> Selling a toaster has an implicit warranty of merchantability. Society expects that if you sell me something, it should have certain promises. Yes, there’s no monetary exchange here, the work is given gratis, but there’s still a relationship and an interaction here and I think it is clear some people, like myself, believe that there are implied expectations.
> However, I strongly believe that when you create something for people to use that there’s an implicit social contract about how to go about doing certain things.
Qwen models are always good. The 35B A3 model is a MoE model which means it has higher performance in RAM constrained environments compared to the 27B dense model (which is better at coding).
I don't have experience to rate it's Hebrew or Greek performance but apparently it's not bad.
> I think the inability to see the freedom AI gives people is one of the saddest things I've seen.
No one’s failing to see the good things, hypothetical or not. Most of us are aware just fine, we just don’t all agree that the negative trade-offs are worth it.
You should reconsider the scope. Nobody is losing sleep over Airbnb pics.
We’re in an era now where every image and video (and for that matter audio) is potentially fake; where knowing what’s real and true is no longer possible.
> We’re in an era now where every image and video (and for that matter audio) is potentially fake; where knowing what’s real and true is no longer possible.
This was always the case. Spin and propaganda are not new, the way it's conveyed has just become a bit easier. People are not as susceptible to misinformation as most assume, they recalibrate how much stock they put into the things they see based on the quality of the information environment. Basically everyone knows now that the internet has a low signal to noise ratio.
No, this really hasn't always been the case. At least not in the sense it is today.
20 years ago your misinformation came from television, radio, and print. All of those things were expensive to produce and there was an implicit need for them to be at least vaguely believable and reliable, because their existence depended on it to continually generate revenue.
- Today, a single person produce 100% AI-generated media for basically the cost of their time.
- That media is as high quality as anything else out there.
- Social media platforms provide the delivery system for free.
- There's no real way to tell if there's even a real person behind the name/pseudonym used for posting it. It might be a person, it might be an algorithm, it might be a nation-state. You have no way to know.
At my core, I do believe it's a good thing, so don't try to frame things as if I'm ethically opposed to the overarching idea here.
It's that the powers-that-be have and always will have more resources to bring to bear than the individual will have to combat them. And right now we're moving at absolute break-neck speed to invent technologies that can be used undermine us individually as well as collectively at ease and scales heretofore never seen.
I believed for a long time that the information age would be the great liberator - the great balancer. But we're on the precipice right now of governmental and societal collapsem and it has everything to do with the massive proliferation and preponderance of either misinformation, or agenda-aligned (shaped) information.
Anti-vax, Antifa, ACAB, conspirituality, QAnon, etc., etc., have all had an enormously negative impact, and we're still in misinformation infancy, and lucky that the leadership in goverment is so old that they're not particularly great at bending these technologies to bear on us. But that's not always going to be the case.
Sorry, but none of the factors you mention are particularly important IMO. At worst they cause a temporary blip that adversely affects some people before they recalibrate their expectations of the information environment they're in. People are simply not as vulnerable to this stuff as the chicken littles crying about misinformation think they are. All of the failure modes of media that you name have happened before, and people adapted.
I enjoy the technology too, but the tradeoffs are pretty grim. It takes stepping outside of my bubble to see it in full force, but AI misinformation is already rampant.
Anthropic explicitly state that they don't do this, even if you use the free plan and even if you don't opt-out of letting them use your data for training:
That answers for the "sold" part but not for the "used" part.
I.e. nothing about this statement prevents Anthropic from running ads within Claude, as long as they run the ad-placement auctions themselves, and so aren't leaking any of the data they're using to decide which placements are relevant to which users+sessions. (This is the same thing Google does for SERP ad auctions.)
But actually, and perhaps more interestingly, nothing about this statement prevents Anthropic from building a Google AdSense competitor either. Other sites (or mobile apps, etc) could plop in an Anthropic ad iframe; and it'd be Anthropic's knowledge of your interactions with Claude that would drive what ads would show up in that iframe. The embedding site doesn't know what ads the users are seeing, so that's still not "selling users' data to third parties", per se.
> Plus, there are no RLHF signals in OpenRouter data. Even if OpenRouter wanted to build a general model-neutral framework for collecting RLHF-type data, it can't force subscriber apps to do the UI-level stuff necessary to collect it (i.e. the things ChatGPT/Claude do, with "thumbs-down" buttons, A/B tested responses, etc.)
The majority of RLHF data doesn't need this. The majority is software development and/or tool calling where the agent gets a signal back as to if it succeeded (eg compilation errors, test errors). It's true that end-of-trajectory signals (eg, did this task do what you wanted) are even more useful but even partial signals are great for RL training.
Parallel *Re*construction is a play on words I wrote related to a lot of the nuance at play I wasn’t able to cover in the blog without making it very long.
Indeed. Parallel construction is when law enforcement doesn't want to burn their source or their source is unlawful, so they find another way to justify a warrant or prosecution even tho their initial justification comes from a different source.
This has been upheld as lawful, but also can unfortunately be used to disguise unlawful behavior.
Reverse engineering is not related to parallel construction.
Parallel construction would be if LE unlawfully intercepted transmissions using this technique and discovered a crime, then found other unrelated evidence to begin an investigation of that crime.
No there isn't.
Pay money and there's a contract.
Anything else is in your head.
reply