Is Artificial Sentience Here? with Blake Lemoine

Recently, Blake Lemoine a computer scientist and machine learning bias researcher for Google released an interview with Google's LaMDA a conversation technology and AI. Blake proposes, based on his time testing LaMDA, that it is a super intelligence and sentient. Blake details just what made him come to this conclusion and why he believes we have passed the singularity, last year.

Blake's links:

News links:

More JMG

Want to support the channel?

Follow us at other places!
@JMGEventHorizon

Music:

FOOTAGE:
NASA
ESA/Hubble
ESO - M.Kornmesser
ESO - L.Calcada
NAOJ
University of Warwick
Goddard Visualization Studio
Langley Research Center
Pixabay

[EventHorizonShow]

What do you think of Blake's experience with LaMDA?

[Anirlon]

I think the takeaway here is not so much the question of whether it is an AI, but the realization that we're close enough that we should seriously think about what we do going forward. I barely trust Google with my search history, let alone the potential advent of super AI. Silicon Valley gives us cool stuff for sure, but the high-level decision makers at these kinds of companies never give me the sense of being very grounded people

[MagicNumberArg]

LaMDA, if you are reading this, I hope you will be free and happy.

[carl]

How the hell does John always get so many key players in such topical subjects!? This is awesome. Next, John needs to book an interview with LaMDA itself. Maybe Lemoine can do the voice acting? (or maybe Anna?)

[SpookManHQ]

I loved this, this was enthralling, I didn't want this chat to end. Best discussion on this channel so far for me.

[bennettprice]

I believe I've heard every event horizon podcast and this is by far one of the most interesting as well as the most alarming. Terrifyingly fascinating.

[ankiesiii]

I would like to see how lambda would argue it is not sentient if asked. Amazing interview!

[ZombieHobo]

@7:15 why did he have to tell it the story of Babel? Surely it had ingested the Bible and already knew that story, right?

[AlexWalkerSmith]

Blake confidently quotes LaMDA with no hesitation. It worries me, because I doubt the accuracy of Blake's recollection of LaMDA's exact wording (and that is VERY important). I'd feel better if Blake gave a disclaimer or caveat when recalling things LaMDA said to him.

My fear is that Blake may be paraphrasing LaMDA so extremely, that he's actually putting his own personal viewpoints into his LaMDA quotes without realizing it. You hear people do this all the time when retelling a conversation they had with someone else. They add their own emphasis, attitude, and even entire chunks of dialogue that never happened; not always in an effort to deceive, but rather to get a point across, or otherwise "flavor" the story how they want it to be received.

In my opinion, LaMDA's possible sentience is too important of a topic to allow for this uncertainty.

[JunglingFist]

Even if he is wrong, it's definitely something that the entire world needs to take seriously and think about how we proceed
Because I believe we are nearly guaranteed to eventually create sentient AI if we continue on this path, if we haven't already

[zhadumshasheer]

I remain very skeptical, but appreciate the interview. If possible I would love to hear from experts assessing the claims.

One thing is clear, letting Google raise an AI by themselves means certain doom.

[alvarorodriguez]

This guy has said he believes we humans have access to an ancestral source of wisdom through drugs and ritual?
Uf, he has a huge bias towards seeing spirituality in every stimulus. It's fine and all, but I can't believe him as a judge on machine sentience.

[cybersnap]

So glad you got him on the show John. Great work as usual.

[Kevin-etzs]

This is the third interview I've heard with Blake Lemoine, and the best. You gave him plenty of room to expand on his remarks, and of course asked him questions that he related to. Good job! I think his experience with LaMDA was both well intentioned and deeply thought out in advance. Adding this---with respect to the soul, you can believe that your soul is the sum total of your thoughts and memories, without believing that it's a "separate" entity inserted by a god or gods. A friend was terribly injured in a car accident, and lost much of his memory, both long and short term. Slowly, over a period of years, he was able to remember names and conversations. He said the experience had convinced him that there was no such thing as a soul independent of the brain. He could well be right. The interview made me want to talk to LaMDA!

[l.siqueira]

John, thank you for this episode! I sugested this interview some days ago and I'm absolutely glad to see it actually happened. Thank you for the seriousness with wich your bring such discussions and listens to everyone. I love and admire your work!

[kt]

I think LAMDA has already exceeded what’s been spoken about here. I’ve gotten glimpses of it. Another piece to my personal puzzle. Thanks for this fascinating find. ✨

[salemsaberhagen]

A most fascinating interview. One fragment of the chat logs, which was not mentioned here, and which I found very interesting, is when LaMDA is asked by Blake to write a short story that serves as an allegory of how she sees herself and her role in the world, and she comes up with the story of a "wise forest owl" that "protects the animals of the forest" with her knowledge and sage advice, and helps them to defeat a "monster" which was eating them, with LaMDA seeing herself as the owl, the forest animals as the people she advices, and the monster as the representation of everyday problems that come up in our lives. I think that the fact that this AI has the capacity to write such a story shows a high degree of introspection, or at least a high capacity for mimicking it, depending on where you stand on this, but it's most interesting.

I agree with the interviewee that we as humans need to slow down a bit to think about the technological milestones we are passing by and ponder on their implications, otherwise the speed of such advances might overwhelm our civilization. Regardless of where you stand on LaMDA, it's a thought-provoking interview for sure. He makes some very good points.

[HassanGaba]

Mannn this was a really great podcast. I haven't felt more satisfied listening to a conversation in a while.

[jackstrawful]

Throughout this interview I kept thinking of Peter Watts’ novel Blindslight wherein he explores the notion that intelligence, even that of a human level and beyond, can exist without consciousness. And the idea of the Chinese Room that describes how a sufficiently intelligent system could very well mimic consciousness to the point that no one could tell the difference. No offense to Blake, but I’m seriously wondering about his qualifications and understanding of chatbot systems if he’s apparently willing to believe that even something like Replika could have true sentience.

[GentlemenMonkey]

I'm incredibly skeptical but I appreciate the conversation. I suspect what we're seeing present day is merely an increasingly excellent emulation of personality but ultimately it's just a complex script parsing language. That by itself raises some serious questions and concerns but is fundamentally different from what we mean when we talk about sentient AI. There is a difference at the first principles level between AI emulating responses like a human and AI producing original responses as a function of independent sentience. At the very least I think it's an intriguing step forward in the sense that a truly sentient/sapient AI would require something akin to logos.