aposm

This, to me, is the strongest evidence that even with all the money frontier labs are throwing into pushing the boundaries of model training and fine tuning, they still don't actually understand what they're doing & don't have as much control as they want you to believe. If they had any real ability to tune & adjust the model & its output, they would have fixed "LLM voice" by now.

Even if you don't think it's a problem, even if you believe that it's just different than human writing, it seems the majority of "regular" people can tell when something is LLM-written and find it distasteful or unpleasant to read, at best. That alone should be more than enough incentive for frontier labs to train this out of their state-of-the-art models, if only they could. The fact that they haven't suggests that they cannot, despite all the claimed advances in model analysis and "explainability."

There's a pretty significant difference there, even in some of the examples you picked to make your point. TCP and HTTP are well defined standards that, for all their flaws, were designed and conceived to be useful abstractions that you didn't need to understand deeply in order to use. The things that LLMs allow novices to skip nowadays aren't designed with that in mind, and generally aren't something you can just skip over without getting yourself into hot water later. If you mess up Stripe or Auth0 setup because you don't understand the implications of the boilerplate the LLM spit out, you can very easily get your whole app compromised or end up on the hook for a lot of fraudulent credit card charges, etc. I see the point you're making and I don't entirely disagree, but I think current iterations of LLM coding tools still make it far too easy for beginners (or even fairly experienced engineers) to convince themselves that a detail falls into the former category of abstractions they can safely skip understanding, when they're very much the latter.

I'm now pretty convinced that this is the best argument for persuading non-technical and non-internet-privacy minded people why this is a problem. It's hard to explain all the technical factors that make it impossible to implement without compromising privacy... but it's pretty easy to get them to understand the real intent. If you start them thinking through what mechanisms would make sense _if_ the original premise truly was protecting kids from content they aren't ready to see, it's easy to arrive at that answer, and equally easy to see why "everyone must provide their ID so we can verify their age" probably comes from other motivations.

I only got hallucinations of random combinations of my (fairly unique) last name & first names that do not exist, combined with very accomplished and completely fictional biographies. I guess I'm not notable enough which is somewhat comforting.

Maybe you should not omit those "dozens of words" that clearly explain what you're trying to portray as a contradiction.

> These molecules have a particular property: When they come into contact with air, they oxidize and change color. One becomes blue and the other red.

When I lived in Japan it was in a relatively recent (last 10 years) but not brand new apartment block - Maybe if you are talking about a rural area or an old postwar Showa era house, sure. But either way the sound proofing is worlds better than any new construction in the US.

All the major metro areas on the Acela corridor are also on a straight line, on significantly flatter land than Japan. Notice how the Acela never spends 10+ minute periods in long, deep tunnels under mountain ranges. The Acela primarily spends most of the trip going below 100 mph because it is operating on 100+ year old infrastructure that has only ever been upgraded piecemeal as it starts to fail.

I think a big part of it is also that (partly because of the necessity of building for earthquake resistance), Japanese construction is a lot more robust than American housing, and also tends to have extremely good soundproofing on windows and doors. Actually, it's most of the rest of the world, except the US.

Yikes. I also just noticed that all the plugins (part of the pro feature set) rely on separate apk downloads from the Play Store, which all appear to be dead/delisted. This is really a shame, I too have thought of this as "the best" Android SSH client in the past.

Yes, it is enabled by default (in fact this caused problems earlier on). Licensed hams (in the US) can increase transmit power (and theoretically use additional spectrum outside of ISM) but even the default "public" channel was encrypted with a known, publicized key. There was some debate whether this ran afoul of amateur radio rules against encryption, even if the key is known, since it cannot be disabled. I believe there was some progress in fixing this and allowing truly unencrypted channels for licensed operators, but I haven't checked back recently.

We frequently get juniors or interns who are perfectly capable of pumping out many LoC with the use of AI in various forms - the issue is that they _don't_ actually ever learn how to think for themselves, and can't fix problems when something goes wrong or the LLM paints itself into a corner. I have found myself doing a lot more shepherding and pairing with juniors when they can't figure something out recently, because they just have not had the space to build their own skills.

A few years later but similarly - I am still running a machine built spur-of-the-moment in a single trip to Micro Center for about $500 in late 2019 (little did we know what was coming in a few months!). I made one small upgrade in probably ~2022 to a Ryzen 5800X w/ 64GB of RAM but otherwise untouched. It still flies through basically anything & does everything I need, but I'm dreading when any of the major parts go and I have to fork out double or triple the original cost for replacements...

How can I take your comment seriously if you didn't read the article? It looks like someone didn't actually make an effort to understand the context of the images.

> Moore's law is supposed to drive down the cost of electronics

The core of the article is this complete misunderstanding of Moore's law. From there, all the rest of the confusion follows, unsurprisingly leading to the author's claim that ~$100 for a long-lasting device is unreasonably expensive.

A human working on an existing codebase does not have any special signal about what is _not_ in a codebase. Instead, a (good) human engineer can look at how a problem is handled and consider why it might have been done that way vs other options, then make an educated decision about whether that alternative would be an improvement. To me this seems like yet another piece of evidence that these models are not doing any "reasoning" or problem-solving.