[infurl]

Another video interview conducted by Lex Fridman with an AI researcher. These interviews are always packed with fascinating insights by the great minds of our time and this one is no exception.

[Art]

https://www.youtube.com/watch?v=E1AxVXt2Gv4&feature=youtu.be

Try this link if you too, receive a "Unable to play video" message.

[infurl]

https://www.youtube.com/watch?v=E1AxVXt2Gv4&feature=youtu.be

Try this link if you too, receive a "Unable to play video" message.

That's weird. Which browser are you using? I'll try to reproduce the problem. I'm using the latest Chrome on Mac OS X Catalina.

[Art]

I'm presently using BRAVE, a very fast browser running off Version 1.4.95 Chromium: 80.0.3987.122 (Official Build) (64-bit)

It is an Open Source product of Mozilla.

Just to double check, I opened my former favorite, FireFox and the video does not play in that browser either. It states:

Video unavailable
<a href="
feature=emb_err_watch_on+yt" target="_blank">Watch this
video on YouTube.[/url] Playback on other websites has
been disabled by the video owner.

Why you're able to view it on your Chrome browser escapes me at the moment but apparently the owner wants it played & viewed on YouTube for whatever reason.

Cheers! Where there's a will...blah...and blah...

[infurl]

Why you're able to view it on your Chrome browser escapes me at the moment but apparently the owner wants it played & viewed on YouTube for whatever reason.

That's the issue then. I never watch embedded videos in situ, I always click the link embedded in them to open them directly in YouTube. On the rare occasions when the embedded video has been set to prevent it being watched outside the host page, I'll either search for a similar video on YouTube or forgo watching it entirely. I'll try to remember to test the link both ways next time.

Aside from that Art, what did you think of the content of the video, assuming that you had time to watch any of it?

[HS]

He defines intelligence as the ability to do well in a wide range of environments. Which is a solid description of the results, but not of the phenomena itself, and could be attributed to other factors such as being a Tardigrade for example. Humans do well in one environment and just recreate it wherever they go (influence), but something like a Tardigrade literally does well in a wide range of environments without the need for clever plans.

Maybe a graphical model would be useful for comparing different types and amounts of intelligence. The volume of someone’s sphere of influence is probably tied to their intelligence. On one hand, a human tends to posses a greater sphere of influence than an organism of a much lower intelligence. However, circumstance, physical capabilities, and social power, all mess with this trend. So maybe, intelligence is the ability to modify your sphere of influence over your environment.

This could scale, because the greater the volume & density of influence the more intricate the possible changes. Intelligence could = (d influence/dt). If the extents of influence = the boundaries of self, then it is the ability to modify your organism-environment system to the greatest degree.

[Art]

@ infurl,

Yes, I took time out of my day to watch the video in its entirety and found it quite interesting. Of course, his AIXI is theoretical, a platform for setting up problems and using predictive analysis along with action and reward conditions (reinforcement learning).

Based on his interview, he believes that in the near future machines will be able to think on human level or better. He references the Chess & Go playing computers and that Alpha Go discovered new openings after only 4 hours of playing itself.

He said that Turing Tests would impress him if they could carry a decent conversation for a few hours with another person. He also joked that many humans would likely fail such a test!

Interestingly he pondered whether AI should be relegated to a Philosophy Department instead of a Computer Department (forgoing computational abilities until some time later).

AIXI is the most intelligent agent which anybody could "build". (according to him)...

I liked his notion of Self-improving AI provided a proof was constructed and accepted. Bigger, Faster, Stronger...

He feels that AIXI could possibly be viewed as conscious if viewed as such but that would be very subjective. IF it or AGI were to become conscious AND have feelings then we (humanity) would really need to be concerned.

He did recommend some nice books for those interested in learning more from AI/ML research.

Overall, an interesting interview and certainly his opinion weighs more than some but perhaps lacking according to some of those computational giants in the field!

Without starting a war, it would be nice to see other viewpoints from Ben Goertzel, Sundar Pichai, and perhaps Ray Kurzweil or others on this topic.

I liked his quip that Infinity is getting closer to us every day. (or to that effect).
 

[infurl]

Thank you for your superb analyses of the interview @Art and @Hopefully_Something. At 100 minutes it was a long interview but it was so fascinating that the time passed quickly. Those podcasts are not ones that you have to watch, they are easy to listen to while doing other things. @SquareBear might like to know that Mitsuku was discussed during the course of the interview. You have fans in high places!

[LOCKSUIT]

Most (and I utmostly) agree AGI belongs in a Computer Science department with proven knowledge, not a philosophy department. He should know this himself as his Hutter Prize contest is all about solid verification of the entries using Lossless Compression accuracy as evaluation. Maybe you misheard him Art. The only thing philosophy that belongs in AGI or physics or Earth etc is storytelling/thinking because all data ex. vision, Earth, motor actions, etc is a story and the AGI brain 'runs' on short cut finding using past experience *Related* stories.

[squarebear]

@SquareBear might like to know that Mitsuku was discussed during the course of the interview. You have fans in high places!

Thanks for letting me know about this. I'm very happy that a senior research scientist at Google DeepMind and professor at Australian National University says that Mitsuku is impressive! 

[infurl]

It's interesting to draw parallels between AIXI and the Free Energy Principle.

For those who missed it:

https://en.wikipedia.org/wiki/Free_energy_principle
https://www.wired.com/story/karl-friston-free-energy-principle-artificial-intelligence/

[LOCKSUIT]

Yes I've been saying that lately too. The future Earth will be a fractal pattern of nanobot modules made of nanobot modules to know where, when, and what everything is so as to use least energy as can and stay immortal with least change as can. Free energy/knowledge extraction from compressed files or gasoline/wood/food/nukeRods is also obtained and used to Sort bits/particles on Earth into a bigger pattern. We are trying to make Earth into a complete pattern fractal so we can easily predict where, when, what everything is and are completely friend network aligned as a powerful field domain of energy communication (global context). We make the future using the past data state of Earth into a better pattern and immortality probability. Free energy/knowledge Extraction (data compression term) is not only extracted for free so we can make Earth into a better predictable place to work in, but also so we save even more energy in the new world that is much more predictable.

[Art]

Most (and I utmostly) agree AGI belongs in a Computer Science department with proven knowledge, not a philosophy department. He should know this himself as his Hutter Prize contest is all about solid verification of the entries using Lossless Compression accuracy as evaluation. Maybe you misheard him Art. The only thing philosophy that belongs in AGI or physics or Earth etc is storytelling/thinking because all data ex. vision, Earth, motor actions, etc is a story and the AGI brain 'runs' on short cut finding using past experience *Related* stories.

LOCK: Watch 1:15:10 for a minute or so...