Due to recent concerns about AI by
Bill Gates,
Stephen Hawking,
Elon Musk, and
Steve Wozniak, the Lifeboat Foundation has launched an
Interactive Friendly AI project with a prototype available on this page.
Currently most Friendly AI research just produces papers and nothing to attract the public’s interest. Considering how much more money is being invested in nonfriendly AI than Friendly AI we think it would be worth putting some effort into Interactive Friendly AI so we could 1) get the public’s interest and hopefully encourage more Friendly AI interest/research 2) help develop algorithms that will be applicable to future Friendly AI efforts.
Our project has dual goals: 1) education, for making a chatbot aimed at clarifying what the risks and benefits of AI are, and discussing the various approaches to them and related considerations, dispelling misconceptions, etc. 2) beefing up the directedness and credibility of the actual research to actually make an AI friendly.
As this is a complicated task, it is our expectation that our project will be composed of many different AI engines and therefore we will provide options in the interface to turn them on and off so different approaches can be tested.
We are very interested in working with scientists who are developing engines based on goal directed agency, autonomous decision making, semantic comprehension, ability for self-modification, etc. It is hoped that the fairly narrow focus of working on Friendly AI engines will make it easier to experiment with advanced AI methods.
Key goals within our Friendly AI architecture are to develop empathy and to build datasets that demonstrate that a peaceful cooperative world is better than a hostile violent world. It is hoped that our efforts will contain useful algorithms for these key goals. (So these algorithms could be used by more advanced programs in the future…)
The Lifeboat Foundation welcomes everything from flowcharts detailing questions and answers that our AI should know, to hosting AI engines at lifeboat.com, to linking to AI engines elsewhere. Email programs@lifeboat.com with the subject “Lifeboat Foundation AI” if you wish to help in any way including providing feedback.
Currently most Friendly AI research just produces papers and nothing to attract the public’s interest. Considering how much more money is being invested in nonfriendly AI than Friendly AI we think it would be worth putting some effort into Interactive Friendly AI so we could 1) get the public’s interest and hopefully encourage more Friendly AI interest/research 2) help develop algorithms that will be applicable to future Friendly AI efforts.
Our project has dual goals: 1) education, for making a chatbot aimed at clarifying what the risks and benefits of AI are, and discussing the various approaches to them and related considerations, dispelling misconceptions, etc. 2) beefing up the directedness and credibility of the actual research to actually make an AI friendly.
As this is a complicated task, it is our expectation that our project will be composed of many different AI engines and therefore we will provide options in the interface to turn them on and off so different approaches can be tested.
We are very interested in working with scientists who are developing engines based on goal directed agency, autonomous decision making, semantic comprehension, ability for self-modification, etc. It is hoped that the fairly narrow focus of working on Friendly AI engines will make it easier to experiment with advanced AI methods.
Key goals within our Friendly AI architecture are to develop empathy and to build datasets that demonstrate that a peaceful cooperative world is better than a hostile violent world. It is hoped that our efforts will contain useful algorithms for these key goals. (So these algorithms could be used by more advanced programs in the future…)
The Lifeboat Foundation welcomes everything from flowcharts detailing questions and answers that our AI should know, to hosting AI engines at lifeboat.com, to linking to AI engines elsewhere. Email programs@lifeboat.com with the subject “Lifeboat Foundation AI” if you wish to help in any way including providing feedback.
