DeepMind

Applied to Atari early by Multicore at 8mo

DeepMind is an AI research laboratory that was acquired by Google [sometime].in 2014. It is known for several record-breaking AI algorithms, often named with the prefix "Alpha", e.g. including AlphaGo, AlphaGo Zero, AlphaStar and AlphaFold.

AlphaDeepMind is a prefix used to denote a set ofan AI algorithms developed by research laboratory DeepMind,that was acquired by Google [sometime]. It is known for several record-breaking AI algorithms, often named with the prefix "Alpha", e.g. including AlphaGo, AlphaGo Zero, AlphaStar and AlphaFold.

These algorithms have many architectural similarities and were developed by largely overlapping teams.

Is there any reason to think that this is actually a coherent family of algorithms and not just DeepMind's label for things they think are really cool?

9gwern1yI don't think it is: AlphaGo/Master/Zero hardly overlap with AlphaStar's architecture, training, or category of RL algorithms, and AlphaFold 1 used a modest sprinkling of RL - but none of its RL techniques were used in either AlphaGo or AlphaStar! It's branding. Or to be more specific, it's DM's label not for 'cool things' but for "things we think will set world-records in their field and which will change them forever".
8DanielFilan1yIn which case, I'm kind of against the existence of this tag. Maybe it could instead be a 'DeepMind' tag.
1Multicore1ySeconded, DeepMind seems like a natural tag to have given that we have tags for OpenAI, Ought, MIRI, etc.
8Ruby1yI have renamed and quickly rewritten the tag. Good work, people.
1plex1yGiven that this is now an organization, it should probably be under the Organizations heading rather than Other.
2Ruby1yGood call! I'll fix that in sec.
1plex6moReminder to do this, since it seems to have slipped through the cracks.
2Ruby6moActually done for real this time. Thanks for the reminder.