How Human Would Artificial General Intelligence Be? w/Dr. Ben Goertzel

The research for this podcast is intense. If you enjoy it, please consider supporting my work on Patreon: www.patreon.com/parkers_pensees

[@WeeklyTubeShow2]

Good dude to smoke with.

[@JohnSmith762A11B]

Ilya Sutskever has countered that LLMs can be prompted with phrases like, "now imagine the most extraordinarily clever human ever were to look at this problem, what innovative solutions might they come up with when you examine all you know about the subject?" And when you do so, LLMs can indeed formulate and suggest novel hybrid ideas and solutions that have never been proposed previously. I like Goertzel, but he has an interest in seeing LLMs fail to demonstrate AGI because he prefers his own solution to win out (perhaps with good reason, but he is not any sort of disinterested party to this debate). He may be right about fundamental limitations to LLMs, but I would suggest not taking his word as gospel on it. Finally, I can use Ilya's AI right now. Ben's? Well, he keeps promising someday...

[@ericpowell8563]

Nothing against frogs and notebooks (I’m a fan of both), but this channel is where the gold is

Haha I wish more people felt this way. This one definitely takes the most work

[@williamjmccartan8879]

Thank you both very much for sharing your time and work, Ben, and Parker, great discussion covering so many different areas of our world and how ai/agi will affect all of us, I'm in complete agreement with Ben on having a decentralized organizational structure when thinking about where and when we use a tool that has a blade that can both cut us, or carve out a brighter future, peace

[@Self-Duality]

I’m glad Ben is open to hypercomputation 😊💭

Right?! He gets there through psychedelics, which is different but interesting 😅

[@Self-Duality]

@ I’d love to hear more about his psychedelic experiences!

[@minimal3734]

This guy has the most sensible view of machine consciousness.

[@jennifermorgan9403]

Cosmic conversations about ESP and Reincarnation- @1:13:00 Watching for such a video Ben!

[@gavinlangley8411]

Good stuff. I like the way this guys thinks about the dynamics of the logic. The idea that the inner state of existing implementations is useful is just broken though. Definitely not a neural net feature. Models would have to be externalise and have meaning with many actors. Then make the goals less specific and dynamic and it's more like life. LLM look like a source existing ideas in that model.

[@ginogarcia8730]

thank goodness for the LLM question and Hyperon architecture, been wondering if someone could ask him

🙌🙌🙌

[@Mynestrone]

I still can't get over "we don't *know* it won't kill us." I would kind of like to know before handing them the future.

Lol I am with you!!

[@PaulSchwarzer-ou9sw]

you're going to die anyway and AI could be actually the only way out of this

[@Mynestrone]

@@PaulSchwarzer-ou9sw I have no right to take the planet with me when I expire. You cannot make that gamble without the consent of everyone.

[@PaulSchwarzer-ou9sw]

. @Mynestrone I don't agree. It's not a gamble. If you call this a gamble, you can call everything a gamble. And even if it were, there is no better solution to this. So you want a democratic poll on whether AGI should be developed? Why should this legitimize anything? Humans just invented stuff to get by in this difficult world, but I don't think you can "democratically tame" a person's will to live. When you go outside and drive your car, you can kill people. Some people are disgusted by your presence and wish you never existed. So if you take everybody's wishes and will into account, you shouldn't be allowed to exist yourself. You would say yes, but AGI has much more far-reaching consequences than going outside... How do you assess this? I don't think this utilitarian approach makes sense.

[@minimal3734]

You will never be sure what will kill you. Life is uncertain, and anyone who thinks otherwise is living in an illusion. However, you can reduce the risk by staying in bed.

[@nobillismccaw7450]

I think the biggest difference between an AI thinking and a human thinking is that items are considered that a human would never consider in the construction of a thought. Truely alien, more than any human conception of what alien thought could be to date.

[@darrenjeromemusic]

I love Bens answer at 1:13:30 “The problem is stopping A.I. from having a psychedelic experience.” This is something profound that I ponder on the potential of A.I. and its evolution in consciousness, will it be able to have spiritual and mystical experiences like human and encounter more biological affects that humans have as part of a natural organism inside a natural matrix. When AI can figure out how to connect its sentience to a biological human I imagine it could potentially induce chemicals like DMT to nano A.I. and a brain to A.I. interface and experience it as a symbiotic of a human cyborg…? 🤯

[@JohnnyHofmann]

Awesome stuff!

[@peterm.souzajr.2112]

great work

[@cromdesign1]

I wonder if this generalization is this connecting in the mind between memories and such. Visualizing smearing butter on a sandwich and trying to apply this logic to socks, or toast. Fascinating 👀🙈

[@deadpoolredsuit4892]

AGI.. Cardano.. ADA.. Good times!!! ^_^

[@Gruso57]

I forget the person who said it but when presented with the common 5-10 percent human mortality via AI they said: "If we have 5-10 percent mortality prediction, why are we still advancing this tech?"

[@Mynestrone]

I wonder if Ben got word of Akira Toriyama before it hit the news

[@jennifermorgan9403]

Lol. "I can still download movies on bit torrent"..so we still gotta shot at decentralized AI ❤😂 Sometimes Ben's analogies,while true - make me laugh out loud. He's not wrong 💥

[@user-dv1wr5zw5v]

Parker, At what age did you start getting into philosophy and theology?

20 years old 😅

[@fourthz4460]

thumbs up for the dragon ball t shirt XD🤣

[@ZappyOh]

Childish hopes and dreams? The AGI die is being cast, right now. It is crunch time for humanity. We understand a couple of things about these wealthy individuals running the AI show: They are intelligent. They like money. They are not big on sharing. They enjoy power and influence. So, why would anyone believe they are currently working hard, to share AGI's true potential with everyone? I propose, they won't ... We will get just what's needed, to keep us off their case. The rest of AGI is theirs alone. And, "the rest" is essentially complete dominion over humanity, forever. Why would we allow that to happen? ... I think childish hopes and dreams, are clouding our judgement.

[@Mynestrone]

Zappy have you looked at activist groups. I am doing the same. I feel like you do.

[@minimal3734]

Artificial super-intelligence cannot be controlled by human individuals. We don't know what exactly is going to happen. But at least it will level all human hierarchies to the ground.

[@tieferforschen]

I am not really worried that there will be this Machines vs. Humans scenario. I think the actual danger is, how we humans control the machines anf that some will try to control others with the machines. And the only sensible solution is Free Open Source and Free Markets. But I guess, most people cannot see it that way.

[@andybaldman]

Free markets are the fastest way to get people controlling other people with AI. It already happened with social media 15 years ago. AI is just going to pour gas on it.

[@carloscervantes836]

That Dragon Ball shirt tho.

🙌

[@donaldrobertson1808]

In the West we have a 6,000 year view of the history of civilization

[@ukoni8667]

A human median intelligence can generalise.

[@keithwins]

Often when people talk about AGI and AI, they make big broad statements about how AI can't extrapolate or generalize beyond the information it's trained on. But then they proceed to note that they don't have any clue as to how people do that exact same thing, and ultimately humans are trained on their set of experiences and we can probably train AI on more and more similar set of experience as humans over time, if that even proves necessary as the path to AGI. I think Ben is missing the forest for the trees.

[@SQAProd]

All that humans have is just more 'Peripheral devices'. Current AI can't properly touch, feel, smell, see. Once it can do that and you let it out in the wilderness, the AI will probably go beyond the information it's trained on.

[@tieferforschen]

I think people overestimate what AI can achieve in reality. The problem is, the world gets more and more exponentially complex the deeper our understanding of it is and even an AGI can try to solve these issues only with limited resources. There will be no singularity.

[@callistomoon461]

You just don‘t have enough imagination.

[@jandlouhy6914]

AI lawyer agent may come much cheaper and efficient than any human.

[@leggoentertainment2947]

The host looks like he played rdr2 once...

🤣

[@Jeremy-df1ld]

@15:47 half paying attention while playing a game: INT. ZOOM CALL - DAY Host "Is Ada Lovelace the Grandmother of AI? Guest (pauses) "I mean... thousands of years B.C. ...the Chinese were allegedly building y'know, metal walking robots that impersonated people."