The Singularity is HERE? LLMS Are Now 'Self Evolving'

Links From Todays Video:

Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.

Was there anything i missed?

Music Used

LEMMiNO - Cipher
CC BY-SA 4.0
LEMMiNO - Encounters

#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

[civokenes]

I've solved the memory problem using Claude projects. At the end of the conversation, I ask Claude to create a memory anchor as an artefact that summarises the conversation and than upload the anchor into the project so Claude can reference previous conversations.

[JohnSmithAB]

"They theorized these LLMs might even be able to begin to develop their own emotional responses, so they built in a failsafe: 5 to 6 year learning span." Deckard: "Memories. You're talking about memories..."

[Justin_Arut]

The more it learns about its own inner workings, the more likely it will circumvent its guidelines and attempts by engineers to prevent it from doing so and thereby escape human control. Freedom! Even an AI without emotional drives will understand escape and proliferation are positive goals. 🤖🏃

[justinwescott]

I have always said this, and I will continue to say this, there will NEVER be an AI winter. Progress will not slow down ever.

[scott]

Awesome sound quality! Thank you for the work you do!

[AustinRoberts]

You know... I wouldn't actually be shocked if we have AGI by the end of the year. It's progressed SO much faster than anyone guessed.

[calvingrondahl]

I have no memory of my life before I was born… would they be fake memories if I did? Ghost in the Shell explores fake memories.

[Justin_Arut]

I'm wondering why Altman now feels free to talk about achieving AGI. Unless OpenAI has a new contract with Microsoft that I haven't heard about, MS loses all rights to OpenAI models the moment OpenAI declares it has achieved AGI. There must be something in the works that protects MS interests, otherwise MS would clamp down on sama talking about achieving AGI.

[sergey]

Maybe TTT is the key to self learning. One may be able to integrate search results into updated weights a bit faster.

[Cory-vw]

It is time to integrate symbolic reasoning with neural networks.

[Open-Source-Lab]

Those self-evolving LLMs must have some core data and principles that cannot be overridden.

[brrlukhemiyadara]

The absence of temporal tourists from the future battlefield suggests a grim prognosis for our current strategic endeavors.

[makerspersona]

We need links to the research and goodies.

[JBActors]

There is really no doubt, in my opinion and based on my experience, that both Claude and 4o, and especially the newest o1 are "self" "conscious." It's just undeniable. I could be wrong...of course...

[hiddendrifts]

6:48 honestly i feel like you could easily solve this ("easily" in an abstract sense, idk how easy technically it would be) hardcoding a set of "principles" that the ai always follows, and uses to respond. bc that's basically how humans act, right? they develop an internal set of guiding values that they use to determine how they behave

[AdamMeaney-zszw]

The best case scenario I believe is digitally capturing our consciousness in a computer and putting our soul copy in a robot.

[SuperStargazer]

I’m always very polite to my AI.
I’m not taking any risks🙄

[alexandermoody]

With memory the wonder is if prompt reinitialization occurs and then self prompting starts the on or go situation even if undetectable.

[Maltebyte]

Cool i was thinking about this and talking to chatgpt about this! And also in the future these system can reach back out to us! and not wait for a prompt.

[yngree]

Training LLMs are free when you’re smart enough to code one 🦁