Will AGI be Open Source or like Open Ai's ChatGPT?

Open Source AGI is what we want, but is it what we will get? OpenAI and Gemini cloud services would like a word with your credit card please. Will AGI be Open Source is the question and how could we get there.

Digital Spacecast https://www.youtube.com/playlist?list=PLarJAzZsWRGAYn6uDW5a2xeX7aNxmSvs9

5090 32GB https://geni.us/GPU5090

🤖 Budzo the Local Ai Server
GPU Rack Frame https://geni.us/GPU_Rack_Frame
Supermicro H12ssl-i MOBO (better option vs mz32-ar0) https://geni.us/MBD_H12SSL-I-O
Gigabyte MZ32-AR0 MOBO https://geni.us/mz32-ar0_motherboard
AMD 7V13 (newer, faster vs 7702) https://geni.us/EPYC_7V13_CPU
RTX 3090 24GB GPU (x4) https://geni.us/GPU3090
256GB (8x32GB) DDR4 2400 RAM https://geni.us/256GB_DDR4_RAM
PCIe4 Risers (x4) https://geni.us/PCIe4_Riser_Cable
AMD SP3 Air Cooler (easier vs water cooler) https://geni.us/EPYC_SP3_COOLER
iCUE H170i water cooler https://geni.us/iCUE_H170i_Capellix
(sTRX4 fits SP3 and retention kit comes with the CAPELLIX)
CORSAIR HX1500i PSU https://geni.us/Corsair_HX1500iPSU
4i SFF-8654 to 4i SFF-8654 (x4, not needed for H12SSL-i) https://geni.us/SFF8654_to_SFF8654
ARCTIC MX4 Thermal Paste https://geni.us/Arctic_ThermalPaste
Kritical Thermal GPU Pads https://geni.us/Kritical-Thermal-Pads
HDD Rack Screws for Fans https://geni.us/HDD_RackScrews

▶️ 5060Ti Full Review https://youtu.be/sCc-y4LXfHs
▶️ $750 Home AI Server https://youtu.be/3XC8BA5UNBs

Ways to Support:
🚀 Join as a member for members-only content and extra perks https://www.youtube.com/c/digitalspaceport/join
☕ Buy Me a Coffee https://www.buymeacoffee.com/digitalspaceport
🔳 Patreon https://www.patreon.com/digitalspaceport
👍 Subscribe youtube.com/c/digitalspaceport?sub_confirmation=1
🌐 Check out the Website https://digitalspaceport.com

*****
As an Amazon Associate I earn from qualifying purchases. 
When you click on links to various merchants on this site and make a purchase, this can result in this site earning a commission. Affiliate programs and affiliations include, but are not limited to, the eBay Partner Network. 
*****

Transcript

Speaker: 0:00

Will AGI be open source? Will Chinese AI and AGI overtake the world? And how is the race for AGI going to impact you both financially and psychologically? Today we're going to dive into a different format of videos where I'm going to actually explore some of these topics. There's no B-roll. This is actually also available on the Digital Space Port Podcast. You can find that in the links below. And I think this is one of the definite places that I hope to see some comments from users on important topics that definitely are coming down the pipeline that I think are good to start talking about now. So let's start with the first one. Will AGI be open sourced? I think there's a lot of indications that AGI and the direction that we're heading towards artificial intelligence developments are led from open source initiatives. However, we also see a very surprising trend in many companies which seemingly release quite a bit open source and then choose to keep some of it back. And what is that trend? The trend is monetization. And they understand that they have to keep back the best. They have to keep back the most productive. They have to keep back the leading edge. And the reason why is because they're turning this into products. If you look at things like Quinn, they have a really great suite. Today, this is actually one of the topics that jumped into my mind because they just released a voice mod which really gives them the full capability suite to enter into voice production with a captive audience of Chinese, of course, but also they have really good products. Now we see things released that are really killer from Quinn. I don't want this to come across as a dig at the Chinese and their open source initiatives. I think it's very good. It pushes the ball much further down the road. But when you look at things like will that actual AGI get released open source, there's a very clear answer. And the answer is no, it will not. Not for many of the companies that are currently potentially in contention for being able to hit artificial general intelligence. Now, the timeline for AGI and my estimation is still a couple of generations of hardware away at best. And the reason why is we simply have not enough bandwidth as silly as that sounds to say, we probably need two generations more of bandwidth to really be able to see a capability to move fast enough. Much of the math in machine learning is not new. Most of the concepts have been around since even the 60s. And what we've seen multiple times is stalls. Most of the stalls that we've seen have been highly related to the fact that we plateau on either a technology, speed especially, or the ability for the math to move past that. When we saw transformers, we saw a revolution. Is this what brings us further is tick token the future? These are really good questions. And I think the answer to those both is most likely no, I think we're going to see rapid advancements. But there will be a plateau effect. And I would actually suggest we are already very much hitting it. It feels like it. The past six months have felt like a plateau effect without really having a huge major breakthrough. Deepsea car one was let's be very crystal clear here, a major breakthrough. Now we've seen before that, oh three and stuff from open AI, really good, really solid. GPT five though. Let me let me know what you think about GPT five. If you're a regular user of GPT five, I have open router. So every now and then I you know, go and check out things like there's a new Sonoma thing that I think is rumored to be grok for on there. Really cool because those are free has huge context window. Is that context window meaningfully useful? Well, I can tell you that we can't get to AGI from context window alone. That's for sure. We actually have to have fundamentally different math powering what is going to be hardware that is faster in a generation or two. I think we can see the convergence of that. And I think open source does lead the way there. Do I expect however, whatever company is first to arrive there, to be releasing that to me and my garage to achieve garage AGI. My realistic expectation is actually no. And that is where looking at how AGI is useful or AI top tools and products are useful now is kind of some of the refocus that I'm doing in the channel. It's become apparent to me that we probably are hitting a plateau period. So maximizing how we are prompting maximizing how we are utilizing maximizing the exposure to the number of tools you guys have. That's one of the main goals that I have right now because that's bringing utility out of people's investment into hardware. And I'll speak a little bit about open source and hardware. The demands for VRAM are, of course, incredible. There's no such thing as enough context window. There's no such thing as enough VRAM. And literally everybody out there can't afford to put the leading edge systems in the garage. Because seriously, I don't think Nvidia is going to sell you just one single DGX B200 or something like that. It's kind of funny. But at the same time, the leading edge hardware is going to be very expensive. And it's going to be in the hands of either the largest corporations out there that are working towards making sure that this is developed in a really sustainable way for them, which means they have to have some profits to show for this. You can't just lose money for infinity. And we are entering into a realm of the AGI race that we see expenditures on new hardware, data scientists, literally just the brains that can make the bits happen through the roof. Of course, we see some players floundering. I don't think met is doing too hot. We see some treading water and actually being remarkably sustainably growing in a right direction, even though they're still reportedly upside down, would be kind of the anthropics and the open AI is out there. We see some as emergent new disruptors, but they also appear to have hit a plateau. And I do actually think that plateau is scaling. That would be XAI and Grok. We see a cohort of really good innovations. The latest biggest news has actually been the open source kind of catch up that the Chinese companies have been releasing companies like Tencent, Alibaba. These are huge corporations, and they definitely have monetization in mind. And turning AGI on for a billion people has a substantial price tag attached to it for them. But I think that it behooves us to be realistic in our expectations for what happens in a garage setting or in a house setting. Is there eventually a demand that is met for AGI? Let's just talk broadly here. And I'm giving my feelings. I think you should give your feelings in the comments below. Is there a demand that's met for a household consumer that actually reaches AGI? Or does it stay in the cloud behind the largest data centers in the world, literally locked to the bandwidth that's able to power it? I think we probably see something that is very similar to that for many generations after AGI is achieved. That would probably be a couple of generations in addition to the couple of generations it's going to take to probably reach AGI. So realistically, do you want your calculator in the cloud? Of course you don't. You want it on your phone. And a calculator is not a great analogy here because what we're dealing with is substantially more hardware intensive. However, for generations from now, you probably could be looking at substantial on-device capabilities with specialized chips. And I think what we consider state-of-the-art then in the household will not be what state-of-the-art is in the cloud. So I do think there will be a growing diaspora between the two. I do think there are some really innovative companies out there that have an approach that very well could lead us towards a real true open source implementation like the Allen Institute. They are thinking outside the box and they are in the minority and they are technically not leading in many regards. They have some great document stuff. You should check out their OMO OCR if you ever get a chance. It's really, really good. But definitely we also have news resource research. They're coming in and looking at making steerable AGI or AI products that really adapt to you, which is great. And I definitely want to give the 14B, which is a QUIN3 product that they just released, a check that is a Hermes class product. And I think one of my first experiences was Dolphin and Hermes when I got into AI locally. And so I think you've seen the local AI Dolphin with the muscles. That was one of the first things I saw and I was like, "Okay, I got to try this out." And I think a lot of that has continued, which is really fantastic. We still see Dolphin. I don't know if there's going to be another release, but we see Dolphin active in the community. We definitely see Hermes and the news research team seeing uplift, but is it anywhere near enough to match what the Chinese open source is out there? And I think it's absolutely not. And how does that get fixed? How could that be corrected? Well, maybe through massive investments, grants, possibly leases that are very at favorable terms to people that are research oriented within open source AGI goal. Is it important for a government to incentivize that? Let's talk about geopolitically, what the thought process might look like. I mean, this is all just guesswork, but I would suggest that no, most governments are going to want to play the game and the narrative that we see kind of fostered by anthropic the most in the US government. And that is, this is winner's take takes all. I think that's a ridiculous proposition. On the face of it, it really doesn't make sense. Because scale technology that is very hard to achieve, that is definitely revolutionary has been achieved many times in the past independently through independent researchers outside of the initiator of the first across the line. Now, it's not to say that the first across the line doesn't have significant manage, they do. But that doesn't mean other people give up. And I don't think we'll see that in the race towards AGI. I don't even want to talk about super intelligence. Because honestly, AGI is very out of grasp. Super intelligence is super out of grasp. In my opinion, we are really far away from that, especially without some very big revolutions happening. Now, could those revolutions happen? Absolutely. Could we wake up one day and have a new deep seek r1 moment? Possibly. Was that exciting for you? How did you feel when that happened? Were you scared? I know that I had some trepidation because I saw Nvidia's stock price. And I was like, well, if it doesn't, I just assumed the narrative the media was giving, you know, they're very brash kind of first take was correct, possibly. That was absolutely probably not correct. Granted, yeah, it costs much less to train. But you make up for that on inference side. And inference is going to be quite big if you have tons of customers. But really, it comes down to utilization. Can customers use AI products? Many people are not going to be interested. Now that's in AGI products or AI products that are directly marketed to them. So I think there is a huge market for under the hood applications to be really good. But if you look at the nuts and bolts of deployment, it's a special kind of class of person. I don't think most people are out here like let's build an AGI rig. The popularity that we see in especially things like the Apple m3 m4, even back to the m1 and m2, all of those popularities are because people can just buy thing, they don't have to build thing. And there's a ton of people that just you say build, they're out, they don't got the time for it. They don't have the frustration level to deal with it, or they don't have the technical competencies. Or this is probably what most of it is. They want to maximize the profitability of their time. So they look for pre built solutions that are already really good. Now we see $30,000 solutions, which pretty expensive out there. We see $15,000 solutions, very expensive out there. We see Apple coming in at about 10,000 with the m3. That is a pretty compelling product for text use. Honestly, the system bandwidth, again, I've said many times just forget AI tops, just think about it as system bandwidth. That's what really matters. Now the capabilities underneath the hood also, you know, matter a lot also. So whether you're using a Vulkan back end, which if you think Vulkan is the answer, I would urge you to go talk to the developers of AI software and try to get them to implement Vulkan. Many have many complained, there's a lot of problems still with Vulkan as a back end. And a lot of things that can happen also with AMD and the framework, you're looking at about a $2,000 rig. And you're also looking at only about 226 gigabytes per second versus the Apple's almost 850 gigabytes per second in the m3 studio. Again, $10,000 product, not cheap, 2000, not cheap either. But being mindful of what you're spending on hardware, and a time of uncertainty is something you should absolutely do. We see the price of GPUs falling for a very specific reason. Do you remember back to December? Do you remember back to January and February? Do you recall how much it was for a 4090 back then? 4090s are still almost $2,000 some of them. Granted, locally, you can get them much cheaper. But there's still if you check out eBay, there's still several models that are almost $2,000. The white Asus ones that I've got the OCs, those have held their value pretty well. But I'll tell you what I'm thinking about for sure for myself is the depreciation of expensive assets. And you definitely don't want to enter into debt to play this game. That's just a fact. You want to be as debt neutral right now. This is just some take that I have on it as you can be. And the reason why is because we could have shocks to the system economically from many factors way more than just AGI. We're talking food shortages, we're talking climate, we're talking weather, we're talking actual developments in AGI displacing potentially. I mean, there are a lot of shifts that could happen. Granted, I think most of them are about five to 10% as impactful as what you're going to hear the news media say. Most of these things are not going to hit everybody. But if you're thinking about what's going to happen in the next year or two, it does behoove you to try to reduce debt when you're thinking about your exposure to things because that reduces uncertainty for you, which is an important thing that I would like to convey. Also, when you're thinking about hardware, that's important because the depreciation cycle on let's say a 3090. I still love my 3090s. Budzo, I have a name for my AGI machine here. Budzo. Budzo has four 3090s and Budzo, if you're interested, you can find out more information about that on the website, digitalspaceport.com and the video that we released and all of the videos feature Budzo for the most part. But 3090s have long-standingly been the best go-to. Those things were almost $1,200 back in January and February. Now, a year ago at this time, they started to creep up to the 700 range. There was a brief period of time almost exactly when I got into this. They were about $600 to $650. We see some in that price range now. So they've held their value remarkably well. But what you may not be paying attention to is the reasons why they have been even lower in the past. So I do feel like there's a future where we're going to see 3090s not in the real near term, but in the near-ish term drop in price even further. So 600 probably is around the corner for 3090s sometime this year. 4090s, I'm not sure how they're going to hold added value. The reason why is because while the concept of a lead GPU is something I think is very valuable to people, at the same time, you see a 4090 having a substantial price tag for the same amount of VRAM. And that same amount of VRAM doesn't have substantially better system bandwidth. You're still at like a terabyte a second. You're 3090, you're at like 938 or something gigabytes a second. So almost a terabyte a second. You do have better activations. So you can actually get some speed ups that are pretty substantial with things like FP8. But at the same time, most of what you're going to run out there is also going to work really well on a 3090 for inference workloads. Now if you're doing image generation or video generation, certainly having a single 4090 could make a lot of sense in that instance, but not at the price they're at. So I think before we see the 3090s come down a substantial amount, we'll see them come down a little bit, but not a huge amount. We're going to see a massive price collapse in 4090s as reality hits of what they're actually worth. And the reason why is because 5090s are now only about 22 to 2300. And you can check the links below if you're interested in finding out more about that. But definitely that is going to create pressure on the 4090s as it should, especially as we see more kind of FP4 optimizations come down in some of the models. Now Blackwell has been plagued with some driver issues. And of course you've got to use open source drivers. I've said this a couple times in comments and a lot of people have asked, got to use the open source drivers and Linux to really be able to use this. But that's going to create pressure on the top end GPUs out there. So I think that we will see 4090s have a pretty big precipitous fall. I'm myself, I'm planning on how I can offload two of my 4090s. Maybe two, definitely one, maybe two. I might keep another one around just so that I have a 3090, 4090, and 5090 to run comparisons on. But I'm not sure that's even that important because for most workloads, I can just tell you and it actually is, you're going to see about the same tax performance for chat generation on a 4090 as you're going to see on a 3090. So I do think there's a lot of interesting things that could happen with some of the pre-built out there from some of the manufacturers. I think AMD and probably given the system bandwidth that we're seeing reported with the DGX that's about to land, we're not seeing good enough system bandwidth. The M3 is really kind of the only class that really gets to the point where you're like, "Whoa, that's really good bandwidth." And the ability to utilize a decent context window for people is something that probably impacts them a lot because you want a large context window. But the context window in many models degrades quite significantly. To my knowledge, only Gemini has maintained their context accuracy throughout the entire million context window that they offered. That's Gemini 2.4 Cloud product. And we definitely don't see that in Gemma 3. I can tell you that for sure. As a user of Gemma 3 27B, I can tell you that we do not see a context window that doesn't have degradation. It degrades pretty rapidly. So optimizing for context window also, somebody asked a really good question, "I want a million context window." Well, you want a million context window that doesn't degrade like crazy. That's what you really want. And there's another thing you really want. And that is the ability to have better prompting skills. And some of the things that we'll talk about are kind of geared towards that. The ability to correct over the course with prompts, if you've been prompting very much, you definitely know, it's pretty hard to get it to course correct if it starts down the wrong pathway. And so a lot of strategy goes into how to correct the initial prompt until you're very happy with the results that you get from that before expanding upon the next one. And that can actually speed things up quite a bit also if you're doing KV caching, so you can get some pretty good results there for local. I think local also has a huge implication for voice applications. And I think we could see huge success there from companies that are releasing products that are geared towards that, especially in the home assistant realm. I've had tremendous problems with getting a stable home assistant running with whisper and Wyoming and all the stuff. Home assistant is convoluted. And if you've used home assistant, I don't think that's a controversial statement. But also you can't get the performance you need on a Raspberry Pi. Like it's created its own problem set almost. And the way some of the custom built hardware that they've produced has problems. So I think that we could see huge changes as far as locally hosted AI in that regard that would be must have products clearly that demand is there for voice integration. We don't really have anything that's great. And we just actually saw I think I think something that's pretty cool, which will be our final segue that we'll talk about here. And that is how corporations react to wild successes. And I think this is interesting because we see in the US, something that just happened, which has kind of happened before. So Microsoft vibe voice, one of the top repos on hugging face for almost a week. Everybody was really having a lot of fun with it. People enjoyed it. They released it under an MIT license. They appear to not have understood that releasing it under an MIT license meant the cats out of the bag essentially. And so they've pulled back their official repo, you can find an updated set of instructions if you were interested in running vibe voice on digital space port comm it's on the front page, second row, last one there. But definitely you see the reaction being, oh my gosh, we have to curtail this. So releasing something MIT and then trying to police it after the fact doesn't work. And I'm very surprised that Microsoft didn't know that Microsoft really does need to have some sort of an outreach specialist, who is a liaison in open source affairs, in my opinion, who can publicly talk to the people out there. We have that at Google, we have that at several other companies that tell lesser degree, Logan Kilpatrick is certainly a great person at Google that reaches out to quite a few people. And he has a great presence. He lets people know what's going on. Google itself has a hard time messaging some of its amazing products. And they've built some user interfaces that I think everybody that's used them would say a little bit on the what side and you don't want to have to set up a developer account and all this other stuff. So they have their own challenges for sure. But if you look at where the messaging came from on this, it essentially they didn't even put out any message, they pulled the repo put out no message. It wasn't until there was massive outcry that they actually put out a message. So just corporate PR wise, this is the response to a wild success. And it's confusing to a lot of people out there. Now, I would also say vibe voice is really good. It's one of the best that we've seen out there and what it does for text generation or text to speech generation. I think it's actually quite fascinating. And yeah, some of the artifacts are a little annoying that it does. You don't want to enter exercise mode. We're plagued with poultry guys since we did this. I mean, our house is just absolutely plagued. It's like poltergeist the movie. But definitely, exercise mode was kind of funny. There in like it's controllable. If you don't diverge your text into something that is a little bit off topic, it'll it'll stay with the script. If it's a cohesive script, it'll stay with the script. If you have an AI do an extension prompt extension, and rewrite your script, it's probably going to come up with something that is going to be followed. But the reaction to a success being that from Microsoft, absolutely puzzling. And definitely they said there were some problems with you know, voice impersonations and stuff like that. I want to read what you guys think about this. But of course, there was always going to be would be my takeaway from it. Like, did you understand what you were releasing? It's baffling. So the alignment, the direction of even large corporations like Microsoft, and the race for better and better AI tooling, which I think I would term vibe voice AI tooling more than direct competition or pathway to AGI, this is more of a quality of life improvement, honestly, allowing things like virtual production to actually be doable in a home setting, which really, this is high quality virtual production that you can do now really opens up quite a bit. And you can bet we're going to explore a lot more of that come more tutorials to do I've got more builds to do. But definitely this is going to be one of the more kind of common themes that I'm going to do is sit down. Maybe it's Mondays, can't guarantee when, but whenever it is, sit down and have a long form conversation with you, which you can also find at the podcast links in the description below. This gives you a no visual needed interface. And I think voice being part of what I think is important. This is just me following through and releasing this podcast out there hopefully is impactful to you can find it on Spotify can find it on Apple Music and whatever podcasting place you get your podcast from. So I hope you've enjoyed this. I look forward to reading what your comments are. I want to again, of course, as always, throw a big shout out to all of our channel members. You guys do actually help me buy new GPUs. It's called buy me a GPU on buy me a coffee, by the way, for a reason because I use that money to help buy GPUs and I will be getting a new GPU soon. Drop your guesses below. Everybody have a great one. I'll check you out next time.