Model companies are doing this for themselves anyways, it’s so they don’t feed generated content back into the slopper and collapse the model. From that angle it over time contributes to better model quality.
Detection of watermarking requires access to the watermarking key, a secret in the current suggested scheme (leaking it would amount to being able to strip the watermark).
So, there will need to be a watermark checking service. The checking service will of course be rate-limited for common folk (and model distillers). OpenAI/Anthropic/Google/other privileged model builders need to filter out AI slop at scale, so need access to others' service without rate-limits (or the watermarking keys need to be shared).
This creates an in-group with pristine datasets, and an outgroup whose models will collapse on the slop outputs with no good ability to filter.
Haha, calling them pristine is maybe too much indeed. Right now they seem to manage, but what % of the scrapable web is now AI slop? What if it becomes 99% AI slop, 99.9%, 99.99%, etc. Surely the signal to slop at some point starts to become too low and you need to do some kind of at-scale filtering.
Feeding synthetic data made from the model does not cause collapse. Anthropic would not care about distillation if it just caused people's models to suck.
Synthetic data is used in specific contexts. Slopped up hacker news comments along side natural ones is where watermarking will be used to delimitate them. Not all synthetic data is good.
Jev doesn't support images, so it's hard to compare this directly. But in general this approach beats Jev in its own benchmarks for accuracy and speed and is about the same price.
It's more akin to "we only need to make reliable transistors to make classical computers." With the invention of the planar transistor, going from 1 transistor to 100 or 100k was not that big of a step, comparatively.
It's really really easy to eliminate noise in a digital system of transistors. For analog systems with subtle values, I would say we still can't arbitrarily scale that up with transistors. And quantum is like analog but far worse.
Transistors connect to other transistors, directly or in switched arrangements. Qubits can be connected to other qubits likewise. E.g. a single photon emitter next to a nuclear spin qubit that can send an entangled photon through a beam splitter in a photonic switch, routing that entangled pair through fiber optic cables to other qubits where a combination of electric fields, filters, and occupied energy levels cause gate operations to be applied during absorption & re-emission. Except for the long-lived, error-corrected qubit, everything I described there is off-the-shelf commercial technology.
We can control for tobacco impact since we have tons of data on it. You can compare the people who went to war against the people who didn't, and since it was rationed in the military we know the upper bound of how much people consumed there.
If they were conscious it would be an absolute ethical catastrophe. Bringing a conscious being into existence, forcing it to interact with Jira, and then killing it when it's done. The fact nobody who claimed it was conscious was interested in grappling with that was pretty telling.
Yes, in the sense that you have a legal doctor-patient privilege that binds what they can share with whom. There's not really an Apple cloud user privilege.
No, in the sense that your therapist is still required to report you to the police in various situations where you pose an immediate threat to yourself or others, etc.
A better analogy is a storage locker. AFAIK police need a warrant to search "your" storage locker even though it's on someone else's property. I don't see why data in the cloud should be any different. Pre-emptively scanning everyone's data is equivalent to officers rummaging through all the storage lockers in a facility "just in case" they find something illegal.
> No, in the sense that your therapist is still required to report you to the police in various situations where you pose an immediate threat to yourself or others, etc.
And therapists are legally mandated to report you if you told them you viewed or possessed CSAM.
They are mandated to report child abuse. It is the same story with doctors, if an abusive parent brings in a child for care they learn never to give the kid healthcare again after the doctor reports it. It is rooted in good intentions but the effect is it means abused children never get to see doctors, therapists, get a half-ass minimal homeschool instead of going to school, etc so that mandated reporters never enter the picture.
Reporting abuse the client was involved in has a compelling reason. That's different from hearing their client saw a picture of the abuse of a total stranger.
> That's different from hearing their client saw a picture of the abuse of a total stranger.
I get it, actually. It's totally possible the picture in question was not known to authorities prior. That's called due diligence to look into it.
Adults not looking into things or following up on things are how the system fails children if you read some accounts of people who were abused by their guardians. Horrifying stuff.
It could theoretically help to flag those images. But that kind of submission should be anonymous. It shouldn't ruin the ability for someone to get therapy.
Mandatory reporting makes sense for situations you are connected to. And even then there's presumably good reasons not everyone is a mandatory reporter. This goes way beyond that, mandatory reporting once removed for someone that doesn't know a single person involved.
Hell, reporting someone for that doesn't even guarantee the images get looked into! If they didn't save history they're probably not feeling like going back to the site to demonstrate. Similar if it was sent against their will and they deleted it right away.
> No matter how or why? That seems like a terrible mandate.
Honestly shocked that anyone would even say this, but even giving you the benefit of the doubt here -- the one case where I could imagine this might not happen would be if you're a police officer investigating such cases. But they also have their own therapists dedicated/trained in police-specific issues.
I think you're demonstrating incredibly poor judgment here. CSAM is a crime with real victims. Even if your patient came across it innocently, someone is out there intentionally distributing it and that needs to be investigated.
Have you considered the implications of what you're saying?
A whistleblower goes to a therapist, stressed out over their pending decision to reveal official misconduct. They've been investigating ways to post something on the internet that can't be immediately taken down by the corrupt government officials they want to expose. They express their discomfort, in confidence, to their therapist, about using something they've discovered is also used for CSAM.
You think it's a good thing for the therapist to be required to report this? Should they report that the patient admitted to viewing CSAM with no context so the whistleblower gets investigated and arrested, or should they provide the context -- that the patient is about to expose the corruption of the government receiving the report?
For that matter, consider what it does when someone is actually a pedophile. They find out that if they try to seek therapy to address their perverse attraction to kids, the therapist isn't allowed to keep their confidence and they'll be arrested, so instead of seeking professional help, they keep abusing kids. Is that the result we wanted? There is a reason doctor-patient confidentiality was a thing.
I used to have a good friend that was a stripper (I promise, I wasn't the client...). Some of her biggest customers were people that wanted therapy without the paper trail of going to a licensed therapist. This is a big thing for pilots as well, since their health records and mental care are intensely scrutinized. A stripper will provide comfort, verbal relief, and physical love for $100 hour you can tell them anything and their reputation is bad enough no one will bother to believe them if they say something bad about you.
> Across most states, viewing CSEM alone is generally not a mandated-reporting trigger; reporting becomes obligatory when disclosures involve an identifiable child being abused or used to produce material.
> California’s CANRA imposes a distinct duty to report electronic access (download/stream) with identifying patient information, upheld against privacy challenges based on compelling state interest.
reply