haxor@derp.fooMB to Hacker News@derp.fooEnglish · 1 year agoGPT-4 is getting worse over time, not bettertwitter.comexternal-linkmessage-square2fedilinkarrow-up14arrow-down10file-textcross-posted to: technology@lemmy.worldhackernews@lemmy.smeargle.fanstechnews@radiation.partyhkns@lemmy.amyjnobody.com
arrow-up14arrow-down1external-linkGPT-4 is getting worse over time, not bettertwitter.comhaxor@derp.fooMB to Hacker News@derp.fooEnglish · 1 year agomessage-square2fedilinkfile-textcross-posted to: technology@lemmy.worldhackernews@lemmy.smeargle.fanstechnews@radiation.partyhkns@lemmy.amyjnobody.com
minus-squareBitSound@lemmy.worldlinkfedilinkEnglisharrow-up2·1 year agoI doubt it, purging any trace of the training data wouldn’t affect the existing performance. I would guess that it’s more something along the lines of tweaking it to use fewer resources, which would probably reduce the cash burning significantly.
I doubt it, purging any trace of the training data wouldn’t affect the existing performance. I would guess that it’s more something along the lines of tweaking it to use fewer resources, which would probably reduce the cash burning significantly.