Bay.Area.AI: AI Alliance: Community is the Way, Alexy Khrabrov
Recording: Bay.Area.AI: AI Alliance: Community is the Way, Alexy Khrabrov
my name is Alexi kov as you can see I wear many hats I'm a director of Open Source science at IBM research uh which is an initiative to connect open source developers and scientists and it's actually under non Focus right so it belongs to nonprofit and uh I'm also a founding chair of J commment at Linux Foundation which is uh part of the lfi and data foundation and this is basically LF generative AI community and recently we set up this AI Alliance and I'm head of community for that from IBM and I'm going to talk briefly about this this is today's metup is area AI mop we just reached 5,000 members so this is uh the oldest a mop continuously running from 2015 right and we deeply technical we've always been doing this we started as an lpop on uh you know before AI was actually uh fashionable again right and uh we run continuously life and this is our home so welcome to IBM and also we have a conference called scale by the bay which runs yearly which is a conference of many meetups including B Ai and it will be held again in November in Oakland so um so this will be a brief overview of the aiin this is a very first event F first of all IBM and Berkeley a me members of the AI Alliance who knows that the company is a member who else is a member of the a Alliance who is from any scale any scale as a member of the alliance do you guys know this uh who else right any uh yes Intel Dell anybody from Dell have folks from Dell right so there are 75 members uh and uh this is the geography and the diversity of this but you know what what I wanted to say first is we're in a very strange moment in history where you know for about 40 years we've been progressing with open source software and uh I don't know if folks remember but IBM was pivotal in ensuring that open source software succeeds who knows what with this key contribution of IBM which made US Open Source a reality Linux Linux exactly right so if not for Linux we would not have commodity PC based data center we would not have Google we would not have big data as epitomized by big table paper we would not have all the data which were Master through the big data so we would not have the machine learning as we know with it and we would not have ai so and another contribution was Java about Java we probably would not have uh I mean those multiple contributions IBM made it made sure it succeeds right so and then e-commerce and the whole thing so uh so I propose that AI is the third time IBM chooses to contribute massively to this open source Revolution but as opposed to the previous times right so Linux users by Microsoft for like bety reasons and IBM actually successful fan in them mo Linux Foundation was set up right to promote this Linux uh however we never had this philosophical opposition to open source which started to happen with AI we never had the intellectual framework which argues that open source is bad right and that safety reasons not now are used as an excuse and legal capture is used to adopt legislation in the US and in Europe and other places right so we have kind of a new array of intellectual forces uh arguing that open source is bad and uh so our alliance in in a way is an answer to this we actually believe that open source is good and uh you know an example Linux Foundation gives this example that at some point uh governments tried to Outlaw cryptography they tried to regulate cryptography and it failed right the attempt to regulate cryptography failed and was basically roll back and uh we believe the same thing will happen with with attempts to arrest AI through uh through this efforts uh and so but we need to make this a reality we need to get together right why why do we need this and you can see the initial response to this Alliance was extremely enthusiastic we we already have 75 members which basically span you know the the Spectrum from Enterprises to startups and I think we have the coolest startups in in in AI right we have any scale L chain llama index user kind of facing application Frameworks we have Enterprises we have meta with llama which is I think one of the key contributions we have hugen face is a key member of the alliance we have Hardware manufacturers we have Intel Dell and AMD right on the purpose we need open Hardware we need alternative gpus right we need to democratize all aspects of AI so the the members of this Alliance spend everything from chips to apps and also they their Enterprise their startups their nonprofits their foundations like non focus and Linux Foundation itself they're all members plus we have top leading research universities uh Berle pan univers of Tokyo uh Northeastern basically if we don't have University just we didn't get to them yet we have super enthusiastic response from universities and usually the very top level do of research Vice Chancellor so and all of them are very eager to go because scientists are facing this question AI disrupt Science And scientists are the least kind of experienced people they do not run AI in production right they do not have a lot of gpus so I think there is a huge opportunity here you know to help deserving researchers to advance their Sciences of Open Source uh and so this growth will continue we have being again enthusiastically covered by the media uh this was the our launch in December uh our next batch is coming up on February 8 and it will include big names so that's is going to be exciting uh and there are three ways we uh want to advance open source AI so first of all this is a Builders Alliance right folks here building stuff first and foremost we want to build things we want to figure out which projects are the most promising for adoption by the community which will move the needle the most which will be accepted by businesses and re sure uh users and we want to get behind this right and that involves basically understanding what everybody has to contribute and figuring it out so D you know from my BM side so here's the guy who who will figure it out so help him do this and Carl with then right so have two tech leads so if you have technical ideas what we should do please ask de and Carl I'll tell them what to do uh now if you are not building right you can still enable others so if you have a bunch of gpus like IBM has contribute them right if you have resources if you don't have money or gpus but you have a lot of enthusiasm and knowledge then go teach people and go Advocate right and so like if you have very standard setup where Enterprises you know do heavy lifting send people boots on the ground gpus hardware and money uh startups do advocacy promote open source and put things on the table which actually get adopted broadly and uh also we will will do advocacy right we need to counter the uh arguments of closed kind of uh AI right we need to uh reason for open source AI but in order to do that we need to come up with technical proof that open source is safe so we need to do benchmarking and evaluation and Safety Research uh CS so all of the private C by law are required to make May profits nowed what put and the community it's a I mean it's a it's a good question but I think the same Dynamics underlines any commercial Alliance if you look at Linux Foundation it's competition it's competition and ceration right everybody wants to create a Marketplace rise and Tide lifts all the bols right so it's it's a I think it's a fair traditional discussion why open source is good for you right and we can have it separately but generally the same arguments apply right but I think in this case we have very key difference that businesses are afraid to adopt this this is not a package you download from the internet which can basically be hacked because of buffer overflow this is a thing that can slurp all your customer data and send it to Putin right like this is not uh this is not a risk you take life ping like pick pick your destination right so so this is why you don't see kids from Silicon Valley coming to a Midwestern bank and being hooked up to sapn Oracle for their financial data and CMS right nobody's going to give it to them until they know that this is not going to do something bad and we need a completely different level of reassurance testing and and and benchmarking to make sure because the model bodies pieces of the data right we don't know yet how to properly handle it so we need to actually create a lot of Open Source tools harnesses and benchmarks make sure that you know we can trust this right so I would say it's actually imperative that we'll get together and create this Market because this Market does not yet exist nobody gave a talk at any of the mops how they successfully replaced a zenes installation with ch GPT right like if you go to open it uses like some kind of JavaScript chatboard which doesn't know what they're talking about like they're not using their own tool for their customer service because nobody's doing it properly yet so uh so we're short of on time so so basically we have six work streams right and they're basically science uh V validation benchmarking uh tooling right so we need tools for developers themselves what I was talking about for evaluation we need this Hardware software could development so this is I think a big difference with this Alliance compared to a lot of software foundations we want to have a robust Hardware component we want to obviously build ecosystem of the models the model so the the heart of this we have meta as a member and I think L is one of the leading models we want more we want the best open source models we want to make sure they are the best and we will Advocate you know strongly for for open source AI before governments and you know oecd and whoever and we have resources and among among the alliance members we have people who who know how to do this properly so uh with that basically that's all I have to say the there is a timeline uh right we have been formed right now all this has been stood up and I just wanted to Dean to mention to say a few words on technical side so you can find out more information at the alliance that AI uh there six areas I'll just mention quickly if you're interested in AI safety if you're interested in open models and how to what's the next generation for building models what should we do Beyond rag you know what rag is obviously a great solution but it's also a hack if you think about it because we're working around the limits of models what should be the successor to rag um obviously there's you spreading the love educating people about AI uh that the hardware portability is a really interesting one Carl and I have been talking about this a lot Carl's going to lead that effort I talk to a few people here that are building AI accelerators so we want to make sure that we're not just locked into one heartb birth provider but anyway any of those interest you please uh reach out to us or talk to us tonight we'd love to tell you about the alliance and how we can benefit in Show