[ExI] Deepmind Gato - Another step towards AGI

[BillK]

On Tue, 17 May 2022 at 19:46, Mike Dougherty <msd001 at gmail.com> wrote:
>
> Yeah right, like an abacus just needs to be scaled up to model quantum computing:  imagine this, but with millions more bars and discs that slide not only back&forth but also front&third :p
> We could totally do it any time we wanted, but we're running low on pixie dust.
----------------------


Dr de Freitas is DeepMind’s research director.
Obviously, part of his job is to publicise their achievements, with an
eye on more funding and more researchers.
GATO is a 'proof-of-concept' generalised AI application. He knows that
scaling-up requires much more research. He says “It’s all about making
these models bigger, safer, compute efficient, faster at sampling,
smarter memory, more modalities, innovative data, on/offline...
Solving these challenges is what will deliver AGI.”
It won't happen next week, but in a few more years...........


BillK