Perhaps GPT-4 is largely marketing and hype. It’s nice and all - yet over sold. More over-sold is the ‘speed’ or rate of innovation.
Again, AI’s (I still call it ML) greatest threat to people in the near term is hyper-targeting and malicious actors who can now mass produce human-enough content cheaply - that for all intensive purposes passes enough of a turing test to appear human generated.
Our brains effort required to critically dissect and identify as ML/Bot created has crossed a significant barrier - for anything we read now, it takes say two times as long to evaluate it’s origin/intent/source. Previously it was a fraction - easy to identify compared to the content itself.
Death by 1,000 cuts is the new reality. Not entirely new, as mass media already does this; however they largely failed at digital (poor revenue). Now it’s possible to make money other ways through similar manipulation digitally - again, with hyper-targeting.
The only way I know of adding a cost to the targeted content is (burnt) proof of work. Seems wasteful - yet I don’t know another way to combat the catastrophic drop in price to generate ML. Humans would pay this cost too - yet humans don’t mass produce, so it ideally should mostly negatively impact mass generated content.
https://archive.fo/FlBgV
wakoinc / Blake
npub1kt…rmrvj
2023-07-05 15:46:53
Author Public Key
npub1ktw5qzt7f5ztrft0kwm9lsw34tef9xknplvy936ddzuepp6yf9dsjrmrvjPublished at
2023-07-05 15:46:53Event JSON
{
"id": "7e3820cea826f3e4c1d97f528e1872a4e8b212b59d19b6bad1949863d6f0bb9a",
"pubkey": "b2dd40097e4d04b1a56fb3b65fc1d1aaf2929ad30fd842c74d68b9908744495b",
"created_at": 1688572013,
"kind": 1,
"tags": [],
"content": "Perhaps GPT-4 is largely marketing and hype. It’s nice and all - yet over sold. More over-sold is the ‘speed’ or rate of innovation. \n\nAgain, AI’s (I still call it ML) greatest threat to people in the near term is hyper-targeting and malicious actors who can now mass produce human-enough content cheaply - that for all intensive purposes passes enough of a turing test to appear human generated. \n\nOur brains effort required to critically dissect and identify as ML/Bot created has crossed a significant barrier - for anything we read now, it takes say two times as long to evaluate it’s origin/intent/source. Previously it was a fraction - easy to identify compared to the content itself. \n\nDeath by 1,000 cuts is the new reality. Not entirely new, as mass media already does this; however they largely failed at digital (poor revenue). Now it’s possible to make money other ways through similar manipulation digitally - again, with hyper-targeting. \n\nThe only way I know of adding a cost to the targeted content is (burnt) proof of work. Seems wasteful - yet I don’t know another way to combat the catastrophic drop in price to generate ML. Humans would pay this cost too - yet humans don’t mass produce, so it ideally should mostly negatively impact mass generated content. \n\nhttps://archive.fo/FlBgV",
"sig": "e11c0d8c16334c934a5c50eb4cabd5c38d393f55e94f30bbf74e2a407fb268784134153d241cc1ccbe19f3a66df7832ed8f27f7803ac4ba344183fdc96b5a51c"
}