Popular Questions about AI[]
In my experience, the questions people are asking about AI are on the order of:
- Could an AI answer complex questions?
- Could an AI protect itself?
- Could an AI program itself?
- Could an AI become self-aware?
- How will AI know what is appropriate, and what is inappropriate?
- Would an AI try to kill us all?
- Would an AI try to manipulate us?
- Who would control AI?
- Would AI ask for rights?
The page is presently answering the questions:
- What is the history of AI?
- What are some approaches to AI?
- What's the current state of AI-like sensing efforts?
The goals of the wiki are:
- Teach about the future.
- Teach about future analysis technique.
It is clear that the "would" questions belong here, as scenarios. What is not clear is if the "could" questions belong here.
My immediate question is:
- Should we answer about the plausibility of AI?
- If yes, where?
- If no, (which is a perfectly acceptable answer,) then we should link people to a place (or places) where people do answer those questions.
- I think we should answer that very briefly and link to other places. Specifically we could write:
- Even before general human-level AI can be programmed to simulate/replace a skilled specialist. An AI could be given independence and then it would become at least partially autonomous. AI could access its own programming, writing and rewriting parts of its software. A sufficiently complex AI could be fitted with........ (to be continued). :)
Artificial Hippocampus[]
I did some research on the Artificial Hippocampus.
Here's what I came up with, very quickly.
- "If it works, the team will test the prosthesis in live rats within six months," - 2003, March, 12
- "They are due to start testing the device on rats' brains shortly." and "If that goes well, the Californian researchers will test the artificial hippocampus in live rats within six months..." - 2003, March, 12
(Didn't happen- they did the first tests, but for some reason, decided not to go forward with live rats.)
Interestingly:
- "While they haven't tested the microchip in live rats yet, their research using slices of rat brain indicates the chip functions with 95 percent accuracy. It's a result that's got the scientific community excited." - Wired, 2004, October, 22
- "The team expects it will take two to three years to develop the mathematical models for the hippocampus of a live, active rat and translate them onto a microchip, and seven or eight years for a monkey. They hope to apply this approach to clinical applications within 10 years. If everything goes well, they anticipate seeing an artificial human hippocampus, potentially usable for a variety of clinical disorders, in 15 years. " - Wired, 2004, October, 22 (again)
(That is, check again in 2007.)
- "...the artificial hippocampus is largely vaporware at this point so caveat emptor,.." - 2004, April, 1
Graphics: