HNHacker News
TopNewBestAskShowJobs

District5524

223 karma · joined March 25, 2023

https://homoki.net/ https://scholar.google.com/citations?user=azMmMJsAAAAJ&hl=en
submissionscomments
District5524··on Part two of Grant Sanderson's video with Terry Tao on the cosmic distance ladder
A great point! I beleive today the main problem is no longer the amount of human labor needed to feed or clothe, but to have access to those universities/labs/companies that have the necessary resources to enable meaningful research that humanity values. Costs of creating food and clothe have plummeted - this is good from any possible point of view, and clearly shows we do no longer "need" inequlity from these perspectives. However, access to human institutions does not follow the same democratization and pricefall. In several fields (space, medical and material sciences), advances we yearn for require lots of capital/military power. In many fields, innovations are tied to huge amounts of money to be spent (mostly accessible only in one particular capital market of the world). But even in "cheaper" fields (like mathematics, economics etc.) simply being heard requires working in prestiguous universities. And many of those have no choice but to select and hire people based on their previous work or their financial capabilities. Thus, reinforcing the same stratification as we've seen with food and clothing, and slowing the process of finding the right bright minds to do what they excel at in greater numbers. It's quite unlikely we will solve any major part of this social problem before 2086 (or any projected population peak)... People seem to be still stuck in celebrity culture in politics and everywhere.
District5524··on America desperately needs more air traffic controllers
Could Congress support AI research and innovation by asking AI company CEOs found guilty of overpromising to prove the reliability of their latest technology by flying in AI-controlled airplanes and relying on AI-managed air traffic, instead of using private jets with human pilots and air traffic controllers? /s
District5524··on Is Iceland getting ready to join the EU?
Maybe surprising, but based on the latest Eurobarometer, the citizens of the following countries trust the EU less than Hungarians (even if Hungary is just below EU average): Germany, Czechia, Greece, Cyprus, Slovenia, France (in this order ...) While the stance of Hungary, Slovenia, Cyprus and Greece is not much relevant in relation to the future of the EU, regarding Germany and France, that's a bit more alarming... Nevertheless, the loud HU government propaganda unfortunately has its clear effect on the Hungarian public opinion - Hungarian people trust the EU less every year, and also, everything else, including democracy in general. https://europa.eu/eurobarometer/surveys/detail/3215
District5524··on Is Iceland getting ready to join the EU?
Exactly the reason you're wrong about this. If there is an official document saying that the successor to A is B, and 450 million people accept this as an "official document" and a fact, there is not much point in you saying that successor to A is rather C, because you THINK that would make more sense...
District5524··on MIT study explains why laws are written in an incomprehensible style
I agree. Building on 200 Prolific answers and inventing names for their "own hypothesis" called "magic spell"? Odd. Lawyers have written like entire libraries on this subject, there are specialized journals examining the legal language used (e.g. in English: https://link.springer.com/journal/11196, https://www.languageandlaw.eu/jll, but there are probably separate journals for this in every language with 10M+ speakers, like https://joginyelv.hu/) I understand this is not about the lawyers' approach to the problem, even if the author has a law degree, but a "cognitive sciences" department trying their hands on a problem that is new for them. But it would have been helpful if they had at least attempted to provide a reference to some prior art in the legal field...
District5524··on The tech utopia fantasy is over
I so very much agree with you... I was first baffled by how unintelligent some parts of the OP managed to come through - while it's still a very nice collection of woes and unethical lies, which is very valuable in itself. But then, I've read the linked CNN post about Jeff Bezos prohibiting presidential endorsement in Washington Posts and the outrage that has generated. No, this is surely not the author's folly - there must be a more generic problem there... Why do so many people considered intelligent think it's normal practice that newspapers give such a very direct and simple political message, like vote for X. And have been doing it for 40 years. It's normal for newspapers and editors to have clear and mostly stable political views and voice that, but I feel it very weird that they publish something as direct as this one. Just like it's very weird for celebrities (who normally work outside politics) to publicly endorse X or Y. By doing this, you lose focus on what you really want to say, and just simplify politics and communication into a binary question. Is it really needed to ELI3 everything even in politics, even to such a binary degree, and if we don't THAT is the "betrayal of responsibility"? While I'm not a great friend of using the usual left-right distinctions, especially in global settings, but that's not really important (people will see whatever enemies where they want to see, but it's still much better than naming your global enemies after a more specific group/country/religion). Technology can give us hope for new opportunities, that we can receive some "outside help" to change our social constraints. But we shouldn't blame yet another group of people for specific technology failing to solve or address some important issues of society. I wholeheartedly believe it is inevitable to convince as many people of the 8 billion to feel compassion for each other, to save the planet, to give better education and equal chances to live a full life for those who can make use of that and social care to those who need it, and, to work together for such important and clear goals... But business-as-usual will not do that, current recipes of "pluralistic democracies led by wise citizen journalists telling us what to do and think" will not get us there. And that's also a problem for LLMs: mandating guardrails will not make them more useful, and as long as LLM safety and "bias" is about how many people will be offended by their responses, then we're clearly just heading in the same wrong direction of blinding ourselves. (My condolences with you! But having a toddler is a wonderful thing, it gives you hope and a good reason to expect many joy from the future ... even if it doesn't look bright now, cheer up!)
District5524··on Woman, 82, still rides same bike she was given at 13
I am surprised that Gazelle is still such a strong brand and is made in the Netherlands. And also that, in general, bicycle manufacturing is resilient within the whole EU, despite the rising costs people rightfully complain about.
District5524··on Being Raised by the Internet
The earlier you go, the easier it is to pinpoint a single demographic (although not sure what's the point in doing that). But from '95 on, I'd say it's pretty misleading to point that to US or middle class engineers. Even in 95, it was already available to non-ranked university students in peripheric countries, and not just for those studying at engineering faculties. Although most web content at that time was related to porn (90%) and jokes (9%) and some other content (1% - probably stem and engineering?), but I don't know exactly in what philosophy that content is considered prominent and default. From 2000, it was accessible and even not too expensive for most middle class households by landline. From that time on, it was more about which demographic is not yet present... I think you had to wait e.g., another 10-12 years for mobile-first minors and for the idle retired people.
District5524··on Halide rejected from App Store because it doesn't explain why it takes photos
No. Data processor has a special meaning (at least under GDPR) and requires you to be a separate entity from the data controller (article 4(8) of GDPR). Just because you process data, will not make you a data processor. A software developer is neither a controller, nor a data processor per se, but the question is why the software processes any personal data. If there is no third person on whose behalf you process data, you might still easily be a data controller and all the privacy related obligations primarily rest on this role. Controllership does not depend on sending any data to remote locations/other apps outside the phone etc., but more on whether there is any personal data processed in the app at all and if there is, the reason for that processing. If you're not processing any personal data, you'll not be a controller, but if you do, you'll be unlikely to be able to skip this responsibility altogether. Considering you'll be the only one knowing exactly what your software can and cannot do with what kind of personal data, you'd better at least explain that to the user. Surely, developers of some software (e.g. compilers) will never be data controllers even if the software can compile software to process personal data... But for many frontend software, like apps, there is no such easy way out unless not processing any personal data at all, including usage data, which is also not an easy thing to do. But that doesn't mean you can just skip the obligation towards the store manageer who says you HAVE to provide a privacy policy even if you don't process any personal data. Similarly, processing for "purely personal or household activity", even by automated means is outside the scope of GDPR in EU, but that will not save you from the privacy policy obligation towards Apple or whomever.
District5524··on Why Don't Tech Companies Pay Their Engineers to Stay?
I think the reason is not that lawyers are irreplaceable, to the contrary perhaps. The reason is most likely that law firms are always led by lawyers, while that's rarely true for firms where most engineers work (or for most other professions). Typcially, law firms can only be owned by lawyers due to regulatory constraints (the UK, Australia being a notable, but not particularly successful exception). While this rule restricts the possible size, scalability, efficiency and profitability of law firms compared to many other sectors, it can ensure a certain degree of independence (insulation) of the management from extra-professional influences. Like "Das Kapital" trying to change the way the firm works to achieve better returns in the next 5 years. That leads to law firm management and owners being composed mostly of people who have previously worked in that exact field and PROBABLY being better at judging the worker's worth in terms of potential revenue he/she may earn, the client trust they have and the chance that they may transfer that to another firm... Maybe law firms also have less of an illusion about what is valuable in their law firm as opposed to those companies relying on engineers who have or believe they have some special moat outside their people (technology, source code, inventory etc.). No professional law firm would have any illusions that their lawyers are special - perhaps outside trial lawyers and some very special field of litigation, the firm just needs burnout resistant cannon-fodder with the capacity to run for at least ten years, so they may acquire the sufficient experience, plus a number of bland human skills and patience in coping with other human beings. There never was a genuine myth of the "10x lawyer", that's something that only ChatGPT invents. (Sidenote: I wouldn't call an engineering job blue collar as long as you work in an office - WFH, we are all collarless workers now.)
District5524··on AI worse than humans in every way at summarising information, trial finds
Summarization is one of those key functionalities of LLMs that laypeople can also easily understand and relate to. I think this article also underpins the hunch that results in ROUGE and similar NLP benchmarks are not necessarily a guarantee for good performance in sector-specific summarization tasks, where the expectations are defined by human users for a specific domain. It reminds me of the Stanford study on the use of LLMs for legal QA including the eye-wateringly expensive legal-specific LLMs: https://hai.stanford.edu/news/ai-trial-legal-models-hallucin...
District5524··on Lidl's Cloud Gambit: Europe's Shift to Sovereign Computing
Maybe you are misunderstanding the gravity of this problem. Thanks to US Cloud Act and the Patriot Act and similar acts, there is no way any US citizen or any US company may EVER be involved in such projects. It's completely legal for the US to rely on extraterritorial jurisdiction leveraging any US companies and US citizens they have access to. But on the other hand, everyone else on out there will want to avoid that, so the only way to achieve that is to avoid involving any US citizens or companies for such sovereign projects. Google will not be able to solve it via subsidiaries, and no nice promises from Amazon, MS etc. will ever change it. Data sovereignity means all this. This will probably escalate a lot more, it might involve the financial infrastructure used (SWIFT) or even currency used in the process.
District5524··on Artificial intelligence is losing hype
I suggest we do not repeat the myth and urban legend that LLMs are good for legal document review. I had a couple of real use cases used for real clients who were hyped about LLMs to be used for document review and trying to save salary, for Engish language documents. We've found Kira, Luminance and similar due diligence project management stuff as useful being a timesaver if done right. But not LLMs. Due to longer context windows, it is possible to ask LLMs the usual hazy questions that people ask in a due diligence review (many of which can be answered dozens of different ways by human lawyers). Is there a most favoured nation provision in the contract, is there a financial cap limiting the liability of the seller or the buyer, governing law etc. Considering risks of uploading such documents into ChatGPT, you are stuck with Copilot M365 etc. or some outrageously expensive "legal specific" LLMs that I cannot test. Just to be curious with Copilot I've asked five rather simple questions for three different agreements (where we had the golden answer), and the results were quite unequal, but mostly useless - in one contract, it incorrectly reported for all questions that these cannot be answered based on the contract (while the answers were clearly included in the document), in an another, two questions were answered correctly, two questions not answered precisely (just governing law being US instead of the correct answer being Michigan, even after reprompting to give the state level answer, not "USA") and hallucinated one answer incorrectly. In the third one, three answeres were hallucinated incorrectly, answered one correctly and one provision was not found. Of course, it's better to have a LEGAL specific benchmark for this, but 75% hallucination in complex questions is not something that helps your workflow (https://hai.stanford.edu/news/hallucinating-law-legal-mistak...) I don't recommend at least LLMs to anyone for legal document reviews, even for the English language.
District5524··on FlightAware Leaks Customer Data (Name, Email Addresses and Passwords)
That's not how it works. I think you are mixing up damages to be awarded by courts to an individual data subject (right to compensation, Article 82) with penalties (Article 83-84) - the latter having a special meaning, in practice covering the administrative fines by authorities (DPAs). There was a case by the ECJ (C 300/21) saying that for damages to be awarded to the individual, they have to prove the material or non-material damages involved. Regardless, any data protection authority can and do fine companies for breaches such as this one and also for late filing, and while DPAs also have to take into account the damages caused to data subjects when deciding the amount of fine, that's not a standard that has to be proven for each data subject separately and is therefore not as strict as the right to compensation. FlightAware definitely should be fined... Like Booking, NTT or Twitter or banks for breach obligations, for more than €400k. But it's still strange that very few US companies were fined so far in relation to no or late breach notifications - not that they were not fined heavily for failure to comply with other obligations. https://www.enforcementtracker.com/
District5524··on Vigorous Exercise, Cognitive Decline, and High Blood Pressure
"The greatest among you will be your servant. For those who exalt themselves will be humbled, and those who humble themselves will be exalted." (Obviously, I'm neither Jesus nor the OP).
District5524··on Japanese words and names sound African (2022)
Speaking of similar sound system of Finnish and Japenese, I think we should highlight the discredited language family claim of Uralic-Altaic, see https://en.wikipedia.org/wiki/Ural-Altaic_languages You can still find tons of other theories of "what else could be related to Uralic"... So far, no African languages were included in these theories. But that's probably just a question of time and politics (and definitely not linguistic research...)
District5524··on Frog or Toad?
I see, thanks for the answer!
District5524··on Frog or Toad?
Thanks for the game! It is a very entertaining and is highly educational, probably thanks to the pictures ... And a good way to show us fascinating species (like the purple frog) that we would otherwise not bother to look up. Except for the most abundant species available in Europe, it seems quite unlikely that people would be willing to give these species "common names" in English. This does not decrease the value of the game at all, but it is perhaps a tad misleading what the objective is: are we to guess a) whether a particular specie is named as a "frog" or "toad" in its common name (folk taxonomy) or b) where they belong in taxonomy (is it a Bufonidae or not)? If we go for a), that is perhaps not as exciting to show species endemic to places without tons of native English speakers (like Thailand etc.). But this does not decrease from the value of the game at all. I think it shows a very efficient approach to gamify and teach taxonomy to people.

Would it be of any use to show a success rate indicator perhaps or a feedback whether the guess was correct or not?

District5524··on Slack AI Training with Customer Data
The Universal License Terms of Microsoft (applicable to Teams as well) clearly say they don't use customer data (Input) for training: https://www.microsoft.com/licensing/terms/product/ForallOnli... Whether someone believes it or not, is another question, but at least they tell you what you want to hear.
District5524··on Ancient Eastern European mega-sites: a social levelling concept?
I think it's pretty far-fetched to draw conclusions from cultures that were at the same place like 4500 years later. How do you think they could have kept like 10000 slaves in place and convince them to remain slaves? And other Cucuteni sites had population estimates around 45 000 people...
District5524··on Anduril Selected for U.S. Air Force Collaborative Combat Aircraft Program
Copyright protection of characters is not harmonised at the international level (https://www.wipo.int/wipolex/en/text/283698), and is a very contentious area even within specific jurisdictions. Like, in the US, there are some tests, like "character delineation" and "story being told" test for protection to exist: how well developed that character is (Nichols v. Universal Pictures Corp.), so cannonfodders are not protected, main characters could be. In other countries, that protection is much stronger, like any uniquely, identifiable character (or even titles of literary works) practically receive such a protection (like in Hungary). But there is always a second layer of defense, like it's fair use, pastiche etc. Like in Germany, the German supreme court has decided that selling costumes of Pippi Longstockings is not within fair use (even if they do not call it "Pippi costume", of course). So, good luck for anyone choosing such names for their products to sort this out in litigation, probably not worth the risk here. It often just takes time for the copyright holders to notice and to react...
District5524··on Serious security breach hits EU police agency
GDPR article 2.2-2.3 does define some exceptions. National law enforcement agencies etc. are subject to a similar directive (LED, https://eur-lex.europa.eu/eli/dir/2016/680/oj). Union bodies, just like in your link yet another (article 2.3), but also quite similar (Regulation (EU) 2018/1725, https://eur-lex.europa.eu/eli/reg/2018/1725/oj) Eurojust belongs to this one. It's a good thing to have EDPS, otherwise it would be much easier to legislate things nobody can comply with in practice. But for most countries, it doesn't make much sense to fine other authorities. (Unfortunately, national security services are not subject to GDPR or LED at all - the only place you can get protection from them is under national law, Charter of the Fundamental Rights, ECHR etc.)
District5524··on Better Call GPT: Comparing large language models against lawyers [pdf]
Could be paralegals as well, but they are usually just human employees of legal service providers that are not licensed lawyers/law firms. It's jurisdiction specific who can do this besides lawyers. In the UK and Australia (and probably NZ as well) they can be practically be anyone, contract review is not a reserved legal activity like in most US states.
District5524··on Better Call GPT: Comparing large language models against lawyers [pdf]
Your reading is correct. Their primary aim was to compare some real lawyers' work with their wonderprompts to show off an upcoming service and how faster that is compared to humans. But they didn't have the funds to ask lawyers to review 15 or 50 or 1000 contracts, just 10, hence the cheat in methods and in conclusions.

This is not about drawing up a benchmark for an industry, let alone validating any useful method in general.

Ten docs reviewed based a single review 'playbook' (I think it's not in the paper, but probably max 20 questions per contract?) and compared across 3 different providers/roles + LLMs...

District5524··on Better Call GPT: Comparing large language models against lawyers [pdf]
You are mixing up important things and probably that's why you don't see a bigger threat here than the livelihood of a couple of lawyers.

People who attended law school/university/have a law degree are not necessarily practising law as lawyers/attorney or whatever they call them in the given country, as a member of the government. But of course, once a marine, always a marine...

On the other hand, the parent was trying to refer to criminal defense (Amendment 6 of US Constitution). That's not something active politicians can participate in (conflict of interest, etc.).

If one has the means to pay for their criminal defense (or family law, anything that truly matters for the average citizen), they tend to be willing to pay a lot and expect some quality of service in exchange. They will certainly not choose ChatGPT, just because that provides the answer faster for long documents.

But a large number of people don't have this kind of money. As long as governments don't have a proper budget for legal aid, they will be inclined to cover the vanishing legal aid budget with false claims that they fulfill "appointment of counsel" with LLM services - while they don't really care about the outcome of the case. (And of course, successful politicians have rarely worked as pro bono counsels before their current job, not that it has any more relevance here.)

Non-Western governments are also interested in having this replacement service - the cost of lawyers is also money there. It's much better for all governments to have chatbots they have reliably trained than less controllable humans.

That's a hypothetical threat for now, but I hope you understand why there is something more here beyond "guildthink" and protecting the livelihood of lawyers at all costs behind the Sixth Amendment. There are lots of countries (or even within the US, like in Utah) where you can review contracts without being a practising lawyer, like LPOs mentioned in the paper. So far, it failed to make the world a better place.

But replacing including LLMs as "counsel" under Sixth Amendment would mean a very different situation.

District5524··on Better Call GPT: Comparing large language models against lawyers [pdf]
The third one (CUAD) is a single paper, not blogposts like the others. I think this paper is still the best in terms of being done by NLP experts and understanding the possibilities and not being just some and mirror. But there are so many papers published in this area nowadays that I might not even notice a new one. The CUAD paper was still based on BERT, so pretraining was needed - that needs a bit more expertise than just prompting GPT-4-32k like in this paper or feeding prompts back to GPT-4 for another round of refining, or doing RAG. For honest research purposes, "contract review" is not really a good area of approach: the subject field is not standardised, there is no good benchmark yet and your paper can easily get into bad company of snake oil sellers cashing on visceral hatred of average people for all professions.
District5524··on Better Call GPT: Comparing large language models against lawyers [pdf]
While this paper is clearly not without merits, it intends to be more like an excuse to make a bombastic statements about a whole profession or "industry" (perhaps to raise their visibility and try to sell something later on?). The worst part is that they have actually referenced a single preprint document as "previous art" - and that document itself is not related to contract review, but to legal reasoning in general. (A part of LegalBench is of course "interpretation", and that is built on existing contract review benchmarks, but they could've found more relevant papers). Automating legal document review has been a very active field in NLP for twenty years or so (including in QA tasks) and became a lot more active since 2017. At least e.g. Kira (and Luminance etc., none of which is LLM-based) are already used quite widely in legal departments/firms around the world. So lawyers do have practical experience in their limitations... But Kira & co. are not measuring the performance of the latest and greatest models and they do not use transparent benchmarks etc. So the benchmark results in this paper are indeed a welcome addition in terms of using LLMs. But also considering its limited scope of reviewing 10 (!) documents based on a single review playbook, they should not have written about "implications for the industry". It is very much pretentious and shows more of the lack of knowledge of the authors of the very same industry than about the future of the legal services industry.

If you're interested in the capabilities and limitations, I suggest these informative, but still light reads as well: https://kirasystems.com/science/ https://zuva.ai/blog/ https://www.atticusprojectai.org/cuad

District5524··on Better Call GPT: Comparing large language models against lawyers [pdf]
All the governments in the world would do everything in their power to get people accept this suggestion as a truth and use LLMs instead of human lawyers especially in criminal defense. Now, why is that? Maybe it's not the technical knowledge that is the most important feature of a lawyer?
← PreviousPage 3 of 3