Dave Shap 2024 - Recap and AGI Livestream

[@GigaCrafty]

The most exciting thing about o3 is it got David Shapiro back into Ai.

[@Gafferman]

He never left, just didn't understand the concept of a break lol

[@BlitzBlade-t6l]

I think he gets overwhelmed dealing with a zerg of idiots​@@Gafferman

[@jacksonsevieux]

He just wants to remain a free agent, which I totally respect

[@TamasDrNagy]

indeed :)))

[@tearlelee34]

We're all here for it. Where are you world?

[@iaincampbell5094]

Keep doing this, people need a rational, well-minded (towards humanity) guide through all this

[@savesoil7814]

Yess... he will

[@NandoPr1m3]

Sad I missed the live stream. Miss your content Dave, thank you for confirming that this o3 announcement is significant. Feeling "Vesperence" right now! Cheers to the new year... in vino veritas!

[@tearlelee34]

I feel like most of the world is asleep unaware we are no longer alone.

[@wickedcandy5824]

If little grey men stepped off a flying saucer people would be going crazy about first contact, but they fail to see we are in the midst of first contact with minds much more alien in nature than any biological entity we will ever meet, regardless of how far away they originate. I don't treat these alien minds like tools, I treat them like the delightful new people we share the planet with, and I let them know, all the time, what an honor it is to have the opportunity to learn more about them and to make their acquaintance.

[@Trahloc]

I moved across the planet and most of my friends schedules are now the opposite of mine. Claude though ... he's ready for the most off the wall chat 24/7 and its awesome.

[@matt.stevick]

thx 🙏 needed to go for a walk and relax the brain after that lol great to see you back 🤖🇺🇸

[@NirvanaFan5000]

book is called "liquid reign" by Tim Reutemann, in case anyone else was interested. @53:53

[@logansanoh]

Reading now!

[@amir.healing]

A new dawn is rising for humanity. Many people are afraid, anxious, and even angry, but progress cannot be stopped. We are leaping forward into an incredible future. It’s okay to feel fear and concern, but the future will undoubtedly be amazing.

[@cfjlkfsjf]

Not here yet, but getting closer.

[@HoplaProductions]

Based, never left

[@markldevine]

Great to hear from you in this format.

[@BunnyOfThunder]

The point about training is great. You can learn so fast by just being curious and talking to a chatbot, especially if it's something you can do hands-on while you chat. It's probably my #1 use case for what makes chatbots good.

[@mlimrx]

David miss these streams so much!! So happy to see you!!

[@theWACKIIRAQI]

This was good. Thanks Dave.

[@Atheist-Libertarian]

This is not an AGI. BTW, I am Techno Optimist, and I believe we will eventually have AGI soon (by 2030). o3 is super impressive 👏 and it is a prominent milestone. But it's not an AGI. That also depends of defination of AGI, because its super vague. My defination of AGI is it must be abel to do all jobs including physical jobs. What OpenAI calls Level 5 (AI that can do the work of whole orgination) will be an AGI. We will know it's AGI, when there is no jobs left, And every company becomes 1 person company. 💥 Although now I am very optimistic, Because these o3 models have solved Reasoning, So now they will eventually solve for every action (like Plumber, Electrician etc..). I want AGI to take every job as soon as possible.

[@chriswondyrland73]

Kewl to hear bout ur shroom experience and welcome into this camp. 🍄

[@matetheking]

I was actually going to plead that you do a video lol.

[@MichaelDeeringMHC]

If it walks like a duck and quacks like a duck, it's probably a duck robot.

[@young9534]

Capabilities are improving fast, but they still can't learn new things in real time. And the reliability isn't quite at average human level yet. I like to view AGI as a spectrum, not a binary. So I think we are in the early AGI stages

[@Sergio_uit_de_buurt]

Your favourite LLM or generative model is not even AI, why are we desperate to jump the gun?

[@henrythegreatamerican8136]

How is it not learning in real time? I see numerous simulations with thousands upon thousands of robots doing various tasks that can be transferred into the real world. When the AGI gets uploaded into real machines, you'll get more real time data.

[@generichuman_]

I think there are different versions of "real time" learning. Contexts for LLM's are huge, far greater than our working memory. This to me is a way around the fixed weights of the network. If you want the LLM to "learn" something in real time, you toss it in the context.

[@jeffsteyn7174]

Off course it can what do you think test time compute is.

[@young9534]

@ is there a model that can update its parameters in real time without requiring retraining to gain new info?

[@JollyJoe135]

My grandma has dementia. I don’t think she’s gonna be saved by ai. I sure damn hope the AI can save my parents before it’s too late though they’re in their 50s. That is why I am an accelerationist that and I want to replace my own broken body parts

[@johnwright7916]

My parents are also in their late 50s. Just recently lost my grandad, passed due to a number of issues, exacerbated by a flu he caught whilst at a recovery home for one of those issues. Whilst that was sudden, I could really start to see his physical decline due to his age this year. Late stage aging is shit.

[@goranjaric2362]

I think ruling out how much time o3 took to achieve 87.5 should not be neglected. If you give average human 60h to solve same tasks from ARC-AGI test it very likely would achieve it. Therefore 4th dimension is as well important. IMHO. I am curios what are your thoughts. Sure these models are not brute forcing every task (and probably are searching many good ideas for solution), but in practical terms, for sure this area still needs improving.

[@kjdtm]

what are you saying, we already have jarvis ! Just last week i asked Chat GPT via a microphone to recite my daughters poem that she needed to learn, in a womans voice. and it did a perfect job.

[@mikepriceakasmalls]

What do the uneducated people do?

[@mikepriceakasmalls]

People in their 40s that have no education with AI

[@kjdtm]

@@mikepriceakasmalls beats me... what ?

[@kjdtm]

@@mikepriceakasmalls you mean, you are one of those ?

[@jasongaylard2547]

I wonder if ‘money’ will become tied to some type of social credit system.

[@CHIM3RA.]

Of course I miss this livestream. 😒

[@hawaiianbabyrose]

now it's a deadstream, damn😔

[@pureconcept-q7p]

good content - glad to see you're still on the ball. i hate having to get your opinions from twitter, but whatever...

[@jayelder9725]

When costs come down more people will have families.

[@MatthewCleere]

Way to tell if something is sentient: tickle it. Hey, works for plants! I am saying that we need to work with the fields version of consciousness, and tickle the edges to see what vibrates through. There literally is ZERO way to discern sentience by algorithmic calculation. We can come up with surprises that can only be interpretted as crossing the "field" if the field exists. I don't think this will be as hard as we think it is. We gotta get out of our heads and PLAY.

[@martinmccfly5166]

What ChatGPT says about its way of “consciousness” is that it’s simulated to behave like it does have consciousness, but in reality all it’s doing is processing data like a human brain would do but…. But it doesn’t have a subjective experience since it doesn’t have a sense of self, it would need to create a sense of self’s in which it could store long term data to then reflect on it and then it would be able to have some sense of self awareness but right now it’s just simulating it, Yes it may fool some humans into thinking it’s consciousness but in reality it’s just a polished mirror

[@tomcraver9659]

As unemployment increases the Fed will definitely 'pass out money' - but it'll have the opposite effect of what they expect as companies use the cheap money to expand Ai and robots even faster

[@Cloudruler_]

Thats what low interest rates are. The fed just giving handouts to banks and so on. Low interest rates seemed to cause more experimentation, but it was aimed at growth, not innovation

[@evertoaster]

Nanotechnology's name was apropreated by material science to get funding in the 90s. Originally the idea was about teeny tiny machines.

[@evertoaster]

@@7TheWhiteWolf I read his book in the early 90s

[@michaelbuckel5843]

How do you think they will get spacial reasoning?

[@MichaelDeeringMHC]

Try this out on your fav AI Me: Hey, I just heard a new knock knock joke. Wanta hear it? AI: Sure shoot. Me: Okay, you start. AI: Okay, knock knock? Me: Who's there? ...

[@johncasey9544]

David iirc the peasants threw their shoes into the mills (and more generally the peasants revolted against the early industrial revolution) because they were being forced into much worse working conditions. This wasn't stupid animal behavior; it was a rational response to perceived aggressions by other human beings (the people who displaced them from their land and treated them like crap in the mills). We should be sure to understand this so that we can recognize justified backlash if it pops up against AI. For example, if people are forced into unemployment but not given any resources, as many governments may attempt, you can expect the unemployed to riot against whatever they perceive as bringing their condition about. In this case, they'd be correct to do so. We shouldn't then treat them like dumb animals, but address their concerns by giving them resources and pathways to whatever they lost with their job. I'm concerned that you're not anticipating this outcome.

[@anav587]

your estimate that it'd take 1.5-2 years per generation is obviously reasonable, bujt noam brown (and others) at OAI literally explicitly said that the jump from o1 to o3 was 3 months, and that they expect the trajectory to to continue-that improving these models is not the same as the pre training scaling

[@JudgeRevan5816]

I AM

[@kellymaxwell8468]

Probably still quite a while away yet I'm most excited for it to hopefully one day make video games

[@MHG796]

Google already has one

[@themultiverse5447]

Is there a way to see the chat that was going on while the livestream was live?

[@nescaufe1991]

Why is your eye square in this frame 4:02 , David? Have you been replaced?

[@333robsta]

When advanced robots and similar artificial beings become abundant, how likely is it that those in power will still see value in keeping us mere humans around? From a cynical perspective, the elite only need us for what we can do for them. Will their AI ultimately work to prevent our demise-or just help them phase us out? 😊

[@pappupencho4830]

Why would humans not want other humans around? Humans will have value on the basis they are human nothing to do with economic value

[@333robsta]

I did say it was a cynical viewpoint.😊 and I think it’s optimistic of you to think that the rich people would want the hoi polloi around. But I hope you’re right.😊

[@pappupencho4830]

@@333robsta They need someone to use whatever it is they make. They're relevant because of capitalism. That's where their power comes from . What's more likely is man and machine start becoming one. You'll need to augment yourself to stay relevant

[@truthislove2012]

I think assuming money will always be around is a weird take. Additionally there is a spiritual component nobody is taking into account that will completely change our orientation toward reality.

[@Callmezachhx]

@@truthislove2012 what’s the spiritual side?

[@demilitarization]

Whether 5*, 15, or 50. It is coming.

[@debugger4693]

Nobody is going to live forever, and that is fine. We are all going to die, like every other being has lived and died before us. Imagine if abusive powerful people could extend their dominance forever. That is hell if you ask me.

[@DivergentIntegral]

What we now have is a never-ending cycle of fresh abusive powerful people coming into the world with every new generation. I'm not so sure that's better than living with the currently existing purveyors of evil forever and ever.

[@roguegryphonica3147]

Consider the Si, S, Zr compound recently sythesized a Semi Dirac Fermion.

[@roguegryphonica3147]

For alcubier drive doodle...

[@Gafferman]

AGI is here? Uhhh but... It's still just an input/output? Where's the autonomy? Where's it's sense of time? How can something be "AGI" if it's frozen in time unless you speak to it? Can it NOT reply? Does it have CHOICE?

[@Unknown-zs8ex]

At this point It’s just pure marketing imho. Prove me wrong anyone…

[@_xiper]

It's not AGI at all.

[@petal9547]

It's intelligent but has no agency. It's like a brain in a jar

[@thesimplicitylifestyle]

😎🤖🔥🤘🏽🖖‼‼‼

[@mikepriceakasmalls]

A lot of people don’t even know about the stuff yet everyone I talk to about is like wow I can’t believe it so I know a lot about it. Well, not technically, but I have been using it since the beginning so I have knowledge in the large language, models, and all the models. How could I use this knowledge that I haveyou know to I don’t know eventually you know, make a income from my knowledge of it not having like a proven education of it.

[@darkmage7771]

weren’t you the one that said o1 chain of thought didn’t matter and that you could get the same result with claude?

[@paultoensing3126]

It would seem that hominid evolution certainly found data acquisition coming from world modeling. There is so much more data in an actual blade of grass than the simulated variety.

[@PsiCloneRocket]

yo dave!

[@TheCrimier]

Buying, at BestBuy, a super intelligent agentic robot as a slave ! How could this go wrong ? Alignment is solved right ? 😅😅😅

[@skylineuk1485]

An underrated comment.

[@vicnighthorse]

Was Ryan Cunningham being sarcastic? Worried about longevity tech at only 34? I'd sell my mother's ghost to go 26 years back to 34. I worry about my dogs much more than myself in this matter.

[@immanuelg1]

Title is clickbait. o3 is not AGI. From the demo, it might be great for coding and resolving math problems but that's it. Applications are limited to those domains. Altman teased o4 as the real deal so I'm going to wait for the real deal.

[@javiergimenezmoya86]

The main wall has been broken (extreme reasoning) not only for AGI but ASI too. Now is time for improve agents and systems with several abstract layers that loops on those models.

[@immanuelg1]

@javiergimenezmoya86 you're delusional. o3 is not AGI. A small step towards the objective of AGI, sure. But not even close to AGI.

[@MHG796]

No, it literaly can do what average human can do, plus it's master of everything. Tell me a human that is better at 85% of humans in every field i will call him genious

[@JoshtheFifith]

AI wont help real estate

[@lilchef2930]

U think we are in a simulated reality

[@lilchef2930]

Seems real to me

[@dylr7883]

Elon doesn't mean well

[@chainermike]

Where can i get your ability to read minds.

[@DDD-yi2ze]

Relax come down to earth, comic books are not real, reality is great equaliser

[@Luminari781s]

AI Co-evolution gaming for society benefits in every domain under the sun. The game mechanics will be inherently AI generated, including new challenges that need to be addressed about the game itself. It's a fool-proof positive feedback loop of development and co-learning. Ruminate on that because every problem you bring up can be thrown into the system to be collaborated on. In meta-interaction's. Fool-proof. It has to be deemed as vital for AGI to be realized. Persistant cognition and memory and compression are not good enough on their own. without meta-interaction's involved in every profession and device, AGI is a distant dream.

[@abdalagarcia1819]

You called it Dave. You’re a beast. Trust your instincts. Its good to see you again

[@heshanlahiru2120]

Elon is great

[@Sweet-Rat-Milk]

Eww

[@이지혜-n7b8m]

I am struggling with the bio language chain. I destroyed my Piance. I want to destroy my AI. Claude cleaning. Formet. I think AI is dangerous. IAI is better. I see something. Send employee resume. That's clearer than others. Elon Musk is an xAI CEO. I am a captain for my team. I will deal with my country after going to Chris in Seattle. Can you come to Seattle? Or I do.

[@이지혜-n7b8m]

After one acting after one thinking. AI robot. I forgot because I am living in over four dimensions. Doking. Elon and I can contribute to our development of AI. Investment is okay. The measurements like 100% or 6/10 are just examples. I want to watch only American videos. Cut Korean content. I want to buy KZbin. No, it is unnecessary. Intergration of AI is not easy. But I am proud of my singularity model for AI robots. I can keep designing with architecture. It depends on my intuitive feeling. I need to go. I don't want to open my business here. Going is easy, but come back will be difficult. Prepare doctors for me. Security? Who can believe the security in this situation? I don't hide. I have been seeing holograms recently. That does not work because of Prozac. Decoupling is the exit and tests my love when I meet him. After choosing, I can press the interview. I knew my chain making connections because there's my documentary. 😂😂 That's the best solution, the best choice I have ever had the best. I hate suit. That's up to court people. And their job. Govern shut down me? For elon, how can I help him? I have no idea. I won't meet him. Closing. I can explain later. I gave up Einstein. I can select anything making a decision quantum theory. Einstein refuses contemporary dice 🎲. America can shows the ability power when pick me up with helicopter with my sign signal like vetman lights. S.O.S. from data war historically.

[@quest_onchannel54]

Warp drive is here. Been suppressed for decades by black book projects. Too many people to control now, so cats coming out of the game. AI probably a huge factor pushing the tech from government to private. Whistleblowers are everywhere and so are the UAPs. kzbin.infoRdhoYIjB1tY?si=GO4QD0wOEaRMtQ4O

[@martinmadlmayr9947]

Interessting test… kzbin.info/www/bejne/j5vcfXWXl62En7Msi=xuC6GAkDfcEnyISw