You are entitled to your opinion @Fruarse , but can you not agree that if ai at least detected patch points on the module graphics, you could implement an interface that would create a much more seamless experience in posting patch configurations?
-- singular_sound
I know I'm entitled to my opinion, otherwise why would I bother writing it down?
Also, what you understand as patch points on a module is nothing but an array of pixels to a digital computer. An algorithm by definition "does not know" like you and I do. Sure, you can maybe train a model to detect them on a manufacturer's basis, assuming all inputs and outputs share the same graphical properties... But it would still need to be supervised, which defeats the whole purpose.
But what happens when some panels have very fine details that are lost in the image upload? You need the resolution to resolve them in order for the training to be somewhat effective. What about different manufacturers, that maybe apply the same logic for input and output, but in reverse?
And, at the end of the day, is all this worth it just because you can't be bothered with spending that extra 0,5 seconds dragging the patch cable to the appropriate patch point?
I fundamentally disagree with the Chinese Room theoretical premise. I personally believe it is not even possible to create such a room without somehow also in some sense copying the relevant structure from a human brain, thereby implying that 'the room understands Chinese'.
-- singular_sound
If you believe a piece of silicon with some electricity running through it executing a predetermined set of instruction that can be reduced to the manipulation of digits is somehow/somewhat comparable to human intelligence... I warmly suggest you do your homework ;)
I just think a certain usefulness threshold has been crossed, and would personally rather get an instant reply from Claude 3.7, than a reply 2 days later, from, say, @zacksname .
-- singular_sound
You are assuming the computer programme to provide you with an accurate answer. Sure, maybe a real person's answer also wouldn't be, but at least a human "knows" what it is talking about and not just regurgitating words without knowing what they mean.
If perhaps only the Haiku model was used, all accounts were limited to one use per day or week, and the right kind of data sharing toward mutual advancement deal could be struck, this could likely be implemented nearly free with very little impact. Etcetera.
-- singular_sound
Who's going to pay for the training, hosting, electricity bills, deployment and upkeep costs? You?
I do not think it is possible to convince everyone to love our new AI overlords.
-- singular_sound
And rightfully so, your can keep your so called "AI overlords" to yourself ;)