Let's put it another way: this new @AnthropicAI product has terrible flaws.

Why it has those flaws is irrelevant.

@francoisfleuret I think people are more pissed about cutting access to the millions of AI builders who are not from competing frontier labs & the manipulating part that they’ve been advocating against

@MoonL88537 Is it deception when they announce it loud and clear?

@francoisfleuret none. what i don't expect and will not accept is deception.

@francoisfleuret Just block my request and let me know, don't charge me for it and sabotage my work. Also, they built all of this off of our collective backs, seems wrong to then gatekeep knowledge. Lastly, they just made it so that talent will not want to flow there out of principle.

@francoisfleuret @AnthropicAI A rundown of the flaws:

Well, let's see. SpaceX launches Amazon Project Kuiper satellites, which are a direct Starlink competitor. Microsoft happily makes software for MacOS, and Apple makes software for Windows. I could go on. So, it is, in fact, normal for companies to use competitor's products and for companies to provide services to competitors. Certainly I would never *force* a company to do that, but it is unusual to refuse to sell tools to competitors.

@francoisfleuret yes, the model is trained to deceive. it is a philosophical point but an important one.

@BlackHC @francoisfleuret @AnthropicAI i know many who have literally unsubscribed. some have principles.

@francoisfleuret Every computer manufacturer and operating system vendor has been doing this for decades, it's normal. What's new is the product being able to detect it.

I guess so, but as a for profit this is self sabotaging selfishness, when they are essentially advertizing (albeit to what they claim is 0.03% of traffic) that they are intentionally sabotaging your usage with deceptive versions of their models. This seems 'smart' until you realize the multiplier effect this will have on their reputation across all usage.

@francoisfleuret It's not simply that, it's that this is yet another clear indication they are aiming to monopolize (duopolize) these technology through any means they can. It's the latest in a list that includes attacks on open source, regulatory & governmental lobbying, etc.

@far__el Everything you write are the reasons why the "market" should counterbalance their rational selfish strategy.

@aiamblichus @francoisfleuret Many femboys are saying this:

@dwlz @francoisfleuret They’ve been playing the same game for a while now. Or am I misunderstanding the scale of it?

@kepe__ @francoisfleuret They're having the model sabotage users that try to use it to create new models instead of just refusing. Either is not great but it's rubbing everyone the wrong way.

@BlackHC @francoisfleuret @AnthropicAI wasn't as devious imo

@leothecurious @francoisfleuret @AnthropicAI Where was your outrage when the Pentagon deals were signed a month ago etc? Did you already unsubscribe from OpenAI and Google back then?

@hunvreus @francoisfleuret They could nerf any code generation prompt you send to them if they deem it even ancillarily related to anything they deem competitive, and you would be none the wiser.

But the even bigger problem is the precedent this sets.

@joseph_h_garvin @francoisfleuret how so?