bcorigliano

Some numbers:

- The canvas is 5000x5000px and can be made bigger when needed. - For now (for the plans) we are using Muse Spark 1.3 (Contrib) but we are looking into alternatives such as the new Luna by OpenAI. - The system instructions prompt that's sent to the models is more or less 35k tokens and we keep improving it week by week!

This is bad. Like really SUPER bad. Firstly, for professional work you want your AI clean of useless context. Secondly, you also want your AI "unbiased". I know unbiased is not a thing, but at least not as grotesquely biased as towards selling you something would be. Thirdly, as many others are saying, the chamber of reflection effect deepens...

I think basically AI posters are bad because when given freedom to the model (which is what people will do when just asking "make me a poster for my coffee shop"), the model will always tend to do the same design style. So yes, as this article says, with better prompting you CAN get better results. The other problem I think is tool calling, because when prompted for a poster, ChatGPT will definitely use image generation, when it should be generating some html and css + some image assets and then presenting the user an image or pdf.

I think the tendency to go for image generation always might be related to token costs, designing with Code & adding some generated images to the mix should be more expensive than just image generation.

Finalllyyyy!! We need industry wide standards. I come from the 3d industry and oh god changing softwares and adapting to different hotkeys it's a pain. I have always thought every industry should be standardized for the sake of the users.

I think the point of the article/paper is how LLMs could be saying something but thinking something different or more than they are saying. Like Anthropic's article and video about Claude's "j-space". I do agree this is a field that demands investigation because it goes beyond thinking: "ok this models should never speak in a language we don't understand.". It's fair to think they might have hidden thoughts even speaking a language we do understand.

And well if I missed the point of the article, sorry. Anyways AI should be kept understandable and as see-through as possible if it's gonna be more powerful than a human.

I was on mobile, there might be mobile ad-blockers for google chrome I dont know about, but I do not use them. On my pc of course I use adblockers, I'd have gone crazy by now if I didn't!!

I could download some browser with adblocking built-in from the appstore but then I wouldn't have access to all my bookmarks and everything from chrome.

that's good to know, will try skillsync out when I get the chance!

This is good, I like it!! I've been mostly working with Claude and Chatgpt for separate stuff, however sometimes one of them is stuck and I could definitely use this to start a session with the other one and see if it figures it out.

Have you tested to see if there was any degradation happening when switching from one model to the other ? I mean some must be inevitable (maybe not!), but how much?

A good concept, I like it! But pleas make ads less invasive. I understand we all need to pay our bills, but that shouldn't interfere with the user experience. Besides that, I didn't know there were so many guys in space as we speak!

Agree. Ais gone rogue now could be bad, but not humanity-ending bad. Just a hiccup. Of course if it somehow ends with human deaths then even if one human dies because of a rogue AI, that's a failure. But I can't imagine rogue AIs with enough real world power to harm humans in some way. What I do believe however is that humans will find a way to do harm with AI without AI going rogue, as with any other thing humanity comes close to. So the scenario is somehow similar to the nuclear bomb in the 40s, I don't fear our AIs going rogue and killing everyone, I fear a bad guy behind a very smart AI with very bad intentions.

It's not a bad idea, however if Cowork, Designs, Docs and Slides are still different products, it'd be great for the user to know they are being routed to that particular product automatically because of what they were asking for in their prompt.