Yeah this is the story I tell myself to make me feel better. I may well have even more regret! Or thought I was an investment genius and invested loads more and then sold it at a bad time.
Medical, banking and insurance are three industries that the European data privacy watchdogs are much more strict about because of the potential for damage.
I'd say the numbers listed here prove the GPs point of poor enforcement. The largest fine is roughly 0.97% of Meta's 2023 revenue, the equivalent of a $600 fine for somebody making 60k / year. It's a tiny-tiny cost of doing business at best, definitely not a deterrent, given Meta's blatant disregard for GDPR since then.
> the equivalent of a $600 fine for somebody making 60k / year
I don't know about you, but on that income I would certainly not brush off such a fine as a "cost of doing business". Would it cause me financial trouble, or would it force me to sacrifice other expenses? Absolutely not. But would I feel frustrated at having to pay it, feel stupid for my mistake, and do my best to avoid it in the future? Absolutely yes.
My bad, a better analogy would be a dealer making 60k / year selling drugs, gets caught by police and is fined $600. I wouldn’t expect them to change much.
1% of Meta's global revenue is a tiny-tiny cost of doing business? At that point, I think I can stop even trying to argue here. It's a massive fine any way you put it. Especially when you consider the ceiling hasn't been reached and non compliance is more and more costly by design.
Their net profit was $60billion in 2024. This is peanuts. It can fluctuate by multiples of this fine in a month, depending on whether or not they've had a bad or good month, nevermind year. This pretty much is just a cost of doing business.
The interesting part is that it keeps going up. You seem to believe we have somehow reached a cap where Meta can just expense it as a cost of doing business. That's not how European law works. The fine maximum is far higher and repeated non compliance keeps making the fines higher and higher. It's a ladder not a sizing precedent.
Unfortunately it doesn't in practice. Meta's total revenue since 2018 when GDPR came into force is just shy of $1T. Even with all the smaller fines combined, the total amount of GDPR related fines is in the range of $3B. It's a rounding error.
There isn't a trend of increasing fines, nor has any fine even reached the cap, let alone applied multiple times for the recurring violations. Even more with the current US administration's foreign policy towards the EU.
While GDPR as a law is fine, with the exception of enforcement limitations, enforcement so far has been a complete joke.
Maximum GDPR fine is 4% of global revenue in the previous year. If a company has 30% profit margin then they can, in theory, treat is as a cost of doing business, indefinitely.
It's 4% per fine. Each violation is a fine and Meta owns multiple companies that can be fined. But 4% of global revenue already can't be treated as just a cost of doing business. Their shareholders would murder them.
Agreed, this sort of thing should at minimum be considered gross negligence at this point, but because regular consumers who buy these products rarely see and almost never understand these news articles it doesn't really impact sales so the company doesn't care.
If this discovery was guaranteed to result in meaningful fines companies would get their act together pretty quickly. 7000 counts of negligent exposure of private data (camera/mic feeds) should in a just world be millions of dollars in fines at the least and arguably criminal charges for management.
Exactly. If GDPR fines can be so high, then something like this that is pretty much intentionally leaking personal data should be in the same ballpark.
His successor is https://languagetransfer.org, which is just a labour of love by a genius polyglot and language teacher.
So much so, in fact, that the owners of the Michel Thomas IP tried to sue him for stealing the methodology. The EFF, back when they actually did anything, shredded them.
I _loved_ Language Transfer when learning Greek. I haven't used it in many years, but at the time remember that I went from just being able to say "Hi" and "How are you" and "Good" to speaking full sentences with my boyfriend at the time in one day. And when visiting Greece later that year I could get by with strangers in everyday interactions easily. It was a mind-blowing learning experience, as someone who is not highly gifted in languages, and I donated to the LT creator.
Now I am learning Swedish. It has been taking me _way_ too long and unfortunately LT doesn't have a Swedish course. Looking at one of these documentaries about Michel Thomas it does indeed look like exactly that kind of approach! And I see he has a Swedish course. I'm excited to give it a try!
I have listened to many hours of the German course but the dynamic between the two students is distracting. They essentially play 'the genius' and 'the grouch'. The good one remembers everything. The bad one snorts and grumbles at every question. The acting is so forced that it completely breaks the immersion.
But I don’t think it is acting. He just only did one take. Usually someone is better then a the other. That’s what I understood from the documentary anyway. He just did the tapes because he was old and someone bugged him enough.
ChatGPT gave a reasonable looking answer to my prompt "Summarize what is special about how Michel Thomas teaches Language".
Maybe just another case of a highly intelligent person coming up with an "obvious" solution that is great: yet is not quite so obvious to others. He clearly was talented - but also he avoided explaining the rationale behind his method for years.
This is where arcade machines should have all gone. More interesting experiences with hardware that are really difficult to replicate at scale.
The best arcade games sell did this - it doesn’t take much - like the pedal for time crisis. Sure you _can_ buy one at home but most people don’t and even then it’s a crap placid pedal.
At least in the US, those "deluxe" cabs with motion just never seemed like a viable deal to me as a kid/teenager visiting arcades.
It was like, $1 per game compared to $0.25 or $0.50 for a normal cabinet.
As a young person with limited income, it DEFINITELY mattered to me... I preferred to sacrifice a little bit of motion and enjoy 2x or 4x the playtime on something else. I mean realistically you'd be spending $20 an hour or more if you stuck to deluxe cabinets. At that point (according to my teenage mind) I was basically halfway to buying a home console game that I could keep forever.
Operators really should have priced those deluxe cabinets the same as regular games during off-peak hours.
They’re being made but I just don’t think there’s a whole lot of demand/spaces for them. People sure don’t want them in their homes and arcades barely exist in many countries now.
I’ve seen a couple of bars open up that try to have an arcade as well but they never take care of the machines/drunk people break them, so after a few months half the games don’t even work. There’s only so many times I can lose a quarter or a dollar before I decide it’s not worth it anymore and I just go drink somewhere else with friends.
The only real arcade left in my city is attached to a laser tag, it would be super weird for a bunch of grown men in their 30s and 40s to roll up during kids’ birthday parties they weren’t invited to lol
Yeah, this tracks. My city has a few "retro" arcade bars with pinball, pacman, etc, and it's fun enough but you're for sure going for the nostalgia more than anything else.
I think part of the barrier to expanding the attached-to-other-things arcade concept is the whole aesthetic: an arcade is loud, with flashing lights, giant and sometimes lurid artwork on the machines. I think if you were able to make some machines that gave a high quality experience without all that side of it, you might be able to install them in other semi-public spaces: airports, train stations, shopping malls, basically anywhere you currently see things like massage chairs.
That said, maintenance is for sure a concern. The state of most public pianos does not inspire confidence.
In Codona's Amusement Park in Aberdeen in the late 90s, there was a Ridge Racer "cabinet" with three massive rear projection screens and an ACTUAL REAL MAZDA MX5 TO SIT IN.
WHAAAAAAAAAT
Seriously insane levels of money-no-object zero-fucks-given design.
This is painful to me on three levels:
1. Real estate costs have gone up so much it’s prohibitively expensive to do something this grand.
2. Advertising is now a race to the bottom where showing car ads on websites has almost zero cost with all return compared to something novel like this.
3. It’s impossible to find a car like a 90s Miata these days because manual transmissions are almost dead and every car had to get heavier to have enough safety features to survive being T-boned by a Cybertruck.
Agree on the rest, but thankfully for #3 a modern base ND Miata with the 1.5 is pretty close to in weight to a NA due to a lot of weight saving work by Mazda.
It doesn't make assumptions, it tries generate the most likely text. Here it's not hard to see why the mostly likely answer to walk or drive for 50m is "walking".
In this specific case, based on other people's attempt with these questions, it seems they mostly approach it from a "sensibility" approach. Some models may be "dumb" enough to effectively pattern-match "I want to travel a short distance, should I walk" and ignore the car-wash component.
There were cases in (older?) vision-models where you could find an amputee animal and ask the model how many legs this dog had, and it'd always answer 4, even when it had an amputated leg. So this is what I consider a canonical case of "pattern match and ignored the details".
I recently had a bug where I added some new logic which gave wrong output. I pasted the newly added code into various LLMs and told it the issue I was having.
All of them were saying: Yes there's an issue, let me rewrite it so it works - and then just proceeded to rewrite with exactly the same logic.
Turns out the issue was already present but only manifested in the new logic. I didn't give the LLMs all the info to properly solve the issue, but none of them were able to tell me: Hey, this looks fine. Let's look elsewhere.
I have genuinely considered if my (and perhaps everyone on hn) life calling should be just to make a better touch keyboard.
Bearing in mind the amount of constant pain and torment the current best keyboards inflict upon the world, can there be any more urgent problem to tackle?
Forget climate change guys. Make a keyboard. Save the world.
There already was one. For all its faults, Swype for Android ~12 years ago was better than any swiping keyboard available now - both in its interpretation of swipes, and in its other features like "squiggle over one letter to indicate a double letter", "run your swipe above the keyboard to indicate a capitalized word", and the incredible editing features (it had an editing keyboard!). The Swype key, located in the bottom left? You could Swype-A to select all, Swype-X to cut, Swype-C to copy, Swype-V to paste. Swype-space brought up the editing keyboard.
Sounds great. When you say 12 years ago do you mean newer versions weren’t as good? Or was it discontinued or something?
I think the really puzzling thing about these keyboards is we all seem to remember typing being easier - even though phones back then were so much smaller. It makes no sense!
I still use it on iOS, and I've tried to remove all other keyboards, but Apple still just seems to "make up" keyboards I don't know are installed. Or switch keyboards on me mid-typing a word to a weird native one I also don't show as installed. It used to be very occasionally this would happen but now it's so repeatable since 26 I can almost not use my keyboard.
One caveat, I have an Icelandic keyboard installed on there. Sometimes web controls will force an input box to a US english keyboard (or numpad), which is annoying but at least that's sort of covered by a spec. What really drives me nuts is when I'm mid typing on the swype keyboard and suddely it switches to a completely square grid keyboard with up and down quotes in the autocomplete (which is not actually autocompleting or correcting(which while technically correct has almost completely fallen out of popularity since the dawn of the internet)
I quit using Android ca. 2016 (not because I hated it, had other work-related reasons why) and Swype for iOS was only around for a brief period before Microsoft killed it (and most of the things that made it great).
I never used the much-vaunted tap-only iOS keyboards of the earlier iPhones. I have large hands (the OG Xbox Duke controller was very comfortable) and typing on those small screens always felt painful even though I was so often told it was great.
Counter argument just to play devil’s advocate. Is that forming LLMs into useful shapes could become the game. If it turns out to be impossible to build a real moat around making LLMs - like maybe China or just anyone will ultimately be able to run them locally / cheaply, then the game of spending a billion dollars training one is much more risky
Just been reading about the crash they’re talking about in the article - it seems like a kid walked out from between two parked cars.
Rather than being a bad thing, this is probably Waymo saving his life.
It says the car reduced speed from 17mph to 6mph before contact. This is the kind of reaction-speed safety an AI car should have over a human driver - instead it’s just ‘waymo hit a kid’.
Do we know if there was space to swerve? I'm wondering if there was a vehicle next to the car, would swerving into it help slow down the Waymo or would it cause lose of breaking power and end up hitting the person at a faster speed?
A human doesn't have time to consider all that, it just seems to me that the instinctual move is to swerve. Hopefully the other cars are paying attention and swerve too.
Cars, no, but truckers have been damned for not just sacrificing themselves when their brakes failed. There was a recent incident where a truck had its brakes fail on its way into Denver from the Rockies and killed several people. They should have taken an off ramp, but IIRC, both in public opinion and in court it was argued that failing that they should have just committed suicide by going off the road before they hit a family.
There's a difference in expectation between the reactions of a person that has under 10 seconds to react and a truck driver that had minutes of time to plan, including bypassing an emergency ramp designed specifically for runaway trucks.
After passing the Genesee exit, Aguilera Mederos's truck began to smoke as he passed a runaway truck ramp, without taking it, and instead drifted into the left lane nearly clipping a white Chevy Silverado, and passed the next exit as well. For the next few minutes, Aguilera Mederos reached speeds upwards of 100 miles per hour (161 km/h)[6] and passed the next four exits.
Alternatively, maybe the car should have been driving more slowly. There is not enough information to know whether a human driver would have had a better outcome.
If I know kids are likely to be running around, I drive well below the speed limit.
I'm not saying a human driver would have done better. I'm disputing the claim that a human driver would have necessarily done worse. There is not enough information given to reach either conclusion.
