Scaling laws in terms of loss are already starting to bend, as seen in published model cards from multiple companies.
Not researched Checkable, but outside the sample that was researched.
Factual: Statements about the present or the past, checked against the record: did they hold up?
Said for an organisation they lead or work for; it counts on both records.
Cite this claim
Dario Amodei, “Scaling laws in terms of loss are already starting to bend, as seen in published model cards from multiple companies.” (said 8 August 2023). Verdict: not researched. Reliability API, claim c_904c1a0d8dc1, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_904c1a0d8dc1
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Dario Amodei
-
Mostly true
“the amount of computation that goes into the models doubles every four months.”
The amount of computation going into AI models doubles every four months.
-
Mostly true
“Distillation of frontier models allows lagging companies to narrow the gap using a fraction of the cost it would take to develop their own AI independently.”
-
Mostly false
“I thought it was a very plausible blocker one or two years ago.”
-
Mostly true
“Importantly, because this type of RL is new, we are still very early on the scaling curve”
-
Resolved true
“I'd guess the number today is maybe ~4x/year.”
The current rate of algorithmic curve shift is maybe about 4x per year.
Verdicts are made automatically; how verdicts are made.