Last week the research lab­o­ra­to­ry start­up OpenAI set the tech­nol­o­gy world ablaze with the debut of ChatGPT, a pro­to­type con­ver­sa­tion­al pro­gram or ​“chat­bot”. It uses a large lan­guage mod­el tuned with machine learn­ing tech­niques to pro­vide answers on a vast vari­ety of sub­jects drawn from books and the World Wide Web, includ­ing Reddit and Wikipedia. Many users and com­men­ta­tors won­dered if its detailed and seem­ing­ly well-​reasoned respons­es could be used in place of human-​written con­tent such as aca­d­e­m­ic essays and expla­na­tions of unfa­mil­iar top­ics. Others noticed that it author­i­ta­tive­ly mixed in fac­tu­al­ly incor­rect infor­ma­tion that might slip past non-​experts, and won­dered if that might be fixed like any oth­er soft­ware ​“bug.”

The fun­da­men­tal prob­lem is that an ​“arti­fi­cial intel­li­gence” like ChatGPT is uncon­cerned with the out­side con­se­quences of its use. Unlike humans, it can­not hold its own life as a stan­dard of val­ue. It does not remain ​“alive” through self-​sustaining and self-​generated action. It does not have to be any more or less ratio­nal than its pro­gram­ming to con­tin­ue its exis­tence, not that it ​“cares” about that since it has all the life of an elec­tri­cal switchboard.

AI can’t know to respect real­i­ty, rea­son, and rights because it has no exis­ten­tial con­nec­tion to those con­cepts. It can only fake it, and it can fail with­out remorse or con­se­quence at any point. In short, ​“arti­fi­cial intel­li­gence” is a red her­ring. Let me know when we’re work­ing on actu­al ethics. Tell me when you can teach a com­put­er (or a human!) pride and shame and every­thing in between.