Fungible/non fungible is a good alternative, and maybe the technically correct word. But I think in that case it doesn't apply and the change the author did is better.
>Dario and Boris have us convinced that “coding is solved” with their loops. But microwaves didn’t solve cooking.
If you want a look at the timeline where the microwave solved cooking, this was an interesting article: https://malmesbury.substack.com/p/my-journey-to-the-microwav.... You can apparently sear meat with a microwave (provided you have the necessary pan).
Training data has stopped being a good predictor of LLM abilities ever since they started doing heavy RL runs. I'm not sure how much corporate dashboards/I can't believe it's not excel stuff were in the training data, I guess not that many considering that stuff is almost always corporate and kept inside companies, and LLMs are still great at it, good enough to make people that used excel and used it well daily for 10+ years stop using it for lots of stuff.
This isn't my experience at all, LLMs do graphs and more complex excel-like web pages very well. They also do dashboards very well. They even seem to do 3D stuff with three.js like video games pretty well too, although I haven't tried that myself. Maybe they can't do something "great", but they sure can do good enough in most cases, and good enough is already better than most websites.
>I'm not a frontend dev, but these statements are starting to get outright disrespectful to those that are.
Agree here, especially considering that usually "niche scientific codebases" have terrible code so you don't need a super smart model to get a good bost in software engineering.
>and also no doubt that using AI will be slower for tasks where it's less about writing code and more about context/world knowledge or building understanding
This isn't true in my experience, AI is great at gathering context through slack, repositories, emails, web pages. For building understanding too, provided you use it well.
Not in my experience, it tends to pick up subtle orientations given in a question (like "which is better, A and B?" and in the context you add you list a few things for A and B) and will absolutely run with them even if they're not true. Has been an issue with Gemini models since at least 3.0. Maybe that makes them great roleplaying models, but for factual information they just run with the slightest hint in one direction or another and never really push back objectively.
I don't know about understanding but Claude and GPT can "recognize" lots of race conditions/possible deadlocks and then run Go with the race detector to figure out if they really happen (actually not all the time they tend to be overconfident that things are race conditions without testing first!).
I don't really know what to call that, if it's understanding, recognition, but it's clearly helping reduce the number of race condtions.
Artificial analysis shows Sonnet 5 as ~2 times more verbose than GLM 5.2. I wouldn't call Sonnet 4.6 underrated, it's in "chinese open source model territory" and unless you rely only on subscriptions it has alternatives.
It's not a high risk, but it's a thing the author said that's wrong, which puts into question the rest.
I think ideological objections to a law that will be imposed on ~300 millions of people through a process without much democracy, oversight or traceability of the decisions is pretty important, more than the technical details.
Some actually object to age verification in real life, for example some people find ways to consume alcohol or buy it before they're the legal age. Some people drive before the legal age especially in the countryside. So it is wrong to say that nobody objects to age verification in real life.
The way the thing per se is presented is skipping the ideological debate to focus on the implementation, which I think is pretty bad from a democracy/freedom point of view, and then proceeds to be wrong about something quite basic like "you don't need to give your birthday", which makes me worried about the rest, and then lists a laundry list of possible issues. So it seems clear that it can be leaky right now. The date of day easily leaks, we can argue about how important it is, but the author explicitly says:
>To prove to a porn site, a gambling site, an online liquor store, or a religious forum that I'm an adult, I should not have to hand over my name, date of birth, ID number, face, address, or passport. That is a dangerous amount of information to give any private website, let alone one dealing with sensitive content.
Note the "or" and "dangerous amount of information".
There is also no built-in way to ensure the person using the ID is the person they actually are, and no privacy-respecting way to confirm that is presented.
I don't think this problem is related to the fact that they don't have a world model, or because they don't form a mental model of how everything fits together, or a fundamental limitation of LLMs. These claims are often meaningless, and the boring answer is usually something like "software architecture is harder to verify than code/maths so RLing on it is harder, and it's harder writing good evals/benchmarks for it".
I don't think a different way of paying should be considered a subsidy. From what I understand AI companies are making money on those plans.
The limits feels really usable to me, but maybe because I've learned to work within their limitations. It's like, it's always possible to consume more tokens for maybe more output.
- the author claims the website doesn't need to know the date of birth, but as said it is easy to derive it. Therefore the author was wrong on that point, which makes me wonder if he's wrong about the rest too
- the author list 10 "Things that would break the promise", saying we should fight for them to be respected. That seem to imply, to me, that as is they may not be respected, and even if the implementation he imagines is possible in theory, it might not be in practice. The author also hasn't considered that maybe people are against age verification precisely because the promise of a well implemented, privacy-preserving age verification is easy to break.
- the author hasn't considered adversarial use, for example using the identity verification from someone else, and what other checks could be added to ensure those. There is a slippery slope of "first we check your age in a privacy-respecting way, then we realize we can't really enforce it like that but you already agreed to age verification so we can reduce the privacy respect to better enforce it"
- the author is using a possible implementation at one point in time to transform a political/ideological issue into a technical issue, while not considering all aspects of said policital/ideological and technical issues; as explained above.
The article says the website doesn't need to know your date of birth, that the state will issue you a certificate once you're over 18. Since many people I think will do that as soon as they can, it's easy then to see patterns of "this person created an account on age restricted websites with the same email address, they didn't have one before, therefore they probably just turned 18".
I find it also sad (and maybe a bit suspicious) to see someone that has blogged since 2003 use AI to generate blog articles, especially for a controversial law that has been pushed again and again in undemocratic ways.
>Not a prescription, a starting buffet. These are popular, well-supported defaults, not the only right answers. Tap any tool to open its site.
These AI tells are getting really easy to notice. A negative that absolutely isn't needed in a sentence, and you know it's AI that wrote it (, not a human ;) ).