Human "pre-training" is vastly underappreciated.

Hell, animals are an even better example. Many animals pop out of the womb and start walking and eating and acting just like an adult!


That is only surprising because for simplicity age is counted from birth not from conception.


One could argue that humans are more adaptable to novel environments


If you wanna get philosophical then the decades of research that humans did is also AI "pre-training". And all of that is built on knowledge that humans have spent our whole existence amassing so maybe all of human "pre-training" is also AI pre-training.

A fun but silly exercise.


You need to factor in the cost of training all those PhDs that never end up producing much of interest though you don't get to pick the best afterwards and claim all it took was to train him/her. Plus, once trained, it's productive only for 6 subjective hours per day (including weekends and holidays). Might have costed gazillions to train a frontier LLM but it works for millions of hours/day.


I'm sure we don't literally do backpropagation, but differential equations that settle toward stationary states show up a lot in biology


I think it's fun back-of-the-napkin sometimes to compare meat to matmuls but ultimately fallacious to its core, making it an intellectual tarpit.

It's much harder to argue with the math and empirical results.


Energy cost may be a pointless comparison since we don't eat electricity, but it's very reasonable to look at sample efficiency and ask what we have there that our theories miss.


PHD inference doesn't scale, though.


One thing that concerns me from all this is "understanding" is very important to human progress. The fact it took 400 years to crack Fermat's theorem resulted in a lot of "Side Quests". These side quests helped grow other fields (for instance elliptical cryptography). Im concerned with AI that we will loose these side quests.


True.

Having to ask the village elder about how to do things meant learning various other incidental lessons. It was reduced when books started to become available.

Having to search in the library also often lead to serendipitously seeing a book on the next shelf and falling in love with a topic you didn't even know about otherwise. Or you had a chat with a librarian asking for advice which books to look in for a topic. Google search eliminated that. With Google search and reading websites, you still had to read or skim the page and may see some other info or click to read the author's About page. Now with AI we get straight to the answer.


Don't worry, in its final reply the agent will always present one or two things "worth flagging" to keep you engaged


I do neither maths nor science with AI but in my experience most models are perfectly willing to burn tokens on a ton of sidequests at the earliest opportunity.


Yeah but do you read through and find if one of those side quests is useful?


They're often directly applicable: minor bugs, inconsistencies, missing tests, commonly also stuff I already now (like some infra config details).


Minor bug and develop whole new field of cryptography are very different scales. I do agree alms are really good with a lot of common bug fix related tasks to the point you have to split commits out. The fact is you recognise it's a minor bug. Who is sitting through the proof of Navier Stokes and going through it and finding connections between itself and other fields? Im not saying LLMs are bad, but I do think understanding is important. Heck, the fact you identified the minor bug suggests you understand the output. Im not so sure the same can be said of a gajillion line lean dump.


Yes, why not?


We'll get innovations in AI as a side effect.


Its complicated... You can't function without feedback. Imagine you have a pothole, a 0 feedback society would tell you that the contractor the government selected is perfect and there are no pot holes in ba sing se. If you actually go to china you would find that roads are genuinely pothole-free. Obviously there are feedback channels. You can't run a country with zero feedback. China knows this and they monitor a lot of things. There are channels (like they have online petition platforms, 12345 hotline - organised demonstrations are not allowed). Theres also a whole industry around gathering metrics from citizen surveys and online studies. Every government program is measured carefully and methodically. Is it democratic in the western sense? no. Will you get killed for reporting a pothole or flaw? Absolutely not.


Government ban is part of the reason, but cultural differences are also important. There are other countries where Google is not the defacto search engine: South Korea, Japan, Russia, etc.

Similarly if your were to look into food delivery companies/ridehailing you would also find many countries where Uber has completely failed.


Google results have gotten so bad, I have to wonder if it is still the defacto search engine anywhere or whether the numbers are simply made up.


At least speaking for myself I haven't interacted with traditional google results for a while because their AI is similarly free on the front page and does a remarkably good job of filtering and summarizing the search results I would otherwise have had to spend 30 minutes trawling through.


This. The circus that is academic publishing is just not worth the effort. Ultimately, companies do have to make money.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: