[FRIAM] AI possibilities

glen gepropella at gmail.com
Thu Apr 6 11:21:07 EDT 2023

What we really need to try is to have 2 APIs open, one to Bard and one to ChatGPT and *conduct* them to play together ... or maybe 4 pipes open so that Bard's output is part of the prompt, dovetailed with the conductor, for GPT and vice versa. Shirley someone out there is submitting Bard output to GPT and vice versa. Right? I've been living under a rock lately. So I wouldn't have seen social media posts about it.

On 4/6/23 08:13, glen wrote:
> Off the top of my head, I can see 3 ways to get music out of the current chat interfaces:
> 1) algorithmic music - E.g. C programs like this:
> #include <stdio.h>
> int main(int t) {for (t=0;;t++) putchar((((int)(t/12)>>8&t) - (t<<4)) & (((int)(t/6)>>6&t) + (t<<2)));}
> The code I've gotten out of ChatGPT has been irritating. But I've never asked it to write something like that. Or maybe something in PureData or Common Lisp Music. Given the above program as a prompt, Bard gave me a slightly different one and confidently proclaimed that it was in a different key with some extra notes. But it's actually just a *fuzz* version of mine ... which even though Bard's gaslighting me, it's still a cool tune. 8^D
> 2) Time series. If you ask Bard to tell you what the next number in this sequence is, it'll tell you: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34. If I get the chance later, maybe I'll runs some other sequences by it and see what it can tell me. But there's no reason a next-token-predictor shouldn't be able to generate music straight out of the gate.
> 3) Notes as tokens, rather than signals/numbers as tokens. I'm sure such exists. But the closest I've come is https://www.w3.org/2021/06/musicxml40/ I don't see any reason why these machines couldn't compose MusicXML in the same way they can compose source code.
> On 4/5/23 22:15, Jochen Fromm wrote:
>> Yes, if a large language model is trained on all works of Mozart and contemporary artists like Haydn, it should be able to create a new piece of music which sounds almost like Mozart. Finally we can listen to Mozart's lost 28th piano concerto or Beethoven's missing 33th piano sonata o_O
>> -J.
>> -------- Original message --------
>> From: Prof David West <profwest at fastmail.fm>
>> Date: 4/5/23 1:55 AM (GMT+01:00)
>> To: friam at redfish.com
>> Subject: [FRIAM] AI possibilities
>> Based on the flood of stories about ChatAI, it appears:
>>    - they can 'do' math and 'reason' scientificdally
>>    - they can generate essays, term papers, etc.
>>    - they can engage in convincing dialog/conversations
>>      - as "therapists"
>>      - as "girlfriends" (I haven't seen any stories about women falling in love with their AI)
>>      - as kinksters
>>    - they can write code
>> The writing code ability immediately made me wonder if, given a database of music instead of text, they could write music?
>> The dialog /conversation ability makes me wonder about more real-time collaborative interaction, improv acting / comedy? Or, pair programming? The real-time aspect is critical to my question, as I believe there is something qualitatively different between two people doing improv or pair programming than simply engaging in dialog. I think I could make a much stronger argument in the case of improv music, especially jazz, but AIs aren't doing that yet.

ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ

More information about the Friam mailing list