Eu acho que esse é o David Gerrold Quando HARLIE era um, publicado originalmente em 1972 e republicado em uma versão mais polida e atualizada (do mesmo autor) que Quando HARLIE era um, solte o 2.0 no 1988 (versão 2.0 é a versão preferida do autor).
Na página do autor:
H.A.R.L.I.E. (Human Analog Replication, Lethetic Intelligence Engine) is an artificially intelligent machine. David Auberson, the psychologist responsible for guiding HARLIE from childhood into adulthood, struggles to understand his erratic behavior.
When humans begin vocalizing their wishes that HARILIE be shut down, he has to prove his existence and value to his warm-blooded counterparts. Throughout HARILIE’s fight to stay alive, Auberson discovers the machine has vast knowledge and understanding of life, love, and logic, posing the philosophical question whether or not HARLIE is human, and for that matter, what it means to be human.
Desde o meio do livro original (versão 1.0) (ele não possui "capítulos"):
“Well” said Auberson, wishing he were someplace else. “It’s HARLIE’s attempt to prove that he is of value to the company. If nothing else, he’s proven that he can design and implement a new computer system.”
“Oh?” Dorne picked up one of the printouts that lay scattered across the mahogany expanse. “But what kind of a system is it? And will it work? God Machines!”
“Not God,” Auberson corrected. “G.O.D. The acronym is G.O.D. It means Graphic Omniscient Device.”
“Is he planning to change his own name too? Computerized Human Robot, Integrating Simulated Thought?”
Auberson had heard the joke before. He didn’t laugh. “Considering what this new device is supposed to do—and HARLIE’s relationship to it—it might be appropriate.”
“All right,” he said. “Tell me about the God Machine.”
Auberson was holding a HARLIE-printed summary in one hand, but he didn’t need it to answer this question. “It’s a model builder. It’s the ultimate model builder.”
“All computers are model builders,” said Dorne. He was unimpressed.
“Right,” agreed Auberson, “but not to the extent this one will be. Look, a computer doesn’t actually solve problems—it builds models of them. Or rather, the programmer does. That’s what the programming is, the construction of the model and its conditions—then the machine manipulates the model to achieve a variety of situations and solutions. It’s up to us to interpret the results as a solution to the original problem. The only limit to the size of the problem is the size model the computer can handle. Theoretically, a computer could solve the world—if we could build a model big enough and a machine big enough to handle it.”
“If we could build that big a model, it would be duplicating the world.”
“In its memory banks, yes.”
“A computer with that capability would have to be as big as a planet.”
“Bigger,” said Auberson.
“Then, if you agree with me that it’s impossible, why bother me with this?” He slapped the sheaf of printouts on his desk.
“Because obviously HARLIE doesn’t think it’s impossible.”
Dorne looked at him coldly. “You know as well as I that HARLIE is under a death sentence. He’s getting desperate to prove his worth so we won’t turn him off.”
Auberson pointed. “This is his proof.” [edited for length]
Se você está curioso, li as duas versões e o lançamento do 2.0 é definitivamente o melhor livro. A versão 1.0 é dolorosamente datada e um pouco desajeitada.