He's not wrong though. In the 80s I'd watch video shows and over the course of a week I'd probably see some videos 10 times. And it wasn't background filler - I'd actually be sitting in a chair/couch and watching the videos. Kids don't do this anymore. First many/most songs are made popular through TikTok memes, not videos. And videos really are mostly just played in the background as they do other stuff. No one is just tuning into Yo MTV Raps or Headbangers Ball anymore.
To be clear -- this wasn't meant to be a GPT vs Claude post. This is more that frontier models continue to incrementally get better. No human steering for this level of math is crazy. Go back ten years and ask anyone in the field if this would be possible in ten years -- and you would have gotten "no" from every major AI researcher in existence.
GPT 5.6 Pro apparently solved all the IMO 2026 problems (competition happened July 15/16 in China). It was done with no human steering and on first attempt. I wasn't involved at all, just ran across this and thought it was impressive (the bar for impressive keeps rising rapidly though).
Italy was it for me. The train station in Rome was crazy. And just Venice in general. I was probably just in the touristy areas, but it was definitely the most hardcore non-stop street scamming I'd been around.
I was just in Rome last year. The only place I really ran into issues was outside the Coliseum. I just said “huh” and mumbled “I don’t know” to every question until they went away. “Where are you from?” “I don’t know.” I think they just thought I was an idiot, or had very limited English, which is fine by me.
I was told later by a guide that if you say “no, grazie” in a semi-convincing Italian, they’d assume you’re local and leave you alone.
That's weird because in Venice I never had a single problem, didn't feel scammed or anything, ate in very nice restaurants (I do avoid the tourist shops and restaurants though) and the people in general were nice. I almost felt at home, and many shopkeepers even spoke French, maybe better than English (I am French). I don't think anybody in the street accosted me at all, anyway. I'd go back without any hesitation.
Rented a car out of Munich and they forbade me from driving it to Italy. Said thieves are too common to risk driving, even in northern Italy. Train was simple enough though.
> And, I personally find the quality of YouTube Music Premium (256kbps AAC) superior to FM radio.
When it comes to playing music from phones in cars, connection type seems to matter more than the source, and iOS has some weird built-in sound normalization only for CarPlay that drives me crazy.
In my Audi A3 (2018), if I connect my iPhone 12 Mini via AUX or Bluetooth, sounds works perfectly fine. Same when playing via CD or USB stick inserted into the car, no problem. FM radio also works well, regardless of volume.
However, if I play music via CarPlay (Spotify [lossless], YouTube, on phone .flac files, etc) some built-in sound normalization seems to kick in and suddenly it ruins the music when playing even slightly louder.
I've tried for years to figure out what the hell is going on, tried every setting under the sun, but cannot get it to work so only thing left is some built-in sound ruinification ("normalization") that Apple does, only when played via CarPlay, not when playing via AUX or Bluetooth.
Seems to happen with every car I try it with, but I never tried a different phone. So right now I'm choosing between being able to have GPS or listen to music properly, as I cannot do both at the same time...
Maybe it’s trying to split the output into more than two audio channels when in CarPlay mode? If it were only Apple Music doing this I’d be sure this was triggering its Atmos output, not sure how it’d be affecting other players, but maybe.
Tried every conceivable player, Spotify, YouTube Music, YouTube, Apple Music (local files), Soundcloud and more, all of them leading to the same thing, via USB+CarPlay the sound get normalized somehow but if using AUX/Bluetooth, works normally.
> Maybe it’s trying to split the output into more than two audio channels when in CarPlay mode?
I hope so, most of what I play is stereo, and works fine via AUX/Bluetooth.
Hmmm. What other cars have you tried? I wonder if it’s the DSP path used for CarPlay. Or could it be the Audi system is clipping this source? I’d find it really surprising if Apple is doing something here.
I think this partially buries the lede:
"As a single hiring vendor comes to dominate screening for an industry, it may be more likely that candidates are shut out."
If we move to using just a small number of AI models to help do things like hiring, we will amplify biases and possibly completely lock out portions of the population. We need to be very careful when using AI systems to evaluate people in general -- not because they might be biased (which they might be), but because even a small bias, if used by virtually everyone, can be damning.
> We need to be very careful when using AI systems to evaluate people in general -- not because they might be biased (which they might be), but because even a small bias, if used by virtually everyone, can be damning.
I don't think this even requires any bias.
Assume there's some loose ordering of who is or isn't a good hire, and every employer has their own fuzzy view of it. If you get slightly better or worse as a potential hire (pick up an extra degree, let your latest certification lapse, whatever), it gets somewhat easier or harder to get hired.
Now assume that same ordering, but all employers share the same view of it. I'd expect the divide between employable and not employable to be much sharper.
Well, I'd say that specific ordering is the bias. But I see what you mean. The bias is arbitrary, but still very real.
Also, we will of course have all kinds of attempts to "game" the system to get ahead. Optimizing (even more) for the metric. Degree mills, for instance.
If you want to make meaningful change in this avenue you really can't use words like "bias" or "systemic" because anywhere from 49-51% of the population will immediately shut down upon hearing that. Someone can argue (and many do to varying levels of success) that systemic bias doesn't exist, which means this doesn't exist, which means there no problem.
However, "this AI model can decide that some subset of people, perhaps random, perhaps not, are simply not hirable for any job" makes sense to most people regardless of political bent.
The problem with the term “systemic bias” is that it takes a word that’s about differential treatment and changes the subject to disparate outcomes.
For example, the article here shows disparate impact: that different percentages of applications are passed through the AI filter. But it doesn’t show differential treatment of otherwise identical applications based on race.
Capitulation is a bad counterpropaganda tactic, especially with terms that have well defined domain specific meanings.
Note that the OP uses "systemic rejection", while the paper does reference bias, it is in the precise meaning of the word.[0] And this is not targeted at the general public.
You may want to look into the 1990 GOPAC handout "Language: A Key Mechanism of Control" to understand why some groups would simply just weaponize any term that was substituted. Academic papers need to error on being precise, to be effective, not focused on handling the general public with kids gloves IMHO.
Edited to add, listen to Lee Atwater's 1981 Interview on the Southern Strategy for even more context.
It's strange to say it might be biased. Bias is absolutely impossible to avoid, especially with how today's "AI" works.
You might be able to avoid it with a panel of AI, similar to how we try to avoid it by using panels of humans, but even that turns out to be contentious and not surefire.
I have feeling with AI it'll be even worse, since folks / companies can pass the buck (similar to how health insurance companies are now using it to deny folks).
> * Bias is absolutely impossible to avoid, especially with how today's "AI" works.*
Unless you're taking the "there are multiple mathematically incompatible ways to define bias" view of the topic, just do what's already known best practice for high-bureaucracy human review. Which is too define an overly-pedantic standard rubric.
Everyone knows there is bias. The problem this article highlights is that by delegating screening and human judgment to a few AI vendors those vendors will bias all employers in the same way.
Agreed. Humans are also biased, but our biases are different across a lot of socio-economic factors. So when we have different people in these positions, the biases become less bias-y.
But LLMs are statistical models. They are aggregating all biases into a general super bias. And they're all converging towards the same solutions.
I actually don’t think most consumers care about that at all. Consumers loved Napster. They have no problem stealing from artists outright, let alone indirectly.
I think to consumers AI denotes lack of accountability or oversight. They think it might work - but it might not and no one will care.
For example, I’m doing work in standardized test prep and there are tons of new AI products and no one likes it. Consumers feel as if they will get subtle but important things wrong. Most of these companies are now trying to hide that they are using AI generated questions.
reply