The same what reddit did, "this is our content" we'll try to milk it to the tilt. I think this is akin to a reverse "tragedy of commons" but for user generated content verticals. It makes economic sense for them to try to lock it and they will try to lock it unless there is a market force that prevents that. Sadly, market forces in these markets are few and far between.
Cory Doctorow argues that the antidote to walled platforms is to restore the legal and technical right to adversarial interoperability. Users and developers should have the legal freedom to reverse-engineer APIs, build scrapers, create cross-platform bridges, and deploy third-party readers without facing lawsuits under anti-hacking statutes like the CFAA or DMCA Section 1201.
Once again, yet another legitimate comment that was [dead] that I had to 'vouch' for. HN either has broken filters (understandable, combatting spam) or seriously loves censorship
Check the shorter "Advice on upskilling" or "The Math Academy way" for well researched approach.
Basically, using the books as the source for the LLM to limit itself by RAG-ing the text and then generate the required tests, questions and explanation of the concepts.
Indeed. At uni, starting in my second year, I worked as a programmer (no remote work, office at 1hr commute from uni dorm) on top of my studies. It was hard, but doable. My major was applied math, so the CS part was the easier bit, but the math, especially Analysis, Numerical Methods, Linear Algebra, and Analytic Geometry they kicked my butt. We also had a really bad prof for the Math Logic class and no textbook, only lecture notes and his lectures were unstructured shit :(
The biggest problem wasn't the workload itself, but how inefficient learning was back then. It was the early days of the Internet, no coursera, no Google indexed uni sites, no LLMs, no online tutors. I only had a mediocre textbook and some notes I borrowed from friends when I couldn't attend lectures or labs. It took ages. I’d burn hours trying to parse two paragraphs in the textbook, only to ask mates who didn't understand it either.
Now, we have an incredible amount of information & tech at our fingertips: LLMs, Google NotebookLM, online tutors, dozens of textbooks, and sites like Anna's Archive packed with high-quality "free" material. With all this technology, the time it takes to understand, prepare, and master a topic is a fraction of what it used to be. It still takes work and discipline, but man... you can feed textbooks into an LLM and use targeted prompts to debug your exact knowledge gaps on the practice test you get from the LLM.
So I reckon it's easier now than before, simply because learning can be done so much more efficiently. After all these years, I still remember the hardest classes and the most stringent professors with great affection - even though I hated those bastards back then. My Group Theory prof once invalidated a definition I wrote because of a single missing word. At the time, I thought it was an injustice, but later I realized that word was crucial.
University is where young people must face hard work and reality, because it teaches them what true meritocracy is. It should be hard, it's the same way a gym workout should be hard or it’s ineffective.
Check the shorter "Advice on upskilling" or "The Math Academy way" for well researched approach.
So what works for me
- open a project in ChatGPT/Notebook LM
- dump all the relevant and highly cited materials (textbooks, papers)
- dump the advice on upskilling text or a short summary I've written for the LLM
- create "Learning Goals", that contain what I want to learn, and how to estimate is my level good enough
1) Ask it to create a learning path from the materials, following the approach. Give that to an adversarial LLM for cross check. (just for sanity check)
2) Ask it to create an "entry test" to check what I do know and what I don't
3) Iterate step by step on each module/submodule from the learning path that intermingles the approach of: small theory step + small practical task + small test. Log what's missing/wrong in my dept log. Give the dept log at the end of the session to the LLM to incorporate/create another test/task.
What I have found useful in this approach is that it will generate a lot of practical tests/tasks for me and it will explain a concept in many ways until I understand it. Also it finds some prerequisites I might miss, but based on my tests and debt log unexpected things I thought I understood surface.
So with the limits of LLM and while building a mental map of the relevant parts it's usually enough to spot the hallucinations, but if you apply structured approaches these are minimal. And it's super good, because the number of practice tests and explanations is endless.
The interfaces are a bit clunky, but current multimodal LLMs are ok with images or even hand writing.
The modern societies run via those devices and the enforcement will move to the mostly free Internet that was "a long time ago, when it didn't matter as much".
It was the market that delivered BYD and Xiaomi. The government did his part with industrial policy and inviting Tesla to China to help start the ecosystem, but the brutal competition between Chinese auto makers following the rules of capitalism and free-market made those who survive the ones that can compete with other global companies. Before that they just subsidised and all they got were regional "champions" that siphoned out the funds and created shitty products no one wanted.
After Tesla came with special laws allowing it to work without Chinese co-ownership, but with a mandate to source from Chinese suppliers, this created the ecosystem needed for the next iteration, by the capitalist urge of Elon to get into the market and by free-market economy of suppliers to Tesla that are up to the standards to supply it.
The government set up the battlefield, and let the free-maket filter the strongest to survive.
In the Western countries the government and regulations are frequently the bane of the car companies. Good luck opening a mine or a battery factory in Germany, regulation, unions, green mandates, 36 hr. week. You can't fight with 1 leg tied.
This narrative overweights free market competition in China and any influence Tesla or Musk might have had versus the $230B in subsides provided to the Chinese EV industry over 14 years (2009-2023), with over ~$1.5B provided to CATL alone (although Tesla did receive ~$17B in total government subsidies in the US). Tesla did not begin production in China until October 2019.
If there’s any hero story here, its the will of the CCP to provide this level of support and subsidy to the Chinese battery and EV industries as part of their five year plans. This is why China can build ~20M EVs per year, Europe ~4.6M, and the US can build ~2-3M.
I understand this deeply challenges the ideas and belief systems around the cult of personality of risk taking entrepreneurs and entrepreneurship being something more than a series of successful calculated risks and bets.
> When it comes to China’s electric vehicles (EV) industry, its market has expanded exponentially compared to other markets. Now, a new study by the Center for Strategic and International Studies (CSIS) has revealed that China’s EV industry received at least $230.8 billion in government support between 2009 and 2023.
> As a result, this major investment has played a pivotal role in China’s rapid progress to become the world’s largest EV market, with sales anticipated to reach 10 million units in 2024.
Yes, there was investment, but investment without free markets and capitalism doesn't work usually for end products like cars. It's not a linear function of the capital invested, it's not possible to spend your way into good consumer products. That's not infrastructure. That's why I was mentioning that before Tesla came that investment didn't lead to any strong EV automotive players. In 2016 there was a big scandal on widespread EV fraud, where manufacturers just built to get the subsidy(骗补).
The policy shift came around 2017 when the government understood that subsidies and closed market created lazy automakers (e.g. check the US market for the same dynamic in trucks). In 2018 they removed subsidies for cars less than 150km of range and changed the requirements to receive them tying them to battery density.
After Tesla came and subsidies were reduced, most of the inefficient companies went bankrupt. It was the free-market competition that killed them. The reason why it changes was the competition not the subsidy. The subsidy was somewhat important, but as many of us knows, nothing government run produces good consumer products, you need market and free choice to get productivity. So it was the free-market and subsidy cutting that created the BYD, Li Auto, Xpeng and Nio. That's the proof that without free market it's impossible to have strong companies. And without subsidies you can have strong companies, even if it takes more time.
In EU and USA subsidies didn't create a lot (Northvolt?), because there were not enough free-market driven demand and too much regulation to jumpstart the free-market forces that will start the engine of innovation.
So the answer is not "capitalism and free-market doesn't work and governmental industrial policy is the answer", but it's the market that creates the strong companies. You can speed it up with venture/government capital, but without the brutal competition in the marketplace where people vote with their wallets you won't make it work.
As the internet become the place where people do a lot of things, no government (and especially no security services) will be able to keep themselves from trying to control it or at least monitor it. And with the new LLM features they can automatically do much more than before.
Human nature is a constant and when the government sees an easy way to enforce something, many more bureaucrats will try to do it.
There is also another benefit to books, on average they are much better than a random 3 hour podcast.
If you care about what you read, you'd be getting something that the author has spend a lot of time, skill and energy to write, the editor would have spend a lot of time and skill to improve with the author.
I have a measure for all content I consume, quality/hr of reading/listening. If it's just a long video that has 2-3 questions that has caught my attention I'd be listening only those. If it's a long text that I might find something interesting I'll ask the LLM to summarize the main ideas as a filter before I decide to dive in.
Books, and their audiobooks version have on average much more bang per hour than random podcasts, because they're structured, authors had spend more time on them and you can cherry pick from a structure.
I also have caught myself using sloppy content as excuse not working on planned tasks with excuses like "this might be useful", or watching "productivity porn" videos. I think LLMs are good as a pre-filter for that.
This is a frustration for me as well. A few useful tricks, and some areas I'm looking to explore:
- AntennaPod (Android: https://antennapod.org/about/) doesn't specifically block ads, but does make skipping them fairly painless. I'm not familiar with the iOS app space.
- Seek out, and add a classification tag for ad-free podcasts. When you're not in the mood for dealing with ads, play these.
- Protip: if you learn, or want to learn, German, Deutschlandfunk (and a number of other German-language broadcasters) have a set of excellent, ad-free, podcasts. This includes a number of podcasts for learning German (generally through the Goethe Institute or Deutsche Welle).
- If your podcast app permits it, set your forward-skip to 30 or 60 seconds (the length of most ad beds), and backwards to 5 seconds. You'll be able to navigate past most ad blocks more easily. You can also set begin/end skip periods for start/end of episode advertising.
- I've thought of manually editing episodes from a desktop session using audio editing software (Audacity or similar). That's ... a bit of additional overhead, but as with other mise en place techniques, you incur the overhead once and don't have to worry about the interruptions when you're in the middle of listening to an episode. Audacity shows sound signatures and I'm expecting that most ad blocks will be readily apparent. I also suspect AI tools might be able to remove ads fairly reliably, though haven't looked into this yet.
I've definitely noticed that I deliberately avoid listening to podcasts which have ads when I don't have the bandwidth / freedom to deal with them (e.g., doing other tasks, walking etc.). And advertising has become more pervasive, longer, more intrusively inserted, and annoying with time.
Other ad-free English-language pods: Tech Can't Save Us, History of Philosophy (Without Any Gaps), and Philosophy Bites (and several affiliated podcasts). All are highly informative, well-produced, don't fixate on current events and politics (which ... I find maddening). And several of those would appreciate any support as well.
Late add: the German-originated ad-free podcasts include several English-language channels as well, covering not only language-instruction but culture, news, technology, science, and other topics.
Other national English-language broadcasters (NPR, PBS, CBC, ABC (Australia), etc.) have podcast channels, but most have succumbed to advertising. That isn't to say they don't have good programming, but don't expect ad-free listening. For a number of cultural reasons, Germany seems to have incorporated the idea that advertising itself has deep costs and its avoidance strong benefits.
Yeah, I can skip forward 30 sec, then back up, blech. It's just not worth it when you constantly have to interact with the podcast app.
Scott Adams' podcasts were different. He inserted very few commericials, and they were short enough there was no reason to skip forward. I tried many other podcasts after he passed away, and they all were largely long, boring commercials. Yuck. I now listen to Pandora or Soma FM instead.
> Yeah, I can skip forward 30 sec, then back up, blech. It's just not worth it when you constantly have to interact with the podcast app.
Fortunately this can be done much more easily now, with headphone-based controls and smartwatch-based controls. It takes maybe 1-3 seconds for me to get through an ad break and be back to listening.
There are podcasts which are just free-association rambling (or worse), others which are very closely scripted and edited.
I very much prefer the latter, and the best of those approach books in structure and/or value, if they don't directly produce books themselves (e.g., Peter Adamson's History of Philosophy, which is both a podcast and a book series).
I got this problem with my own employees, LLM are fine, but lazy slop is not permitted. Current idea is to have a clear "best practice" template for most of the research/specs/problem definition they submit and it reduced the slop to a manageable level.
But this might work in a smaller company where the management is reading and is strict about these things.
I have one IKEA Apstuga on my desk, sitting right next to a good CO2 monitor.
Since Apstuga uses worse approach (heat) rather than light as the good sensors, it diviates around +/- 100 ppm. For example the correct CO2 is 610 ppm and IKEA's sensor shows 552 ppm with is reasonably close. So the trend will be correct and the values will not be.
But when it goes over the safe limit it should be enough to decide to ventilate.
Cory Doctorow argues that the antidote to walled platforms is to restore the legal and technical right to adversarial interoperability. Users and developers should have the legal freedom to reverse-engineer APIs, build scrapers, create cross-platform bridges, and deploy third-party readers without facing lawsuits under anti-hacking statutes like the CFAA or DMCA Section 1201.
reply