"just one more model until GPT-5"

ICYMI - Years since OpenAI released the groundbreaking GPT-4,

then ever since they’ve been releasing a ton of “more efficient” horrible lower-skill models, refusing to even start training of GPT-5, for years,

But now FINALLY GPT-5 is near

All expecting a HUGE step forward with GPT-5, in large part just due to the primary “bitter lesson”

“bitter lesson” = model IQ scales with the log of the input compute resources, all else being equal, and no amount of tricks is enough to bypass the fundamental need for huge input resources

— and GPT-5 will FINALLY be a new model that they put BIG compute resources into training,after refusing to do so for all these years since GPT-4

= GPT-5 expected to be a massive step forward, finally

Massive AI wave inbound

🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶

So we’re back?

Or it’s so over?

🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶

Combined M2, i.e. “liquidity”

Man, gotta run a lead-lag correlation plot on these, because they look correlated af

🄳🄾🄾🄼🄿🄾🅂🅃🄸🄽🄶