AI singer-songwriter ‘Anna Indiana’ debuted her first single ‘Betrayed by this Town’ on X, formerly Twitter—and listeners were not too impressed.
AI singer-songwriter ‘Anna Indiana’ debuted her first single ‘Betrayed by this Town’ on X, formerly Twitter—and listeners were not too impressed.
I put “AI” in scare quotes specifically because I do not believe we are having an “AI revolution”. These are not AI.
I think AI can exist but that’s not what we have right now. What we have are jumped up algos that can somewhat fake it.
Even those future “real” AIs are going to be taking in human input and regurgitating it back to us. The only difference is that the algorithms processing the data will continue to get better and better. There is not some cutoff where we go from 100% unintelligent chatbot to 100% intelligent AI. It is a gradual spectrum.
I believe a real AI would be able to generate its own inputs without humans to give it input. It would have an actual subjective experience, able to actually imagine new things with zero external inputs. It could experience the redness of the color red.
Is this how you see human intelligence? Is human intelligence made without the input of other humans? I understand that even babies have some sort of spark before they learn anything from other people, but dont they have the human dna input from their human parents? Why should the requirement for AI intelligence require no human input when even human intelligence seemingly requires human input to be made?
Sorry, lots of questions, just food for thought I suppose.
The very fact that “babies have some sort of spark before they learn anything from other people” shows there’s something missing.
I think intelligence requires the ability to think about your own thoughts and then draw new conclusions. LLMs can’t do that.
Yeah, to be clear, I’m not arguing that current LLMs are as creative and intelligent as people.
I am saying that even before babies get human language input, they still get input from people to be made, the baby’s algorithm to make that spark is modled on previous humans by the human data that is DNA. These future intelligent AIs will also be made by data that humans make. Even our current LLMs are not purely human language input, they also have an algorithm that is doing stuff with that data in order to show to us its, albeit relatively weak, “intelligent spark” that it had before it got all that human language input.
Chatbots are not new. They started around 1965. Objectively, gpt4 is more creative than the chatbots of 1965. The two are not equally able to create. This is an ongoing change, in the future AI will be more creative than today’s most creative AIs. AI will most likely continue on its trajectory and some day, if we dont all get destroyed, it will eventually be more intelligent and creative than humans.
I would love to hear an rebuttal to this that doesn’t just base its argument on the fact that AI needs human language input. A baby and its spark is not impressively intelligent. What makes that baby intelligent is its initial algorithm plus the fact that it gets human language data. Requiring that AI must do what the baby does without the human language data that babies get makes no sense to me as a requirement.
Without humans to curate the inputs and outputs, LLMs hallucinate and go insane. I think this is the precursor to creativity, but they need the ability to curate themselves (i.e. the ability to think about their own thoughts) before I’ll call them intelligences.
Yeah, you are definetly onto something there. If you are interested in checking out the current state of this, it is called “AutoGen”. You can think of it like a committee of voices inside the bots head. It takes longer to get stuff out, but it is much higher quality.
It is basically a group chat of bots working together on a common goal, but each with their own special abilities(internet access, apis, code running ability…) their own focuses, concerns, etc. It can be used to make anything, most projects now seem to be focused on application development, but there is no reason why it can’t be stories, movie scripts, research papers, whatever. For example, you can have a main author, an editor that’s fine-tuned on some editing guidelines/books, a few different fact checkers with access to the internet or datasets of research papers (or whatever reference materials) who are required to list sources for anything the author says(if no source can be found, then the author is told by the fact checkers and they must revise what they’ve written) and whatever other agents you can dream up. People are using dwsigners, marketers, CEOs… Then you plug in some api keys, maybe give them a token limit, and let them run wild.
A super early version of this idea was ChatDev, if you don’t want to go down the whole rabbit hole and just want a quick glimpse, skip ahead to 4:25, ChatDev has an animated visual representation of what is happening. These days AutoGen is where it’s at though, this same guy has a bunch of videos on it if you are looking to go a bit deeper.