New Ethereum talks, every Monday. The week's conference uploads by event, in your inbox.

Loading player…

Greg Osuri - Democratizing Access to AI Resources - TOKEN2049 Singapore 2024

TOKEN2049Wed, Oct 9, 2024, 12:00 AM

Transcript

today I'm going to talk a little bit on the on the reason why we are here why we do what we do because I feel like you know crypto in general has kind of lost its way at a little bit uh in to getting too much into the weeds and really forgetting the roots uh of what created this industry so looking back into uh the history of of computing uh reviewing the current state and really showing where we're going right so this talk is not too much on the AI side although I'll touch a little bit but really on the general framework general principles how uh how we got to doing what we're doing my name is Greg i'm the CEO for overclock labs and the core contributor to Akash network uh been a programmer for most of my life and prior to Akash I founded a company called Angel hack uh I'm an open source developer uh being uh ranked as one of the top developers in the world for code I've written libraries have published um you ranked as number four on um on GitHub golang trending uh category was the only human being here I believe in between Mozilla and Google um a lot of the code I wrote is now part of Nvidia part of US Department of Defense part of AWS a lot of that source code which we originally wrote for Akash uh it's kind of incredible to see how it comes back to actually us repurposing some of the Nvidia code as well so it's like gets into the cyclical uh dependency anyway the talk today the outline really is about if you really think about how the birth of the internet uh was born out of decentralization and the need for decentralization and that transition into what we have which is extremely centralized right um it really began with with mainframes right if you comes come down to first time computation was used at scale main frames where you have to go to the system literally borrow time from the system in order to use the system which led to innovation in personal Computing uh more distributed Computing in the hands of people that really created the internet on the foundation of decentralization connecting these individual devices in a hive network sharing intelligence and removing the need for a centralized coordinated system to communicate it gave birth to the internet or based on decentralized principles and historically decentralization is what really spun uh The Innovation and the accessibility that we have today uh but eventually somehow the centralization came with large hyperscalers and uh large uh data centers essentially filling the need for coordination right so as devices got lot more smaller uh lot less powerful the the the new layer called the cloud kind of came around consolidating this power into few companies really uh companies we call them hyperscalers today Amazon Google and Microsoft the world uh if you look at the it spend uh over closest 70% of all the it spent is spent by four companies so that tells you the kind of control these companies have it kind of gives you an idea how the vendors to the companies are optimizing Hardware to serve the hyperscalers and not really to serve the General Public are are are are the common good right that presents a few challenges when it comes to sovereignty because it's no longer the case when you have smaller data centers uh that can Place Large orders uh into those vendors because a lot of the vendors are really optimizing for for hyperscalers and that Gap is actually widening quite a lot it's very concerning and if you look at uh the Gated access to the resources there was a story on semaphore u i mean this is a very very widely known uh fact that if you are doing any AI development it's extremely hard to get access to this accelerated Computing namely chips like h100s or A1 100s even likes of Elon Musk who is famously known for saying getting access to GP is harder than getting access to drugs even though that's a that's a low bar in San Francisco but that's a reality right if you try to get any decent amount of compute uh to do AI training or inference it's you'll find it extremely difficult I mean until recently it was like impossible to get it on uh on on centralized uh servers like Amazon so the world I mean they will select who gets compute based on the profit margins they're able to get and if we look at a costs right I mean uh the costs are getting out of control over 50% of every single penny we spend on online services goes to one of the three Cloud providers all of us pay a cloud tax with without knowing who the who we paying the tax to so your Netflix of the world your Google of the world your ASAS of the palent even Department of Defense which uses palent CLE pays a cloud tax right so you can see here there's this invisible opaque layer that we all have to pay into to for for society to function and it's a is in and hor is this study it's called the trillion dollar Paradox uh it's a fascinating study highly recommend folks to go through that um and and and the worst part is reliability I mean how many of you heard of crowd strike or having been impacted by crowd strike quick show of hands here I never heard about the company until recently uh crowd strike happens to be this company that provides critical software for Airlines and a whole lot of critical infrastructure services that going down literally made thousands of people cancel their flights without food without accommodation delt alone had to cancel 7,000 flights because somebody pushed bad code to production it's not a state actor attacking in our infrastructure just somebody pushed bad code that's how opaque This Cloud is and none of us know what's happening right I mean this is not open this is controlled by a few people and somehow this code got pushed through you can imagine what a scenario tomorrow where a serious State actor is trying to attack infrastructure what's that going to look like right and that is current situation we have and if you look at the industry practices and cloud is known for stifling in Innovation they built this competitive remes which makes it extremely hard for newcomers to come and participate in this economy U it reminds very well I mean reminds you of the railroads uh uh you know encroachment on on on stifling Innovation or in from the 1800s or late 1800s to to pre-World War I uh cloud is exactly in that position and right now US government is actually investigating a lot of these Cloud companies for antitrust um you know a lot of lawsuits uh by open source companies and the cloud companies because they tend to steal open source software uh and their business models right um I mean it's it's very obvious the central centralization leads to higher costs uh less access and they're you know stifling our Innovation and the economic economic inefficiencies and escalating cost is a real problem that we need to pay attention to um and what are the risks of going down this path what are the risks of hyper concentration of power the real risk is us as a society going back to feudalism I call this digital feudalism there was a time in history that few people had most power in Europe and peasants had to depend on those people for accessing basic things like food and water and that is a future where where we're facing right now uh and the future is Bleak where a corporation or a business dictates who gets to do eii dictates a r ration of resources uh to use Ai and we're kind of seeing that right now right you're seeing Amazon and Google of the world rationing computing power rationing or we call the spice you know but the spice must flow right so and the industry sentiment is obviously very clear people want Solutions and that's where we have we have a new era of innovation we have a New Era of decentralization uh in the cloud industry and uh and why because we believe compute should be a public utility compute should be as free as the air we breathe and water we drink and that's the future we're all building towards uh what that means is restoration sovereignty we need to have control over our data we need to have control our our computer and we need to empower communities uh rather than corporations uh to drive the technological progress um instead of few people uh having access to this compute the access should be ungated actually should be open actually should be public and and efficient right uh the margins if you look at Amazon of the world Amazon is the Amazon web services unit is alone a a a trillion dollar business for Amazon right and profit margins are over 60% and that's not okay uh when you you can't have a normal ly profitable uh layer and you call it a public utility for example the cost if you look at uh decentralized systems like aash is 80% cheaper than what you would otherwise pay on Amazon and that's because there's compute available everywhere you don't have to depend on you know these hyperscalers to drive that that depend on margins to drive their business and reliability I mean we don't have to rely on these opaque Services open source systems are very capable in a case like aage we have been live for about four years and there hasn't been a single time the network went down or slowed down it's always available it's always live and that's possible because it's open source people are looking at the source call all the time people are pushing updates all the time it's decentralized there's not a single system that controls uh the entire network but rather a distributed nodes that all coordinate together that control the or that run the network right you have this New Era of innovation that really looks like this right this is a a new industry called deepin or uh or decentralized physical uh infra infrastructure networks um Akash happens to be a category Creator and you have all layers of cloud infrastructure all the way from general purpose compute where Akash and you know flux and a whole lot of uh uh you know players U are are are are innovating to storage networks like filecoin to machine learning networks like bit taner yeah this new layer of infrastructure called deepin and that's very live that has product Market fit and that's growing really well in a case like Akash which I happen to be a core contributor of Akash is the world's first decentralized supercloud that relies on a distributor network of cloud capable clusters and leverages underutilized compute it offers that in a perm missionally decentralized manner um the way it works at a higher level is the users uh that require compute Define what they want uh in a in a format called sdl Place their order on chain and providers that can offer this this compute bid on the workload in a reverse auction mechanism upon which the tenant selects the provider that they want to go with uh and a lease gets created and execution uh post lease creation happens offchain whereas coordination and payments happen onchain in a decentralized manner uh the Nvidia I mean the incredible ecosystem Nvidia happens to be a biggest user of Akash in fact uh Akash is the only crypto protocol Nvidia uses uh it's Akash is integrated into Nvidia products and uh incredible companies like uh you know news research which happens to be a top tier uh AI team uh that's working on decentralized training to University of Texas to Roch of Technology uh you know Akash has a a has a wide uh ecosystem in the traditional AI world and also in the decentralized world some of some of my favorite uh happens to be uh Venice AI which is a company created by Eric wores U and it's a privacy optimized chat GPT replacement that runs on aash which I personally happen to use as well um the sorry growth has been incredible I mean this quarter alone we saw 50% growth in the Lees that were created uh from from last quarter and U and that is uh bound to increase this quarter as well uh so before I leave the question I Really Leave is uh with you is as we stand in The Crossroads of technological development today the choice is really ours to continue down the path of concentr of of centralization or to embrace decentralization and unlock a new era of uh Innovation and and empowerment with that I conclude thank you so much

Automatic transcript — names and jargon may be misspelled.