This was literally always where "AI" was going
-
RE: https://typo.social/@colinmford/116613158941426952
This was literally always where "AI" was going
@mcc Dystopian science fiction was far too optimistic.
-
RE: https://typo.social/@colinmford/116613158941426952
This was literally always where "AI" was going
@mcc oh my god i hate it
-
@mcc I wonder if this won't just affect output when asking "consumer-product" style questions but like code generation as well. This is amusing.
@zyd pay to get your 0day added to someone else's code
-
@mcc this shit cannot burn down fast enough
@astraluma @mcc my one quibble at this point is that this is not really the terminus. we will know AI will have reached its apotheosis when these token bids are conducted on the blockchain and have real-time algorithmic futures contracts on kalshi
-
@astraluma @mcc my one quibble at this point is that this is not really the terminus. we will know AI will have reached its apotheosis when these token bids are conducted on the blockchain and have real-time algorithmic futures contracts on kalshi
@astraluma @mcc do I need to say that this is satire and that google should not follow it like a recipe? pointless, I guess, since the kind of person who would need that warning has already had this idea themselves with no need of my help and is implementing it enthusiastically as we speak
-
@mcc ad bidders (on RTB exchanges, I don't think google supports this) have been bidding based on user context vectors for years and years, and using RL to do it. this is just the logical progression of that
@mcc I bet the way this actually works is ad bidders get an embedding of the context window and bid for an ad insert. your ad bidder wouldn't need its own LLM for that, you can do Q-learning with the embeddings as your state vectors
-
@mcc I bet the way this actually works is ad bidders get an embedding of the context window and bid for an ad insert. your ad bidder wouldn't need its own LLM for that, you can do Q-learning with the embeddings as your state vectors
@mcc the reason I don't think advertisers will bid on outputting a single word at a time is if you have multiple bidders the output will be incoherent
-
@astraluma @mcc do I need to say that this is satire and that google should not follow it like a recipe? pointless, I guess, since the kind of person who would need that warning has already had this idea themselves with no need of my help and is implementing it enthusiastically as we speak
@glyph @astraluma @mcc Maybe you should preface it with a catchy title like "Please, for the love of all that you hold sacred, do not build the Torment Nexus"
-
RE: https://typo.social/@colinmford/116613158941426952
This was literally always where "AI" was going
@mcc one question that pops in my mind when reading these is: was there, at any point in the conception of these ideas, someone to ask "but is that what our users want"?
-
RE: https://typo.social/@colinmford/116613158941426952
This was literally always where "AI" was going
-
RE: https://typo.social/@colinmford/116613158941426952
This was literally always where "AI" was going
@mcc alright, well, it's local models for me forever I guess.
-
@mcc I don't see this as worse than keyword auctions, tbh
-
@mcc alright, well, it's local models for me forever I guess.
-
P pelle@veganism.social shared this topic
