Peter Cheeseman - Implications of Artificial General Intelligence (AGI)

2,590 views · Published 20 August 2017 · 35:59 · Indexed 23 October 2026

Channel: Science, Technology & the Future · 2017 · Science & Technology

Watch on YouTube

Synopsis: It looks highly probable that the first AGI is only a few decades away. When it arrives, the implications for continued human existence are dire! One indicator is that we do not know of any alien intelligences--but why? Given the recent discovery that most stars have planets, and that life on earth got started shortly after it formed, it is likely that there will be billions of planets in our galaxy with life. It is also likely that at least some of these lifeforms will have advanced to human level of intelligence and beyond--yet we see no evidence of this! While there are many possible explanations for the missing aliens, one explanation is that alien civilizations developed AGIs and the AGIs then destroyed their creators. Another explanation is that the aliens merged with their AGIs making them individually super-powerful.
In this case it only takes one rogue super-intelligent alien to destroy them all. Even if we are very cautious in developing a "friendly" AGI, the risk of bad unintended consequences is very high. Trying to out-smart an AGI that is orders of magnitude smarter than us is a losing proposition, so once created, it is unstoppable. While I believe there are possible technical methods to develop an AGI that is benevolent to us, mainly by interacting with us as it develops, this misses the main risk. The main risk is not the AGI, but who owns it! Developing an AGI will take considerable resources, so whoever puts up these resources will want to exploit it to recover their investment, and gain a powerful dominating position. Other organizations will regard this monopoly on power to be intolerable, and will try to develop their own AGI or steal the existing AGI technology. Once we have competing hostile organizations with AGIs, the game is up!

Talk held at 'Artificial Intelligence / Human Possibilities' event as adjunct to the AGI17 conference in Melbourne 2017.

Assessing emerging risks and opportunities in machine cognition

With AI Experts Ben Goertzel, Marcus Hutter, Peter Cheeseman and Joscha Bach. 

Event Focus:  
Given significant developments in Artificial Intelligence, it's worth asking: What aspects of ideal AI have not been achieved yet?  
There is good reason for the growing media storm around AI - many experts agree on the big picture that with the development of Superintelligent AI (including Artificial General Intelligence) humanity will face great challenges (some polls suggest that AGI is not far).  Though in order to best manage both the opportunities and risks we need to achieve a clearer picture - this requires sensitivity to ambiguity, precision of expression and attention to theoretical detail in understanding the implications of AI, communicating/discussing AI, and ultimately engineering beneficial AI.

Meetup details: https://www.meetup.com/Science-Technology-and-the-Future/events/242163071/

Many thanks for watching!

Consider supporting SciFuture by:
a) Subscribing to the SciFuture YouTube channel:  http://youtube.com/subscription_center?add_user=TheRationalFuture
b) Donating via Patreon: https://www.patreon.com/scifuture and/or
c) Sharing the media SciFuture creates: http://scifuture.org 

Kind regards,
Adam Ford
- Science, Technology & the Future

More from this channel