TalkBass has been independent since 1998. Add your voice.
Create a free account to reply to discussions, view embedded media, and browse with fewer display ads.
Join free Log in
Want zero display ads or expanded classifieds tools? Compare plans.
see post #32My company is going full bore AI. I'm retiring in a year, and avoiding most of it. Getting out just in time.
Meta just announced the layoff of 9000 workers due to gains from AI.
It's coming.
That was a bit garbled. I'm laid up with a broken leg and pain meds. Hopefully, people can make some sense of it.One of my tests is to ask about the saxophone solo for The (English) Beat. It's one of the few without any saxophone at all. Last time I tried it, about three months ago, the chat bots waxed lyrically and confidently about the solo.
Our issue was frequently finding good quality data, or ways to take mediocre data and somehow filter it to improve it.AI or people learning from bad information end up regurgitating bad results based on the GIGO theory.
This is exactly what's driving the data grab scenario.Our issue was frequently finding good quality data, or ways to take mediocre data and somehow filter it to improve it.
The only time we tried to use internet data was when bored and considered making an arse/elbow discriminator (UK expression: doesn't know his arse from his elbow) but we discovered that getting training material without losing our jobs was going to be challenging. So that project only lasted an afternoon.
And AI scrapes data a lot of which has been generated by AI anyway.AI or people learning from bad information end up regurgitating bad results based on the GIGO theory.
Correct, training on your own bad data that's then used to train itself to improve accuracy is just like eating your own babies...And AI scrapes data a lot of which has been generated by AI anyway.
So in a sense it's feeding on its own excrement.
Thus the models get gradually worse over time
The whole system will crash as there is no route to profitability given the amount of debt taken on by AI companies
Luigi.
That brings me to my next point. I started my pedal board AI experiment because I am a historian by trade and wanted to test it out with something I would be able to see red flags with fairly quickly. Wrong pedal specs, incorrect features etc... What I have concluded, up to this point, is that Ai requires a semi-knowedgable, inter-active operator. Simply asking it to provide a history of X is probably gonna generate a pretty shakey answer.
I used to use Google Scholar quite a bit; it now has an AI add-on ("Labs") which I haven't yet used for any questions I'm actually interested in, but it does at least provide links to the papers cited, and provides summaries, of what sort of quality I'm uncertain. At first blush it looks reasonably promising, though.It's pretty interesting to decouple the mainstream PR machines from reality.
ChatGPT is full of hot garbage.
Gemini has been helpful to find source material faster, kind of like using old Google in a curated fashion, and then chase its cited sources for actual facts you can verify.
In this manner, Gemini has not been garbage in garbage out but filtering click bait ad infested sites and getting me to the relevant pages of online manuals in record time.
Best name ever.And that right quick.
Earth was like that in the future world of The Expanse series.You'll find, if you wanna have a look, that they have a plan for this, much like kings of old.
They plan on a form of basic income and minimal housing for the masses, based of property taxes they want to define.
So basically, they're trying to reduce their cost exponentially.
Sam Altman even wrote a manifesto about this, called Moore's Law For Everything... And it reads like a dumped down version of the communist manifesto - minus all the pesky democratic control espoused by Marx, of course.
![]()
Butlerian Jihad
Last Jihad redirects here. For the book by Sumer and Kautman, see The Last Jihad. "We must negate the machines-that-think. Humans must set their own guidelines. This is not something machines can do. Reasoning depends upon programming, not on hardware, and we are the ultimate program! Our Jihad...dune.fandom.com
And AI scrapes data a lot of which has been generated by AI anyway.
So in a sense it's feeding on its own excrement.
Thus the models get gradually worse over time
The whole system will crash as there is no route to profitability given the amount of debt taken on by AI companies
Yep- I saw this terrible AI-written post reposted on a Facebook bass group recently that misattributed Larry Graham's "slap story" to a broken amp. I searched for that story and was directed back by Google to basically a carbon copy of that same story in a Facebook post a few months back as the "source" for that info.Correct, training on your own bad data that's then used to train itself to improve accuracy is just like eating your own babies...