AStrangeMorrow

Basically compared to standard LLM models it is an order of magnitude cheaper and faster. (Note: I didn't get to actually try jev yet, just looked at demos/specs/pricing etc)

You can definitely do similar things with say hosted LLMs + a lib like outlines , or with API models and the proper output validation layer, but again way slower and more expensive.

And on the opposite side you can train dedicated classification models that will be even cheaper and faster to run than jev. But, well, you need to train them (costly, time consuming, and data might be hard to come by depending on target). Here you are a nice zero-short system, that can handle complex/messy data out of the box.

I mean now with modern LLMs a lot of good packages and tools get forgotten about. When you have hammer everything looks like a nail. Even if said "old" tools are actually orders of magnitude faster, and sometimes better too for that specific task. (And I remember spacy being basically SOTA for generalist NLP tasks not that long ago, like 2020/2021).

I do agree to some extent, but really the overweight threshold for BMI is 25. And 30 for obese.

At 6 feet / 183 cm that it takes 184 lbs / 83 kg to be overweight. From what I could find, for regular gymgoers the typical weight for people around 6feet is 180-190lbs, which put many people around the overweight threshold. I am myself at 24.9, and while not skinny I am definitely in the skinnier half of the people at my local gym.

Yes, it is a pretty high level metric with good correlation a bunch of diseases, but really many of these is because it is ALSO correlated to percent of fat, which is often the more relevant metric. But as you said BMI is so much simpler to measure.

Many active gym people have pretty high BMIs but fairly low fat (because muscle is dense), and unsurprisingly have better outcomes than the average person (if you ignore the share that uses/overuses anabolic steroids and co)

Same for France and Spain, which is what I am familiar with. I am expecting it to be true all around.

The map of France shows maybe 100 castles? If you include castles, forts and chateaux (which it seems the map aims to do) France has around 45000 of them, including 15k medieval castles (military), an other 15k from the Renaissance period (mix of military and luxury), and 20k more recent ones (some military, many luxury type).

I know if at least 20 medieval castles within 1h of my hometown and not a single one is here.

For Spain maybe 10k to 20k castles and forts. Out of which at least 2500 genuine medieval castles (not just big houses with fortifications)

I know, I am in the same range of points and still asking questions has always been a bit scary.

I remember spending 2h writing a question for what I thought was a complex c++/compiler issue. 10s of thousands of lines proprietary codebase, so I couldn’t include everything obviously, but also couldn’t create a “minimal working example” to reproduce the issue. So I included as many things is I could to try to get pointer on how to track that behavior I was seeing. Of course the second I post it I got a -1 plus “can’t reproduce”/“please add minimal example”.

An other time, I had a question that was very similar to an existing one, but different setup and the answer did not solve my problem at all. Mentioned all that, linked the other question and specifically wrote that it was NOT addressing my problem. Posted it, soon after tagged as duplicate with that one answer that did NOT solve the problem.

After that I rarely asked questions again.

Also the points system made it frustrating as a new user: someone 2 years ago asks a basic language question “+50 upvotes”. You asked a similar question, asking extra clarification on an aspect “-2, already answered, read the doc” and so on. And with such a big deal made about reputation it felt like just being born early and being able to be an early adopter meant you got east points. For new users, though luck.

The evidence that humans would naturally be designed to sleep “multiple times a day” is quite mixed. “Multiple” does a lot of heavy lifting here, when basically most evidence points to two batches, following either of these patterns:

- a long uninterrupted night cycle and a short (20-60 mins) afternoon nap. Around 2pm. - a night cycle split into two halves. With a 1-2h break (maybe up to 3h) starting around midnight to 2am.

The former is still very common, and imho stretching the definition of multiple cycles. The latter is more historical (more common when there are long winter nights and no electricity).

Also making it a “Western” problem is kinda weird? There are other cultures where single cycle sleep has existed. Even hunter-gatherer groups with little to no contact with the west. And alternatively afternoon naps are still quite common is some western areas. I guess the main thing that prevent it would be the classic work day schedule.

Well of course as you pointed out the legitimate one would be copying voices WITH permission (yours, someone you know who gives authorization, through contracts for movies/bots etc). The model can’t differentiate between voices for which you have permission or not.

But more generally while recordings might be copyrighted, the voice itself isn’t so copying a voice isn’t a crime, at least as it currently stands. You cannot however use said voice for deceptive practices. You can however for advertisement (needs permission). And in the US you can for satire, at least in the US, withOUT permission (falls under the 1st amendment).

This drives me crazy when dealing with various contractors in my area. They often only answer by text, and can take a little while to do so. Not a big deal.

But If I send a message with more than one question, there is what feels like a 10% change they will answer all my questions and a 90% chance they will either answer one or none and I’ll just get an “ok”. And I am talking about 5 lines messages not 50 lines.

So I have to send my questions one at a time, wait sometimes minutes to hours for an answer, then send the next one.

Of course I can also call, but often can’t reach them. Or can reach the front desk that doesn’t have the answers. I understand people are busy but it turns something that should be one message into a cat and mouse game

From looking at how that was done, it seems they (the paper you linked) used an older paper which looked at which names are frequent enough and more biased toward a certain demographic (90% of that name occurrence falls within that demographic).

But they picked 9 family names per group. Which sounds quite low. And combined that with first names to reach 500 first+last names per group.

I wonder how much of the bias we see has to do with the names actually picked versus it being racially motivated (absolutely not denying that this probably is a factor, but might not be the only one).

For example, in France there is the national BAC end of high school exam. If you you at the names X grade distribution, and look at the higher “very good” bracket: some names are heavily under-represented (less than 5% of say “Jordan” get that grade) while some are over-represented (35% of “Josephine” get such a grade). The exam is for the most part anonymous, but some names are definitely heavily correlated with lower/higher income groups. So nothing surprising: Josephines tend to come from richer families, thus in average get better education/support, thus better grades. Same thing is true with family names to a smaller extent.

So I wonder how much of the bias we see, be it from real persons or the AI has more to do with a class thing than a racial thing. Again those are not neatly separate things, but still

Yes possible. But really that video of them features the word prominently (even on the thumbnail) AND that vocabulary estimation website. The video/podcast is just slightly over a week old.

Anyway doesn’t really matter, it was more to see if anyone else was a listener of that podcast.

Someone watched “The rest is Science” I imagine!

Not sure how to understand that. You mean as the best engineers?

Funnily at my company, the few engineer that did the majority of the work before AI still do the majority of the work now. By majority I mean tackling both more issues and better.

However there is a general verboseness and over engineering trend across the board.

I have a similar issue. I tend to have a very “structured” type of writing. Say on slack or Reddit for example. Using markdown formatting. Lists with bulletpoints etc. And I tend to write long detailed explanations, sometimes too long if I am being honest.

But now I find myself adding noise and imperfections to my writing (not that it was perfect) to make it more human, which is kinda silly.

Yes. I mean if you look at the corpus basically HALF of recipes are Chinese/Korean.

They do quickly acknowledge it, but definitely not a balanced set.

I still love learning, especially outside of tech. Been working in the ML field for over 8 years, and while I went into it because I liked the field, I did lose some interest in learning things, but mostly because of the sheer volume of publication and the rate of change. Learning stopped being something I enjoyed doing and went to something I had to do to keep up. And it just stopped having the same flavor.

I also like meaningful commit names. But am sometimes guilty of “hope this works now” commits, but they always follow a first fix that it turns out didn’t cut it.

I work on a lot of 2D system, and the only way to debug is often to plot 1000s of results and visually check it behaves as expected. Sometimes I will fix an issue, look at the results, and it seems resolved (was present is say 100 cases) only to realize that actually there are still 5 cases where it is still present. Sure I could amend the last commit, but I actually keep it as a trace of “careful this first version mostly did the job but actually not quite”

I mean “mistakes” can be hard to define. IMHO there is an area of responsibility between the LLM, the LLM user, and the code itself.

Did it make a mistake because I didn’t follow instructions properly or hallucinated some content?

Did it make a mistake because the prompt was unclear/open to interpretation or plain wrong?

Did it make a mistake because it lacked some context? Or too much context and it starts getting confused?

Is not handling edge cases automatically when that was not requested a mistake?

I am not just trying to defend LLMs, in many cases they make obvious mistakes and just don’t follow my arguably clear instructions properly. But sometimes it is not so clear cut. Maybe I didn’t link a relevant file (you can argue it could have looked to it), maybe my prompt just wasn’t that clear etc

I have still mixed feelings about LLMs.

If I take the example of code, but that extends to many domains, it can sometimes produce near perfect architecture and implementation if I give it enough details about the technical details and fallpits. Turning a 8h coding job into a 1h review work.

On the other hand, it can be very wrong while acting certain it is right. Just yesterday Claude tried gaslighting me into accepting that the bug I was seeing was coming from a piece of code with already strong guardrails, and it was adamant that the part I was suspecting could in no way cause the issue. Turns out I was right, but I was starting to doubt myself

I mean I do agree, and on iNat I can clearly see my house and the house of a few other people in the neighborhood. However you can easily find the current owner information for a given house in the state I live in, and since we bought the house, our name.

I guess it is different once you look at people renting, and also you could track a specific person posts to see when they are posting away from home for example. But as far as revealing your home address, sadly there are many other ways in a lot of cases

I imagine they are taking about 4D golf by CodeParade, as seen here: https://youtu.be/y53UNskR-zU?si=iUfCkxYqkACx955t

Steam link: https://store.steampowered.com/app/2147950/4D_Golf

The person goes over quite a few technical details on their Youtube, though they talk about a bunch of other coding experiments too.

I am not sure, but you don’t have to technically bet on assassination. You can bet on an event which would happen as a result of said assassination. X won’t get re-elected. Company Y CEO will change in 2027. This is artist Z last tour. Athlete K won’t participate in this event etc.

The issue is the combined risk of insider trading coupled with the bias of disaster-centric betting, or at least event-centric betting. This means if you have the means to create an “out of the ordinary” event you have a strong incentive to make it happen and to bet on it. These must be controllable events, so not natural or complex systems. On the gentler side it would be sports fixing, which has always existed. On the worse side it would be causing war, making economic decisions that will impact many, betting on people death and so on. These kind of things are seemingly already happening to a certain degree.

Arguably so is “aim/ein etc” and “in”, though more dialect dependent and more subtle.

The former for me have a bit more exhale and round sound while the “in” are a tad drier.

For example “fin” and “faim” are distinct for me. However “faim” and “feint”

I am curious about how much energy needs to be expanded to contain the anti-matter. Say it the matter/anti-matter is to be used for propulsion/energy generation can we reach a threshold were we are actually energy positive

Yes it feels like a full time job just to try to keep up. And I’ve been in AI for close to 10 years so I feel like I have to keep up at least a minimum.

An other thing for me is that it has gotten a lot harder for small teams with few ressources, let one person, to release anything that can really compete with anything the big player put out.

Quite a few years back I was working on word2vec models / embeddings. With enough time and limited ressources I was able to, through careful data collection and preparation, produce models that outperformed existing embeddings for our fairly generic data retrieval tasks. You could download from models Facebook (fasttext) or other models available through gensim and other tools, and they were often larger embeddings (eg 1000 vs 300 for mine), but they would really underperform. And when evaluating on general benchmarks, for what existed back then, we were basically equivalent to the best models in English and French, if not a little better at times. Similarly later some colleagues did a new architecture inspired by BeRT after it came out, that outperformed again any existing models we could find.

But these days I feel like there is nothing much I can do in NLP. Even to fine-tune or distillate the larger models, you need a very beefy setup.

Might be surprising but I am kinda willing to believe it. Since we bought our house, we had quite a bit of work done by professionals. But whenever I can I do things myself.

Like I had multiple companies quote me $300-500 based on the job for things that take me maybe 2-3 hours total to do, including learning about it (will be faster next time), getting the materials, and doing the job.

When you have a few of these a months they add up. It is usually nothing for a month and then 4-5 things to fix/improve the next

I remember having to describe a standard model to predict online shopping behaviors for my ML class exam in university. That was close to 10 years ago now.

Also remember a teacher telling us about that story of a company finding a woman was pregnant from her shopping behavior and pushing relevant recommendation. Prompting people around her like her dad or something to find out she was pregnant

Idk basically everyone is my org has seen some good value out of it. We have people complaining about limitations, but would still rather have that tooling than not.

For me the main difference is now some people can explain what their code does. While some other only what it wants to achieve

For me, the main thing is to never have it write anything based on the goal (what the end result should look like and how it should behave). And only on the implementation details (and coding practices that I like).

Sure it is not as fast to understand as code I wrote. But at least I mostly need to confirm it followed how it implemented what I asked. Not figuring out WHAT it even decided to implement in the first place.

And in my org, people move around projects quite a bit. Hasn’t been uncommon for me to jump in projects with 50k+ lines of code a few times a year to help implement a tricky feature, or help optimize things when it runs too slow. Lots of code to understand then. Depending on who wrote it, sometimes it is simple: one or two files to understand, clean code. Sometimes it is an interconnected mess and imho often way less organized that Ai generated code.

And same thing for the review process, lots of having to understand new code. At least with AI you are fed the changes a a slower pace.