whatyesaid

If you've got a prompt, along the lines of: given some references, check their validity. It searches against the articles and URLs provided. You return "yes", "no", and let's also add "inconclusive", for each reference. Basic LLMs can do this much instruction following, just like in 99.99% of times they don't get 829 multiplied by 291 wrong when you ask them (nowadays). You'd prompt it to back all claims solely by search/external links showing exact matches and not use its own internal knowledge.

The fake references generated in the ICLR papers were I assume due to people asking a LLM to write parts of the related work section, not verify references. In that prompt it relies a lot on internal knowledge and spends a majority of time thinking about what the relevant subareas are and cutting edge is, probably. I suppose it omits a second-pass check. In the other case, you have the task of verifying references, which is mostly basic instruction following for advanced models that have web access. I think you'd run the risks of data poisoning and model timeout more than hallucinations.

For references, as the OP said, I don't see why it isn't possible. It's something that exists and is accessible (even if paywalled) or doesn't exist. For reasoning hallucinations are different.

Yeah especially as in places like London there have been many explosions and house fires originaying from cheap foreign e-bike batteries.

Some Chinese companies care about a long-term brand and place high standards on themselves but it's not true that anything online has passed safety standards. It's hard to differentiate the two due to the amount of fake reviews also.

Rust is slow to develop and prototype in. It's good if you're a big company not the average Joe.

We need a better C++ ("systems") base language that could have an optional borrow checker or bounded model checker enabled.

Maybe if Opera, Arc, and Brave people joined forces... Without a combined effort they lack the power of Chrome store and developer docs.

Worst comes to worst they can use the Firefox engine. The MV2 extensions don't need much work to leap over for Firefox

If most companies become 4 day work week, that would be the new floor and not a perk no more?

You might still have 5 day work week employees who get paid more but to them the perk is money/earlier retirement.

People who cite interest rates and economy are coping hard. It's all of a sudden a factor now but never was before. Apart from dotcom specific crash, tech has seen relentless growth, until recent years where it was shown firing people had no negative effect.

That's nothing to do with number of engineers you have.

That service does a lot of the matching that you need and provides all roles from PM to designers to programmers. It's unquestionably better as you're dealing with volunteers. Unless you're picky with who you select and then lucky for them to have enough continued time for you.

Companies just want a resume and straight to interviews.

First registration is $2 but any renewal is $20 after. A lot of first year domains use different pricing and they are charged differently I think by the TLD holders, so it's not just stupid marketing by like GoDaddy who hopes you don't transfer to another registrar after them taking a loss the first year and huge hike after

A lot of super cheap ones have expensive renewals.

There is already socialcoder for non profits, which also provides project management and liaison. This thread would not do better than that service for non-profits, simply because it's more organized.

.io is simply the new .net, what you get when you can't get .com but don't want to rename. Although .io and .ai are really expensive so I wouldn't be surprised if something new comes up.

I have set my OS to dark mode on windows and android.

It doesn't work without extensions like DarkReader etc if you don't provide the theme/CSS as the website. I uninstalled DarkReader because it just guesses how to map things over and some websites became unreadable or in subtle ways things don't display properly.

Maybe I will use whatever is suggested here but you really shouldn't have to for a site as big as HN.

Damn, poor little Google seems really scared. I'm not sure if that info is true. You have to provide the whole source of the entire product and not just a segment of code (eg library) you forked which they do often?

It's a research paper. You can write a theoretical paper and let others apply it practically, which others can figure out the practical aspect and report results of benchmarks, or others can also build on the theory.

This paper only has 2 authors. The other solvers are probably applying technique specific tricks and speedups, and you're working with approximate optimization, it's not that easy to move everything over.

Why HN still only comes in white without these extensions is crazy. Theming is a day of work unless you use Lisp I guess. This site barely changes.

Pretty sure dark mode is much easier on the eyes at night at least.

Age is pretty easily guessable by degree dates and your appearance, they don't need to ask it explicitly.

Your first sentence is describing the web based LLMs. There are others that operate against the project structure and has read all the files. But a lot of CRUDs also involve a much easier modular domain structure.

LLMs amplify you quite a bit in productivity if you're doing CRUD which is what most people are doing. You just clean up the code and clarify potential improvements.

Only if the rate of newer services/companies keep up and outweigh the ease of development we should eventually reach. People are currently still maintaining jQuery PHP and other tech that makes doing anything much more complicated than it needs to be.

Bullshit jobs are jobs that contribute not even an economic value: https://strikemag.org/bullshit-jobs/

A good game developer provides entertainment at least. Contrasted to Twitter who under Dorsey&co shipped NFTs and 1 redesign in 10+ years. They had thousands of engineers sitting doing nothing. Imagine firing 90% of a school staff or 90% of a McDonalds staff.

Not to say that moving forward with development constantly is worthwhile either. Zuckaberg spent billions on a Metaverse which is somehow objectively much worse and buggier than VRChat.

It's going to be normalised in the coming years, across small and huge companies. Not sure how all the web devs will adjust.

Also we aren't in 2010. You don't need to build as much in house anymore or patch up tech that doesn't scale. Modern frameworks, cloud, and open source tech makes things at least 10x easier than they used to be. We will only improve in this regard.

Are there any stats on how many people are being hired?

And eBay would also fall into similar because from an engineering side, there's been no real changes for years (not that any product changes are needed).

You can have double blind papers and double blind research funding I guess to help that a little?

Not sure what else you can do. Table rankings are actually only applicable for undergrads in practical terms. At research level, You have world leading experts at random universities often, and then people who worked in top universities who advanced the field by none after decades.

I wonder how many universities actually mega climbed ranks in the past decades though

Grade inflation has been happening at literally every institution. So you can't point just to Harvard. But a middle rated school can't take a stand, only the very top schools can start to, and they'd have to do it slowly to give employers and students time to adjust.

There are apparently advanced classes like Math 55 that had huge dropout rates that had to be cleaned up though. So those courses have been completely changed instead of just awarding higher marks I'd imagine.

How do you know who they're firing? Nobody knows what "core tech" is. They are clearly getting rid of people in LoR. And game companies aren't bloated with sales people.

I also think that experimental culture is gone due to things like startup culture. Any half viable idea, they leave and create a startup.

So Google has really only been doing obvious ideas. Like Pixel phones, Pixel buds, getting into cloud too late.

Yeah but Twitch is so big it would never pay retail rates to any cloud service, even given the rates for the bulkiest users. And because it's Amazon owned it's not a question of going to AWS but how much they'd negotiate. So on paper Twitch can lose them money but Amazon makes money through AWS.

YouTube is break even and imagine how much content is uploaded and stored there per second, forever. Twitch VODs are deleted after 2 months. Most streamers don't even get any viewers. Those that get viewers get people paying real money on top of ad spam. The numbers don't really add up like how Amazon often pays $0 in tax.

How is a software written in Java having less assurances in terms of leaks/crashes etc compared to Rust? All Rust gives you is a borrow checker that isn't very smart, but means you don't need GC/VM, so you gain a bit of performance compared to Java.

But if it's not software that needs to go fast, then Rust is not adding anything for you. And if you considered Java to begin with you probably don't need top tier performance.

In that link it says they are required to pay all restaurant workers $20 an hour due to new laws. From my understanding that job gets a lot of tips in America so they don't want to cough up $20 an hour.

the idea of having in-house delivery people for any place is outdated compared to gig apps probably. You only lose a bit of money by not having your own people.