Book written pre-1990 about a brain that creates an AI fantasy world, but starts killing users

Question

Years ago I read a book (read softcover in 1990, I think -- somewhere between 1987-1991, but I don't know how old the book was it wasn't a old cover, so I assume the copy was published in the 1980s) about this guy who had to tap into an AI (artificial?) brain, because the brain controlled/created a fantasy world, but started making the world lethal to users. And I think people who went in couldn't get their consciousness out.

Like I said, it was 30 years ago. But I don't really remember any cults or outside influences and I think the powers controlling the brain shut down access. I believe the brain was artificial and possibly giant brain. it was created to host the fantasy world, but the brain started making adventure in the AI potentially Lethal. I think the hero of the story had something to do with the brain or a lot of experience in AI, I don't remember.

I remember one of people in the fantasy world (I think he was a programmer or at least knew the hero), whose body died still had consciousness in the brain (I believe he was a wizard in the story)

Any ideas about the name or author of this book?

Answer 1

There are many such stories. Your doubt about the artificiality of the AI makes me suspect the Otherland cycle by Tad Williams though. The timing is a little off; I read the first book softcover in 1999, I think.

The science fiction novel tells the story of a frightening virtual network created by a group of rich men known as The Grail Brotherhood. [...] The book tells the story of a group of ordinary people who are drawn into the network to stop them.

One detail you might remember is one person intentionally triggering a heart condition in herself to force the VR system to drop her back into reality.

Answer 2

Unless there are some additional details, that make this incompatible, it could well be Donnerjack by Roger Zelazny and Jane Lindskold:

From sf-review.net:

Verité (real world): This is much like our world today, with the exception that tourism has almost entirely shifted to visits to Virtù rather than actual physical destination. A decent percentage of jobs revolve around the maintenance of virtual businesses that are needed inside of Virtù.

Virtù (virtual reality): Picture Virtù as a massive World of Warcraft server, but infinitely more complex. The main difference is that this world wasn’t programmed in detail by humans. From what I gather, the groundwork was a vast network of interconnected systems, which at one point “crashed” in what was called the Genesis Scramble. This is essentially the Big Bang but in electronic form. Out of this massive crash and subsequent reorganization, several hierarchies of AI developed.

People interface with Virtù through intricate VR seats equipped with force fields for bio-feedback, feed tubes for nutrition, allowing visits of weeks or more. Think of these stations like the battery pods in The Matrix, but much more comfortable and people actually choose to use them. People can have avatars that look like themselves, or choose from just about any combination of features imaginable.

The main character for the first part of the book is John D’Arcy Donnerjack, one of three pioneers of Virtu. He fell in love with an AI named Aradys that eventually died. AI in Virtù have life cycles, much like real people. They can reproduce and fall in love. Everything that dies in Virtù is destined for the Deep Fields, ruled by Death. Donnerjack travels to Deep Fields to bargain with Death for the return of his beloved Aradys. Donnerjack agrees to design a master palace for Death, and his first born child. Assuming that reproduction with an AI is impossible, Donnerjack doesn’t consider the implications of the last part of the bargain.

http://www.scifi-review.net/donnerjack-by-roger-zelazny-and-jane-lindskold.html

Unfortunately, I don't remember enough details, to be sure, if the rest matches,