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 Lessons from interviews on deploying AI Agents in production
Not that it's incorrect but there is some data showing variability even with the very same input and all parameters. Especially if we have no control over the model behind the API with engineering optimizations etc. See Berk Atil et al.: Non-Determinism of "Deterministic" LLM Settings, https://arxiv.org/abs/2408.04667v5
District5524··on 1X Neo – Home Robot - Pre Order
Plenty of opportunity to use forced labourers in a DIFFERENT country while complying with all the immigration laws possible, and also saving the owners from having to meet real poor people. (I hope this will not work well...)
District5524··on Who's Submitting AI-Tainted Filings in Court?
That's an important observation. It's not easy to get filing data outside the US Federal courts (PACER), because it's not typical at all that courts publish the filings themselves or information on those who file the pleadings. But you can find statistics of the legal market (mainly law firms), like class size (0-10, 10-50, ... 250+ lawyers per firm) of total number of law firms, number of employees per class size of total law firm employees, or revenue per class size. Large firms only dominate the UK, especially in terms of revenue, US is less so, EU is absolutely ruled by solo and small firms. I did some research back in 2019 on this, updates, the figures probably did not change, see page 59-60: https://ai4lawyers.eu/wp-content/uploads/2022/03/Overview-of.... The revenue size statistics was not included in the final publication. You can fish similar data from the SBS dataset of Eurostat https://ec.europa.eu/eurostat/web/main/data/database (But the statistical details are pretty difficult to compare with the US or Canada, using different methodologies, different terminologies.)
District5524··on I spent the day teaching seniors how to use an iPhone
I used to feel the same, but nowadays, I have very few normal calls, mostly from close relatives only, but a dozen of spam calls a week. And that used not to be the case before 2024. I know robocalls are/were a terrible affliction in the US for like the last 10 years, but they were very few of these in Hungary. I wish I'd know what caused this change..
District5524··on EU court rules nuclear energy is clean energy
This is similar to the reasoning of Austria vehemently opposing nuclear reactors to be built in neighbouring countries, even if downstream on the Danube, even if 200 km from their border.

The latest decision (although on the surface, not on an environmental issue like the article is about, but on state aid measures - but actually not the real reason for Austria's opposition): https://eur-lex.europa.eu/legal-content/en/TXT/?uri=CELEX:62...

So, I believe, yes, low HDI countries with high corruption do have the right to build nuclear facilities. This is not like a combination of low HDI and high corruption index awarded by some international organization has the approval rights to such questions of sovereignity. There is a whole range of special regulation regarding who can build nuclear stations and under what conditions, with a special agency to ensure the safe use (IAEA) - that should be the only criteria for letting nations build nuclear stations, not corruption, HDI or how rich the countries are.

District5524··on A ChatGPT Pro subscription costs 38.6 months of income in low-income countries
I believe it's more about witnessing a deepening AI-divide between rich and poor countries. Or the maturing of the AI market. But in the end, this will only matter if we see these more expensive tools/subscriptions actually having a market advantage that cannot be mitigated by low income countries. (Certainly no new startup will be able to train new models in low income countries, hire world class AI professionals etc.)
District5524··on A ChatGPT Pro subscription costs 38.6 months of income in low-income countries
That's for international students going to the UK to study CS. There is not much point for anyone to go to the UK to study CS for that amount of money (unless they already live there, but they get their degree for a third or even less of that money). It's normal for international student tuition fees to be inflated by many universities, they try to collect some extra revenue based on a perceived extra prestige, especially the US and UK. Similar to charging different prices for tourists than for locals.
District5524··on AI Ethics is being narrowed on purpose, like privacy was
That reminds me of the new draft standard of CEN/CENELEC (EU std body) on "Competence requirements for professional AI ethicists" https://standards.cencenelec.eu/dyn/www/f?p=205:22:0::::FSP_...

But by the time they'll adopt it, singularity will already have happened... For some reason, my instincts suggests there will be no MA in Philosophy needed.

District5524··on Intel CEO Letter to Employees
No, because jurisdiction of the competition authority will still be based on the market activity (or place of registration etc.) of the companies colluding, not on the physical location of the colluding/anti-competitive activity itself. While such conspirative theatre may decrease the risks of discovery, collusions usually come to light by one participant notifying the authority in exchange for immunity or decreased fines.
District5524··on AI Market Clarity
that would be fun, but what i've heard that it's named after Harvey Specter, from Suits. But that's not something they'll officially ever deny or confirm...
District5524··on AI Market Clarity
Although there is a part on the use of AI in the legal market, but it is neither comprehensive, nor informative. A much better view of the law-specific state of the art models is available at https://www.vals.ai/vlair (although registration is needed to access the report). That benchmark in the report clearly shows that there is not much added value or moat in the custom training of Harvey (costing supposedly ~5M$), while a Llama-based model can achieve quite similar performance. Of course, no benchmark is perfect but it is still more informative than this blogpost, plus all the linked websites of the "legal" products linked that fail to give you the slightest idea of how they work. Besides claiming to be be disruptive...
District5524··on I know genomes and I didn’t delete my data from 23andMe
The exact same thing happened in Hungary (sort of, we can't blame it all on the nazis, Hungarians did it enthusiastically themselves). They used the census data of 1920 and 1930 (but not declaring your religion and ethnicity was illegal) in the numerus clausus acts and then in the mass killings and holocaust (600k of 850k). But after the 2nd ww, they used the same census data against the German minority as well, to evict them and move them en masse to Germany.
District5524··on Felix Baumgartner, who jumped from stratosphere, dies in Italy
Communism in Hungary was introduced and kept up by Soviets after the 2nd ww. The very considerable difference between Orban and Kadar (the longest serving communist leader) is that Kadar paid attention not to make his direct family and direct friend the most exorbitantly rich guys in 15 years, which probably also made lots of mid-income people poorer in Hungary. And despite very strong democratic backsliding in many areas, it is still more democratic than during Soviet-led times. But that's probably not due to Mr Orban's character or self-restraint. Baumgartner's idea is probably related to fashionable libertarian ideas of those rich people who never had the patience to feel sympathy for other people, nor to study history or humanities, but feel like they should have a say.
District5524··on Doom Didn't Kill the Amiga (2024)
I'm sure that was an important reason for the US and UK markets. I don't know much about the Swedish market, but in the Eastern European markets, game consoles were not popular at all until the very late 1990s. Gamers in the early 1990s used Commodore (and less affluents ZX Spectrums and clones), and later, as the article says, more and more PCs. In these parts, people were very short of money, which caused piracy to grow HUGE. Computers were also preferred due to being multipurpose. And console games were difficult to copy, so if you wanted to be a serious gamer, you had to have a PC. Even if you had the money to buy games, you couldn't really do that, because nobody really was selling actual games in these markets. With different currencies and limitations on hard currency exchange, you didn't spend those limited funds on games, even if you travelled abroad. Probably also as a consequence, playing on consoles was considered to be lame. It was for kids only. Hah, for those born in the 1990s only! So that was a bit different over here.
District5524··on Fei-Fei Li: Spatial intelligence is the next frontier in AI [video]
An immaterial side note: funny how obsessed she seems to be with her age. She said once that people in the audience could be half or even third of her age. Given that she's 49, is it really typical that 16-year olds attend these fireside YC chats?
District5524··on ChatGPT's enterprise success against Copilot fuels OpenAI/Microsoft rivalry
The pipeline Microsoft is using for Copilot products hides actually what models they are using, and you have no influence over it. Sometimes they use smaller models, but I have no clear source from Microsoft saying this ...
District5524··on AI Saved My Company from a 2-Year Litigation Nightmare
It's not a problem that you try multiple sources at all. That's a smart move as long as you do realize what you are doing and give these extra sources appropriate weight. But that's very difficult to do with LLMs due to a) pointy-haired bosses and the capital very much wanting to believe they can do away without third-party expertise and the unpleasent dependency on humans in general, and b) the market hype around technical tools that technical people are much more prone to (excellent, it works well in my field, it should also work in other fields, we're smarter then the others etc.). The difficult part is that if many people go with bad contracts and bad legal advice, that will be visible only after many years of litigation, and in general, by rising legal costs in courts. And that also means a lot more people will have to swallow agressive counterparty propositions from the better financed side, bad default judgements, and in general, a lot more third-party financed litigation (which will not benefit lawyers in the end, but most lawyers do not care). That will also be seen just as a further rise in legal costs... Which will probably be promised to be solved by less human judges and more LLMs, run by people interested courts spending more on LLMs. LLMs do not work in the way that humans do, benchmarks are very primitive and unreliable, even the best tools marketed for lawyers are unreliable on real life scenarios, while evaluations costs a lot, so very few people do them.
District5524··on AI Saved My Company from a 2-Year Litigation Nightmare
I think the major point is that if you fancy yourself a beginner businessperson, sooner or later you will have to find some professionals you can trust. Regardless whether that's a lawyer or an accountant. Doing business will have its costs, and you will not necessarily pay those costs in the way you can anticipate.

Maybe you pay those costs before you sign a contract, and avoid a jurisdiction in a contract that you have no connection to (if you can). But you will definitely pay these costs when you have to litigate. If you think Claude or ChatGPT is the best answer to this, and all you need to know is using advanced voice and upload documents, I believe that's not that useful piece of advice.

Please bear in mind that it's not just particular case law that these tools hallucinate. They give incorrect advice in very different ways how humans do. If you can find a lawyer you can trust in a jurisdiction you're interested in, and you got along with him/her quite well, you can trust them for as long as they're in business, even if you never pay them once again. And they might even give you referrals or point to other professionals, understanding your budget and situation. Of course, they can still give you bad advice, they can become gambling-addicted, and run away with your money. It happens, even if very rarely.

But with LLMs, you won't know when they hallucinate, and what they get wrong. Even if you have the full statutes and case law uploaded and updated, it will give you wrong answers as well, but they're not working in the same way as humans do.

All we know that it's more likely to give you incorrect advice in jurisdictions not being covered by their training data (like outside New York or Delaware), and it will draw incorrect inferences in other cases from materials it has seen. Usually only lazy lawyers not reading the output will face the problem of incorrect case law.

But legal advice is not like coding where you can filter out most of the incorrect answers with tests and compilers. The author have happened to chose the wrong lawyers for the first time. I accept that not everybody has reliable lawyer friends qualified in Delaware, but in terms of advice, it would have been more practical to discuss the possible directories to use, beginner's methods to select lawyers, warning signs to look out for etc. It's sad that many lawyers act for trolls and do bad faith litigation. It's even worse that bars have no methods to exclude and disbar such guys. But you cannot avoid these problems by trusting a cheap AI tool instead, that's just wishful thinking.

District5524··on The UI future is colourful and dimensional
We ARE a loop ourselves: https://en.wikipedia.org/wiki/I_Am_a_Strange_Loop
District5524··on Widespread power outage in Spain and Portugal
This is a bit more than just red tape madness, it's a strategy to make businesses more transparent. This is about trying to reduce non-reported transactions and too many people dodging their reporting. Even if the rules for cash registries and reporting are detailed, a) that's not really expensive for businesses - it's easy to automate and there are quite a number of competitors; b) compared to accounting and tax rules, they are dead simple.

Receipts or invoices are the basis for a firm's whole economic activity, including the underpinning of their financial reporting, their tax burdens etc. And businesses failing to provide receipts erodes not only the tax base, but also any rights a consumer may have.

District5524··on The Gruen Transfer is consuming the internet
I understand your concern, but that's not how the law prohibiting it works. That just says that unfair commercial practices are prohibited, and it is unfair to use aggressive practice, such "any onerous or disproportionate non-contractual barriers imposed by the trader where a consumer wishes to exercise rights under the contract, including rights to terminate a contract or to switch to another product or another trader". https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A...

So, requiring an ID check for termination, for no other reason than to make it more difficult than necessary, would still fall under this prohibition.

District5524··on No Pay, No Work; Early Career Lessons
I hate to say this, but this is quite common in Italy for smaller businesses (and most businesses in Italy are small). They have become accustomed to their courts being slow so that this is a negotiation tactic you often hear. A shame on such a nice country...
District5524··on Visual Reasoning Is Coming Soon
The first caption of the cat picture may be a bit misleading for those who are not sure of how this works: "The best a traditional LLM can do when asked to give it a detective hat and monocle." The role of the traditional LLM in creating a picture is quite minimal (if there is any LLMs used), it might just tweak a bit the prompt for the diffusion model. It was definitely not the LLM that created the picture: https://platform.openai.com/docs/guides/image-generation 4o image generation is surely a bit different, but I don't really have that kind of more precise technical information (there must be indeed a specialized transformer model used, linking tokens to pixels, https://openai.com/index/introducing-4o-image-generation/)
District5524··on AI coding mandates are driving developers to the brink
Why do other people's criticism ("talking shit") destroy your fun? They make you unsure of your opinion? If you believe this emotional investment in AI tools gives you a competitive edge in the long run, then so much the better for you. AI tools should work the same way with or without your love. The only difference is that if you have fun using these tools, you'll be better motivated to use these tools better then the others. (Regardless, existing knowledge probably helps more in using these tools effectively in the longer run than pure fun.)
District5524··on An image of an archeologist adventurer who wears a hat and uses a bullwhip
I asked Sora to turn a random image of my friend and myself into Italian plumbers. Nothing more, just the two words "Italian plumbers". The created picture was not shown to me because it was in violation of OpenAI's content policy. I asked then just to turn the guys on the picture into plumbers, but I asked this in the Italian language. Without me asking for it, Sora put me in an overall and gave me a baseball cap, and my friend another baseball cap. If I asked Sora to put mustache on us, one of us received a red shirt as well, without being asked to. Starting with the same pic, if I asked to put one letter on the baseball caps each - guess, the letters chosen were M and L. These extra guardrails are not really useful with such a strong, built-in bias towards copyright infringement of these image creation tools. Should it mean that with time, Dutch pictures will have to include tulips, Italian plumbers will have to have a uniform with baseball caps with L and M, etc. just not to confuse AI tools?
District5524··on US Administration announces 34% tariffs on China, 20% on EU
That's a bit misleading. Make no mistake, current US tariffs are UNIVERSAL. Not specific to e.g. chicken or automobiles or purebred horses. They cover absolutely everything. While you may see 38% or more tariff on chocolate, chocolate export to the EU from the US is hardly an important issue. And for dairy products, the tariff is a fixed euro amount per weight, not %.

https://ec.europa.eu/commission/presscorner/detail/en/qanda_...

https://trade.ec.europa.eu/access-to-markets/en/search?origi...

District5524··on US Administration announces 34% tariffs on China, 20% on EU
Let's not forget the biggest of EVILs, Lesotho, with its 50% tariff. Mr Trump has just noticed that they have cornered the US economy. (A strange coincidence, Trump has recently said nobody has ever heard of Lesotho being a country...)
District5524··on US Administration announces 34% tariffs on China, 20% on EU
Exactly. Calculation for actual figures is based on trade deficit, everything else is just a fat lie. "White House officials said its levies were reciprocal to countries, such as China, which it said charge higher tariffs on US goods, impose "non-tariff" barriers to US trade or have otherwise acted in ways the government feels undermine American economic goals" (from BBC) Non-tariff barrier usually means regulation US exporters don't like, including data protection, EU not wanting to buy junk food from the US, most of which apply to EU companies the same way anyway... There is no point in asking for a reduction of these tariffs, and will probably never happen.
District5524··on Apple Loses Top Court Fight Over German Antitrust Crackdown
Norway is not a member of the EU. Yes, precedent can be used with different meanings, and many clueless ministries of justices in continental law countries tried to innovate by calling some measures they made as turning that country into partly based on precedents (usually to restrict the freedom of judges, like in Hungary - which clearly failed). But more importantly, there is a world of difference between precedents 1) having in practice some relevant legal effect in a jurisdiction (e.g. using them as arguments or just as interesting case law) or 2) in the sense that lower courts are bound by some decisions of upper courts in a country, and 3) having a worldwide legal system built around stare decisis that is actually working the same way for hundreds of year, and where decisions from, say London, are read by lawyers in NZ or Singapore because they may have as much legal force as an act of a parliament. It's not just a spectrum, it's a very different beast for practical purposes.
District5524··on Apple Loses Top Court Fight Over German Antitrust Crackdown
Nope. 1) Not all of the "EU law" behaves the same way. There are legal acts of the EU that are directly applicable (like most provisions in regulations), there are even some provisions of the primary legislation (treaties) that can be applied directly (i.e. are binding and can be litigated). Of course, there are also legal acts that need implementation (like most provisions of directives). 2) EU law is upheld both by the EU Court of Justice (Luxembourg) and national courts, this depends on the cause of action you have. 3) An "agency" has a special meaning in the EU, that doesn't mean the same thing as a federal agency in the US (as a branch of the executive). There are instead EU institutions and national public bodies (whose exact nomenclature depends on local law) that may also have the task of upholding legislation. 4) No precedents are necessary or used at the EU level. Precedents are more a common law thing, they have an explicit binding nature for courts there. While there are common law countries (Cyprus, Ireland, Malta) with precedents, their precedents have no special place or role in how EU law works, there is no stare decisis how it works in common law. So, this is a national decision based on German national enforcement reviewed by a German court.
← PreviousPage 2 of 3Next →