Big Think
@bigthink·8.8M subscribers·10.1K videos
Why Superintelligent AI Could Be the Last Human Invention | Max Tegmark /big think
Posted
September 17, 2017
Views
35,591
Likes
1,105
Comments
214
Engagement
September 17, 2017
35,591
1,105
214
3.71%
Every word spoken in this episode is indexed. Type any phrase to jump straight to the moment it was said.
Type any word or phrase that may have been spoken. Click a result to seek the player to that exact moment.
Try a name, a topic, or a quoted line
See what was published immediately before and after this episode.
4:51Now PlayingWhy Superintelligent AI Could Be the Last Human Invention
New videos DAILY
Join Big Think Edge for exclusive video lessons from top thinkers and doers
----------------------------------------------------------------------------------
Max Tegmark has a bone to pick with Hollywood. We shouldn't be afraid of AI or, for that matter, a robot uprising. We should be more afraid of the next few years while we try and get AI through this early phase. Right now, just the same way a child would, machines take us literally. The key to the next few years is getting them to understand and adopt human logic—i.e. killing is bad and that just because you can doesn't mean you should—because if we don't set those boundaries now, in the future we may be viewed as nothing more than ants in their way.
Max's latest book is Life 3.0
----------------------------------------------------------------------------------
MAX TEGMARK:
Max Tegmark left his native Sweden in 1990 after receiving his B.Sc. in Physics from the Royal Institute of Technology (he’d earned a B.A. in Economics the previous year at the Stockholm School of Economics). His first academic venture beyond Scandinavia brought him to California, where he studied physics at the University of California, Berkeley, earning his M.A. in 1992, and Ph.D. in 1994.
After four years of west coast living, Tegmark returned to Europe and accepted an appointment as a research associate with the Max-Planck-Institut für Physik in Munich. In 1996 he headed back to the U.S. as a Hubble Fellow and member of the Institute for Advanced Study, Princeton. Tegmark remained in New Jersey for a few years until an opportunity arrived to experience the urban northeast with an Assistant Professorship at the University of Pennsylvania, where he received tenure in 2003.
He extended the east coast experiment and moved north of Philly to the shores of the Charles River (Cambridge-side), arriving at MIT in September 2004. He is married to Meia-Chita Tegmark and has two sons, Philip and Alexander.
Tegmark is an author on more than two hundred technical papers, and has featured in dozens of science documentaries. He has received numerous awards for his research, including a Packard Fellowship (2001-06), Cottrell Scholar Award (2002-07), and an NSF Career grant (2002-07), and is a Fellow of the American Physical Society. His work with the SDSS collaboration on galaxy clustering shared the first prize in Science magazine’s "Breakthrough of the Year: 2003."
----------------------------------------------------------------------------------
TRANSCRIPT:
MAX TEGMARK: Hollywood movies make people worry about the wrong things in terms of super intelligence. What we should really worry about is not malice but competence, where we have machines that are smarter than us whose goals just aren’t aligned with ours. For example, I don’t hate ants, I don’t go out of my way to stomp an ant if I see one on the sidewalk, but if I’m in charge of this hydroelectric dam construction and just as I’m going to flood this valley with water I see an ant hill there, tough luck for the ants. Their goals weren’t aligned with mine and because I’m smarter it’s going to be my goals, not the ant’s goals, that get fulfilled. We never want to put humanity in the role of those ants.
On the other hand it doesn’t have to be bad if you solve the goal alignment problem. Little babies tend to be in a household surrounded by human level intelligence as they’re smarter than the babies, namely their parents. And that works out fine because the goals of the parents are wonderfully aligned with the goals of the child’s so it’s all good. And this is one vision that a lot of AI researchers have, the friendly AI vision that we will succeed in not just making machines that are smarter than us, but also machines that then learn, adopt and retain our goals as they get ever smarter.
It might sound easy to get machines to learn, adopt and retain our goals, but these are all very tough problems. First of all, if you take a self-driving taxi and tell it in the future to take you to the airport as fast as possible and then you get there covered in vomit and chased by helicopters and you say, “No, no, no! That’s not what I wanted!” and it replies, “That is exactly what you asked for,” then you’ve appreciated how hard it is to get a machine to understand your goals, your actual goals.
A human cabdriver would have realized that you also had other goals that were unstated because she was also a human and has all this shared reference frame, but a machine doesn’t have that unless we explicitly teach it that. And then once the machine understands our goals there’s a separate problem of getting them to adopt the goals. Anyone who has had k...
For the full transcript, check out
Sentinel Indexing in Progress
Metadata and chapters are available. Claim extraction for this episode is pending.
All video content is delivered via YouTube embedded players in accordance with the YouTube Terms of Service. Sentinel provides research tools that promote discovery and accountability across political media.
