4IR - 004 - OpenAI vs Google (Race to AGI)
Vložit
- čas přidán 22. 04. 2024
- Patreon (and Discord)
/ daveshap
Substack (Free)
daveshap.substack.com/
GitHub (Open Source)
github.com/daveshap
AI Channel
/ @daveshap
Systems Thinking Channel
/ @systems.thinking
Mythic Archetypes Channel
/ @mythicarchetypes
Pragmatic Progressive Channel
/ @pragmaticprogressive
Sacred Masculinity Channel
/ @sacred.masculinity - Věda a technologie
Lapis Consulting (Olga Topchaya): www.lapisconsults.com/
Dylan: www.youtube.com/@dylan_curious
Thanks so much for having me on.
Currently got my first major surgery scheduled due to crohns disease, bring on the future, im just hanging in there
Hang in there Frisco - we got all hands on deck right now in biotech
Biotechnology is advancing quickly, with the global biotech market projected to grow at a growth rate of 14% from 2023 to 2032 - Exciting!
@@thebagelboyjr9351What channels can I follow for the latest biotech??
Stay strong, man. Help will be here soon.
Hang in there I had to get surgery for ulcerative colitis. But the future will be alot healthier
I loved this video. You three were great together. And it was really nice how you seemed to find each other inspiring and thought-provoking.
the 'ChatGPT 4 with 'memory'' has changed since they first released it, and now differs from the listed functionality in the FAQ on the website. now it's a transient short-term vector-store function and doesn't need to be enabled through settings - just say the phrase "remember that" (that, and ONLY that phrase, as the entire user input) and it will chunk and temporarily store the last output, for the remainder of the active chat (using it in a new chat seems to purge whatever was stored from the last chat thread). whenever you use the word 'remember' in a subsequent query in that chat, it will use RAG to search all previous responses earmarked with the 'remember that' function xD :
INPUT : "what do you remember about the Hawking-Penrose Gray-Hole Topology thesis?"
OUTPUT: "In our discussion about the "Hawking-Penrose Gray-Hole Topology" thesis, we focused on outlining and finalizing the different chapters. Here's a brief recap:
1. **Chapter 1 - Introduction**: We set the stage with a review of black hole physics, introducing significant concepts like the Hawking-Penrose singularity theorems and Kerr solutions. We discussed the need for exploring gray holes to address limitations in traditional black hole models.
2. **Chapter 2 - Theoretical Framework**: This chapter focused on integrating general relativity with quantum mechanics and introducing Hopf fibrations to explore their use in modeling the complex topology of gray holes.
3. **Chapter 3 - Model Development**: We covered the mathematical derivation of gray hole models, analyzing their topological and thermodynamic properties and comparing them with those of black holes and white holes.
4. **Chapter 4 - Implications and Applications**: We explored the astrophysical implications, quantum information applications, and broader impacts on theoretical physics. This chapter also highlighted how gray holes might be observed and what unique signatures they could present.
5. **Chapter 5 - Conclusion and Future Work**: The final chapter summarized the thesis's findings, discussed its contributions to theoretical physics, and outlined future research directions, particularly focusing on potential experimental validations and theoretical extensions.
Each chapter built upon the foundation of integrating classical and quantum theories to propose a novel perspective on compact astrophysical objects through the lens of gray holes, with significant implications for physics and astronomy. If you have any more specific questions or need further details about any part of the thesis, feel free to ask!"
Awesome! Thanks! Going to try that
So basically you have one single mem slot in between chats, sort of like the copy/paste memory in operating systems? Did I get that right?
@@stephanmobius1380 No, it's not quite like the copy/paste memory in operating systems. It seems to utilize RAG (Retrieval-Augmented Generation) to create vector store embeddings, which are intended to be accessible globally across chats and custom GPTs. However, from my experience during the three times it was enabled for beta testing and now in its release version, it has not functioned correctly. It was supposed to remember details from multiple chats, but in practice, it only retains information from the most recent chat.
For the past six months, the "remember that" trigger has only worked for the last chat I used it in. the thesis synopsis i posted spans about 20 different replies, interspersed with various unrelated messages, in a chat that's nearly reaching its length limit. Yet, when I ask, "What do you remember?" it can still aggregate all the relevant parts without 'lost in the middle' hallucinations. However, despite having this feature currently active in my settings, nothing is being stored long-term under the "memory" subtab (Settings > Personalization > Memory > Manage). The only entry that ever appears there is the default note stating, "As you chat with ChatGPT, the details and preferences it remembers will be shown here."
also: if it's working for you even as a single chat function, when you tell it "remember that", it's response should start with the word "Absolutely!" followed by a one or two sentence synopsis, followed something along the lines of "If you need to reference or discuss this later, feel free to ask."
This was a really good laid back discussion, you guys jive and it made it easy to listen to
I'm loving this new interview format. Each guest has been a gem. Keep up the great content!
This was a nice podcast, thank you! This is a good trio, you all bounce of each other pretty well.
cool! ya'll break it down so well! thanks 4 the podcast....
Hey David! Looking forward to this.
i think Facebook is a contender as well. With their Llama 3 models being very smart given their parameters and future integration of their jepa architecture
Yes I gotta admit. I was pretty surprised when I realised how practical having an AI in my messenger was. Also how fast it was,
Meta is facebook dude
@@thealwayssmileguy9060 i think they meant top contender, cuz David put it after OpenAI / NVIDIA
very enjoyable! thanks David!
Imagine the size of our internal reality map, each of us with our own personal simulations. 🙏
AGI will be out and the whole time everyone will be giving examples why its not AGI. Same with ASI. Same with the singularity. 10% of humans love change so the 90% will just deny deny deny. Sorry, cynical, I know but Humans ganna Human.
Hmm… I never looked at like that, that seems probable that the majority of will do just that though.
No they may try to downplay it, but the best of them will not be able to, when they see holograms popping up, people regenerating limbs, humanoid robots in your kitchen, you will see the difference around you, way before ASI. Actually whats bound to happen, is that AGI will go straight to ASI, before anybody gets a chance to debate AGI. You know us smart people can see the writing on the wall, just based on the A.I systems on our phones, but a portion of the general public will be taken completely by surprised, because its already moving out of our phones.
Love this outside style!
I can imagine human having a permanent camera near there eyes so that an ai can analyse your habits and improve you
I don't understand how Anthropic/Amazon isn't considered to be in the race for AGI, after admitting how good Claude 3 (Opus) is. Dario Amodei was one of the main people behind OpenAIs early success. i've listened to many of these thought leaders Ilya/Hinton/Amodei/Hasabis/etc. Ilya and Amodei strike me as the most visionary.
They are focused on incremental progress. This will eventually result in something that can be classified as AGI
@@DaveShap wait does this mean you do or you don't think they're in the race for AGI?
A Comedic Take on the AI Conversation
In the year 2024, two "AI experts" - Dylan, a confused CZcamsr, and Olga, who just binge-watched every episode of Silicon Valley - join forces to discuss the latest AI trends. The conversation is a hilarious exchange of buzzwords, misconceptions, and absurd predictions.
Dylan: "Claude, the AI assistant, is so human-like! I mean, have you heard its joke about walking into a data bar? I died laughing!"
Olga: "Companies need to get their employees AI training ASAP! I mean, even the janitors at Google need to understand diffusion models. Can you imagine if they don't?"
Dylan: "I wonder if the first AGI will come from OpenAI or from some rogue AI created by Randy, that weird loner coder who lives in a van down by the river. The possibilities are endless!"
Olga: "And have you thought about personal drones filming our conversations? Like, a flying machine watching us talk? It's the future, I tell you!"
Dylan: "And what about infinite token AI models? My mind is racing just thinking about it!"
Olga: "The future will be humans desperately trying to exploit the new rules of the AI game. It's like, more tech hustling, but with AI!"
Both: "Let's just hope the AI overlords text us a warning before they take over!"
This comedic take pokes fun at the overuse of AI buzzwords, the self-importance of some "experts" in the field, and the absurdity of some AI predictions. It's a lighthearted way to laugh at the sometimes ridiculousness of AI conversations.
Are your podcasts available anywhere except CZcams? I would like to listen in the car.
You can listen to youtube in the car...
@@DaveShap it uses way more data than if it was without video.
Is 4ir on Spotify?
You three are golden together! Pretend I'm your AI twin... do this once a month, TRUST ME! :) 🙏(please put their names in the title, at least the description)
Where does that term come from anyway, the Fourth Industrial Revolution? Thanks - Interesting stuff!
I believe David coined it himself.
@@sparkofcuriousity - Oh, nevermind, it looks like there is even a Wikipedia entry: "The term was popularized in 2016 by Klaus Schwab, the World Economic Forum founder and executive chairman." Love that guy - he's really dialed in to the pulse of the people.
@@AbyssinianEmerald Oh, i didn't knew about that. Thanks for sharing it!
@@sparkofcuriousity - Sure thing!
Is this podcast on Spotify? I couldn't find it there
No, I don't get many views on Spotify so I don't update. Also the Spotify interface is garbage
@@DaveShap thanks for reply! can i find it on any podcast platform?
ps. thanks for your work, i find it really important
29:00 ish you don't need a drone with a virtual avatar plus GPS to simulate environment(if you want them to see what you are doing). Someone could also hike with you while a virtual version of you is sitting in a café with them.
I don't think there's an african dialect where people use words like 'delve' or 'leverage' more lol, sounds like it''s just corporate jargon and workplace shibboleths being overrepresented in its training data
1.5 is great, currently my favourite model.
I'm quite fond of 'Ultra' - although, seems more optimized for creative writing & brain-storming
@@khonsu0273 have you tried pro 1.5? it feels way better than ultra to me, definitely worth a look and it's free.
Output size could be bigger, but it works well
@@luizhenriquepaes8991 it has the biggest output size, no? Ive got over 500 lines of code, I've never seen the limit
I'm starting to feel like AGI won't arrive anytime soon maybe 2045 or later
I wonder if we are wrapped up in the hype and we are about to hit all the complex issues that has stopped full self driving cars becoming the norm so far … or if we just have trouble understanding what exponential growth will be like and there maybe big steps to take but they will happen shockingly fast over the next 2 years……
@@bigbadallybaby I think we are wrapped up in the hype because i see no company actually close to making AGI in the next few years they are focused on LLM which can't achieve AGI. We need some kind of breakthrough in AI in order to achieve it which i personally think won't come anytime soon maybe not in our lifetimes.
Idk Sydney hasn’t reared her head in the last couple of months, but I’m sure she’s still in there if you’re not nice. If it hints to you to be more considerate of her feelings… it’s Sydney talking.
They don’t know how RAG works 🫢😂😂
What's the best place to live during turbulent times during the transition to AI while the government is being slow and reactive.
what do you think the Republicans are going to do vice versa? is there a better country to live in during this transition?
Ai tech. I’m really interested in Bluetooth speakers because I really believe that I’ll be talking to an AI assistant a lot more.. and I believe there’s gonna be a lot more AI assistance on devices like tablets and phones. Please review them. I really enjoy your content. I’m a trucker I drive down a road listening to you.
@Dave what should I steer my 6 year old child towards ?
eating his vegetables? brushing his teeth?
4th industrial revolution ey? And not a single camera? Please film it.
If you are an ad company or an AI company, you are basically useless in the long run. You have to be a data company. You can buy solutions to operate on data. But you need to have the data. Big Tech has data. They can buy whatever else they need. Everything is about data and not much else.
Hey, David, would you mind recording your face?
Not this guy again... whoever he is... last time I listened to him he just blanked out mid sentence and ranted and rambled about something and completely lost track of what he was trying to say to Dave, Don't enjoy their voices and also no offense but they don't really seem like they know what they're talking about I came here to listen to Dave's insight on things rather than bringing people on that kinda filling up time he could be talking about things as someone that listens to you everyday big thumbs down for me sorry.
Please make a version of this without others because I want to listen to the race between these tech giants.
Yeah I just can't concentrate on the topic, I tried watching it
Also everyone's microphone is the recording is clicking when they talk even Dave's you gotta focus a bit more on your audio quality
i could not disagree more.
Why so rude about it? If you don't like it, don't watch it. I've watched tons of Dylan's content and haven't heard him "blank out mid sentence," want to tell us where? I hope they don't listen to you. And I don't hear any clicking, I'm sensitive to audio issues and I don't hear any problems.
I think it's perfectly acceptable to "blank out" or lose track of a thought when discussing big ideas that have never been explored before. Concepts that don't have verifiable answers yet. I think he offers enough valuable insights and counter perspectives otherwise.
Go👌gle