TDM S3 Ep5 - How Slack is Changing DevRel with AI - Kurtis Kemple (Slack & Salesforce AI)

regular fix of the best guests from the world of data and analytics this is the data mix with Brian buen and George [Music] Beer big George hello big George how you doing I'm good thank you how are you today pretty well pretty well like I feel like technically in control today which is like sometimes a difference for us a difference of opinion but like camera's situation feeling good the lighting is working to my favor so I'm I'm having a good start mate how about you good um I'm kind of running aroundall over the place we had some pretty bad storms uh earlier this late last week um and there's still a bit of clear up going on around that I'm at my brother's house at the moment we had some a power cut at the end of last week which we're kind of we've resolved um but I'm still sort of messing around here so not quite feeling grounded not as grounded as you anyway well yeah I I hear you on the AO in front right if if the data fix was a weather event then I'm sure that's whatwe'd be talking about but it's it's not so yeah we're we're in recovery mode on on the weather side of things but we're getting there we're getting there so um absolute pleasure to have you here as always and uh again no secrets here with our with our guest who we will have on very shortly but to the order of the day first of all right let's get our our bits and Bobs out the way right so like we we're really lucky to have you as an audience guys we we are not ashamed to say that we we we love the fact that youyou you download the pods you you join us on YouTube um we're working really hard on the YouTube channel just now to sort of make that more accessible for you um in terms of you know pulling all that together so please if you do have a second please just join us and and sub us on on YouTube Spotify or apple we really do appreciate it and it helps you keep up to date with all the latest stuff that's going on George any comments I think the um actions are speaking for themselves we've had our um subscriptions uh are going fantasticallyever since you started this Brian we're uh we're I think we're way over six 00 in fact pushing 700 now which is fantastic yeah I think I think we just landed another like 60 or 70 Subs in the last few days as well so I think we're actually even higher than that now which is totally awesome let's keep it going folks we appreciate your support thank you very much it means we can get better and better guests not that that but it's is incredibly important so thank you very much for everyone as wellso um so thank you very much um so into our regular segment of the data fix so George how are you feeling about this shall I line you up and get you in there yes please because we all right let's do it hope one today all right okay let's get you in so here we go so today's episode of the data data fix is brought to you by your friendly jingle in the background so George what have we got going on in the world of tech today I have a sneaking suspicion I might know what this is but I'm going to let youtalk I think you do it's the one that's been setting the stock markets um uh a light at the moment um I think um uh I heard somebody call it um America's Sputnik moment when the Americans thought that the Russians had won the Space Race um it is of course deep seek uh so this um is a new AI model um by a Chinese AI startup that was only uh founded two years ago um they have uh released a model an AI model um that is um reportedly outperforming openis um 01 in all sorts of maths and reasoning um criteria um it's justit's blown onto the onto the scene um and really taken the World by surprise um I think we've seen um stock prices tech stocks tumble um I have have some Nvidia shares um I was watching them crashing yesterday I would say crashing but they dropped 12 and a half% um I'm still up slightly um but uh it it really is sending the the shock waves through the the stock market and I think it's raising questions I think that's the key thing it's rais raising questions around why does it take at least 100 million uhfor open AI anthropic Etc uh to train a model when deep seek reportedly trained their model for under six million so 6% or less um of what these other um big AI companies are and this this is trained with data so apparently supposedly allegedly up to like um q1 2024 right so it's not right up to date but it's getting there so yeah I've got a I haven't tested it myself a couple of my friends have tested it and they're saying damn this this is real this is good so um it's it's one to watch um andit'll be really interesting to see how the market goes I did have a a a conversation with some of my buddies last night about what what their feelings are it could be good for the industry if you think um that this could mean that you can have uh um uh your own personal AI running on an Nvidia Jetson for example um or on a a modern iPhone um in your pocket this is really where what this technology is saying so maybe it'll make it um far more ubiquitous everybody will have one um and uh in their pocket they'll be able to keeptheir data to themselves there won't be this latency of sending information over the net so who knows but it's very it's it's kind of fascinating right because it puts Sam Alman and Co in quite a situation doesn't it after they've not only a couple of months ago they launched their um their $200 a month sub or whatever it was for the very latest models of chat GPT and now out of nowhere the little upstart has has turned up and all of a sudden everyone can have an llm in their pocket so it's it's quite a it's quite amarket Shaker at this point in time specifically with the political elements of it as well we've already seen a little bit of the the blowback with that um anyone who's turned on a TV over the last few days we'll have seen that so it's interesting times it really is and this announcement of 500 billion for Stargate which is America's um sort of belts and Roads program to build AI data centers across the country now there's going to be a question mark do you need 500 billion to do this well exactly and I think you canargue whether um whether this was politically motivated in terms of the time of its release depending on everything that's happening with current events in the world but we're not a politics show and we wouldn't surise on that but does seem quite um quite a coincidence shall we say that it's all happening at this particular time so but I think if to turn a positive spin on it I think um and you know we'll chat to Our Guest very very soon but um you know he has big views on on democratizationof AI for for everyone that that can possibly get it so um this this could be a bit of a game Cher I don't want to sound like a chat GPT description right by using words like game but I think honestly it is a bit of a watermark and and time will tell how how quickly this pushes forward but from what we've learned over about AI over the last 18 to 24 months it won't take long for this for this to accelerate so all right fabulous anything else on your plate mate or um well there's all sorts of things we could talk about butI think just about everything in Tech nowadays has got a political angle uh to um probably let's just um underscore it we've talked about deep seek we mentioned Stargate um I think AI is still definitely number one up there um in the uh in the technology news everything seems to have a an AI angle at the moment you just can't get away from it well well that's it and um again we are relatively it's important to say we're still relatively in the dark we're still really at press release stage wedon't really know what this thing is where it's come from how it Works how it's going to impact um everyday life so it'll be fascinating to see that and I can already see some comments in the chat so we might just have a quick quick chat about that um but thank you very much mate that was a that was a storming data fix this week it's on everyone's minds and um it needed to be talked about so thank you very much for that all right let's let's pop out of this segment then and get back into theshow for real thank you very much mate that was awesome so the data fix is here every two weeks keep you up toate with tech in the current day lovely lovely so that was if that was like the entree then I guess it's time to move on to the main course right do you want to talk a little bit about our guest today mate before we bring them on yes I've met uh Curtis um on a call we had was it before Christmas but honestly the uh the energy uh Curtis's story um where he's got to in his career all inspiring umall uh riveting actually I think we are in for um a fantastic segment here and I can't wait to get them on well let's do that right now let's get let's get Kurt on hey Kurt can you hear us okay yeah I can hear you all uh am I coming through okay oh yeah you're all good perfect superb to see you mate thank you so much for coming on and spending time with us today uh yeah absolute pleasure to be here uh happy to talk about data anytime big fan awesome well well look I think before we get into introductionsand this thing I think it's fair that we we call out some of these comments in the chat and give you a chance to see see what you think about these recent developments as well Kurt but um welcome anhel thank you for joining in are we are we positive that this thing does not run an oxident now I don't know if that's a terminology thing for me I'm not sure what idental Hardware is but I don't know is the answer I don't know if either of you guys have a viewpoint on that but I'm assuming we're talkingabout the Chinese AI here but I guess not so we'll um we'll we'll swiftly move on anel if you want to ask another question please do and our good friend Luke who's a previous friend of the show um has been on for an episode before legally acquired data in line with global data protection legislation oh my goodness yes this is a massive can of worms that we cannot possibly cover today but look seen the rumors back discuss in more detail yeah I mean there's there's a lot of gray areas here cut right and you'vehad quite experience of AI and that like new tools just coming out of the woodwork and becoming globally dominant is not a good sign for gdpr or any other data regulation that exists out there right yeah I mean we have to imagine that any generalized model is you know um pulling lots of information from the internet they do what they can you know to control where they're scraping from but you know um a lot of times you know like you said it's a very big gray area who knows where they're getting thesedata sets it's also uh a lot of times companies will use thirdparty data sets instead of doing a lot of the bulk work on their own and so you know you have to like kind of follow the trail there so definitely gets into a gray area I'd be very curious to see if they open source like the um data that has been trained on uh or at least give some sort of indication of where they're pulling it from I think you know that's really for me like the next step in like the true open sourness of uh AI uh we see thiswith hugging face right like you can upload your data sets there and so it's pretty transparent like what you've trained your models on I would love to see more transparency around that for sure I think that would help move um uh like AI trust a very long way especially for end users trust is a big word there right and we'll I'm sure we'll get into that as as we talk more but um you know again I I just wanted to to say welcome again Kurt and obviously you know um it was it was awesome to meet you out in in SanJose um it feels like a long time ago but it really wasn't that long ago it just feels that way so yeah Kurt and I were lucky enough to meet out at Tech field day and it was our first exp both of us our first experience of that how did you find that whole thing mate it's a bit of a whirlwind for me so I'd be interested to hear what you thought about it yeah I mean I it is definitely a bit of a a whirlwind but I really enjoy that structure um you know to me it feels much like uh so I work in developer relations I would put this AKinto like an accelerator program right it's like you're closing the feedback loop between uh companies and people who can provide really useful insight into like you know what they're doing uh so I think you know when you look at that and then it's it's both ways one it helps the companies understand maybe where they can make adjustments or following the right track right like green l oh we're doing the right thing here uh and on the flip side it gives experts an opportunity to expand the breadth of uhyou know information that they have available to them to help make even more and more informed uh decisions context is everything so that's something I live within developer relations uh and I think anytime you can shorten a feedback loop you know everyone's going to benefit but yeah overall very great experience cool well I'll I'll just chip in with one more and then I'll let I'll let George get Unleashed on the other side there um so why don't you just tell us a little bit about um you know whoyou are what you're and what you're doing as well because obviously we met um in Zan Jose but um tell us a bit about yourself C to start with yeah so hey everyone Curtis kle uh currently senior director of developer relations at Salesforce but I focus on slack as a platform so I've uh been there for a couple years now now uh I was first focused on growth and engagement for platform and then uh took over deel pretty recently about six seven months ago so it's been a very uh fun and interesting challenge I've enjoyed itquite a lot uh before that I was doing consulting and developer relations helping uh startup scale ups and Enterprise businesses understand devil what they need from it you know uh given their different characteristics and goals uh and just like sometimes how to like build those teams and help them start on those Journeys uh and so uh prior to that i' you know I've been in engineering uh software development for over a decade uh that culminated uh I was Tech lead uh of the UI team for Major League Soccer soanything fanf facing we like rebuilt the mobile app uh Greenfield uh uh migrated to graphql for a lot of their apis um rebuilt all the like the sets processing Pipeline and stuff it was real fun I also built their Fire TV app uh their uh Facebook um chat box my first intro to AI like eight years ago uh back before uh you know we really hit the mark on natural language processing so I was really trying to do a lot there to improve the experience for our chat bot so it really cool uh but yeah that's kind of you know I've always enjoyedhelping folks be successful and that kind of led me into Education and Training which is you know what led me to developer relations uh and then I think what you know really grounds this in this show in particular is I take a very data informed approach especially when you're working in an Enterprise company we cannot um undervalue the importance of having quantitive data you need that qualitative as well but you really need to understand the quantitative data that is available to you uh and what people care about inyour business so that you can show impact and continue to uh grow and succeed in developer relations yeah fantastic so um are you are you remote based or are you actually in the in the Salesforce office uh I am remote based so I'm on the East Coast excuse me roughly the Virginia Beach area uh but I'm in San Francisco all the time so yeah okay awesome um the I mean we certainly the industry that I'm in um we hear about Salesforce all the time it's a huge player in Enterprise software um I don't want to labor the AI Point toomuch but um from an Enterprise point of view um it seems that um Salesforce is um right there at the Cutting Edge of how Enterprises are using um uh Ai and I think um uh you've got quite a a strong program um and road map um of AI and some like quite ambitious targets as well do you want to talk about them for a second I would love to yeah absolutely yeah uh you know I think Salesforce with agent force is like you said extremely well positioned in the Enterprise space I mean Salesforce is such an an expansive platform right like you canintegrate anything into it and because you can integrate anything into it that means you can bring it into agent force uh on top of that there's been a lot of work put in uh within the atlas reasoning engine and trust and privacy layer right like to make sure that you know all of your data is protected in and out you can even connect like thirdparty llms and still like the data is protected on the way in and out like you know what's going out exactly and what's in um and so it's just a reallynice setup but long story short you can bring all of that anything all of that into these actions for your agents and so like you can you know um I I just think like it's fundamentally about to uh uh change uh we know that like smaller more specialized uh models uh with guard rails and and better instruction perform better it's why we have like uh Cloud projects and your own custom gpts that you can upload your rag data to and Define how you want them to behave right like so we know that that is that uh and so I just think that ayou know sales force and agent force are so well positioned and then when you put slack as that conversational interface in front of these agents they can definitely be working externally with customers as well but you know when you have these internal facing agents Gap filling expertise picking up all of that you know low-end Toil and automation stuff off of your team's plate to free you up to focus on the the work that matters right like the stuff that drives impact you know and we're just at the beginning of it I'm already buildingagents um that are doing pretty incredible things and you know I just think about how we're in the infancy stages of this the MVP like in a few years Atlas reasoning engine will be like unreal it's already impressive some of the things that it does um and so yeah long story short I do think we are well positioned uh I do think agent force will provide a lot of value uh to a lot of businesses and it's already providing value to me and my team like we're building agents so I'm I listened a couple of weeks ago I listened to apodcast um where Mark benof was being um was being interviewed so it's an hour long u podcast he was talking about the AI strategy in Salesforce s sorry to labor Salesforce I know your focus is slack oh no this is yeah it's all I it's all trust me one company yeah awesome but but what came across really was um Mark's um sort of vision and enthusiasm so I was kind of keen to know how much is um do do you kind of get that energy that that Benny off energy down through the rest of the business um into thestuff that you're doing on a daily basis that's a great question yeah that's a that yeah it is absolutely a great question I'll tell you this like uh agent force uh fundamental to Salesforce that energy is literally in every team like everybody uh is understanding you know how we can uh contribute to the successive agent force it's a very exciting time and unprecedented um uh product you know it feels like uh you know when mobile first came out or the expansion to cloud like all these things like fundamentallychanged the the shape of of the company product offerings and so you know I just think that you know um agent force is well it's a it's a force and yeah that energy is carried all throughout everything uh we all see a lot of potential with it um you know and on the slack side it's like an excellent um integration between those two platforms that just unlocks like a lot and so we've been working at finding those real sweet spots of keeping slack slack and tying into Salesforce real well and agent force unlocks a lot it really doesit's really nice and and car I know that um you know we've already used words like Ai and agentic and everything that we possibly could have used we have already right but I I know from I know from you from speaking to you personally how much this like means to you and I don't just mean from a technological standpoint right and I want to just shift a little gears a little bit into how Ai and data and all of those things that have been important to you how much you feel should be important to other people aswell and not only important but accessible right because oh yeah this is a bit we we spent a lot of time talking about this and um I I was pretty amazed to hear your thoughts on it so why don't you tell us a little bit about where you're coming from yeah yeah so much like what I just described with this world that we have in Salesforce where you'll be able to you know uh build these uh agents right that can handle anything from you from teaching you code instruction to taking action uh on on different apps now that's very powerfulright but that's like trapped in the ENT Enterprise right but like you know as we look at things like deep seek and your ability to have this on your phone right like wouldn't it be really nice if the end user like the just the everyday person could also have access to just like you know an ecosystem of models or agents that can help them do anything from figure out how to start a restaurant because they're an amazing chef but they don't have a business background or maybe they're a civilengineer and they want to uh you know take into account uh climate effects on on what materials they're planning to use and so they could you know rely on these these models and uh agents to help them navigate and figure out these again Gap filling expertise or removing toil from people so that they can focus their energy on times and time on things that matter so I just I think there's a ton of opportunity but especially in the education space um you know with my personal background so I'm formerlyincarcerated right I was incarcerated in uh 2004 roughly uh released in 2010 and that's actually where I got my introduction to programming now I didn't really have access to anything I had like a book we had a computer with no internet and I had a teacher who had like a rough understanding of HTML CSS and JavaScript luckily like it just really fueled a fire in me and I I was persistent and I stuck with it and I continued to educate myself and upon release I continued to educate myself and follow that path that path couldhave been greatly shortened for me and I could have learned a lot if they had given me a tablet with no internet but did have an agent on it agentic that could review my code for me and offer suggestions right and like show me diffs and where I M made my mistake and like teach me further and or you know now everyone in the the room's questions can be answered at the same time to free up uh the teacher but hey we can keep human in the loop if we're unsure about the results here from what we're gettingfrom this agent and I still have a teacher there right like something to reach out to so I just think is like as these models get smaller and as these reasoning engines get better and we start making them more specialized right and and uh uh powered by contextual data you know rag like all these things you know um they matter right so they matter U because they all can lead to the democratization of like information and AI to to everyone uh you know and you could look at uh humanitarian use cases that this could benefit you know asidefrom educational use cases that this could benefit and so you know as someone who like you know would benefited from discovering front-end development which only requires a computer to have access to which is still a lot for some people I I do understand that as well but you know it's a lower barrier to entry than a lot of other for of like software development or other fields and so you know I think if we could continue to reduce that barrier to entry or provide more and more resources uh to people whouh might lack access to them you'd be amazed at like what that might do um uh globally like you know the impact that that could have and so all that to say is I think as models get smaller and more specialized and and as we push this scale uh more towards userfriendly and like hugging face is a great example again because they've pushed the scale to developers it's an amazing DX now to work with models train them right uh uh uh deploy them share them it's going to push further to the end user and Soeventually we'll see something like a Marketplace or marketplaces where people are deploying models and agents and there's like some sort of like Revenue sharing system with the people who are putting them up there and it gives access to you through your phone or computer or whatever access to all of all of these things and I think like I love the cost um aspect of of deep seek because like yeah you know traditionally you know you're trapped in the Hundred Million Dollar Club right like if you don't have aund million you can't trainlike a a really significant model and so you know to see the price drop to six billion right like if you know once that's like actually um you know consistent right like that will continue to happen right like we'll we'll have more and more Innovations and how we train how we build models and you know the price will continue to drop and so I think all things lead to democratization but like you know I think we could get there quicker if more people were thinking about it you know uh and so from my perspective it's it's just youknow I want to see AI democratized because I understand the intrinsic like value and and the true transform like I don't want to sound like AI but the transformative nature that like this Ai and smaller and and and better Tech actually unlocks for people you know like AI is a very big umbrella and a lot of it you know probably not as you useful honestly but like there's a lot of parts of AI that are gonna change things pretty fundamentally um and you know I think it's just again everything's on a scaleright nothing is is binary and so I'm very interested to see how it shakes out but I just hope it gets here faster than you know than later sooner than later yeah look and I'm smiling in the most warm-hearted way that that I can CT because I know like how much of a personal difference that's made to you and how much of a personal mission this is for you as well given your background that I think look I think it's amazing that you're you're so open to just talking about it and just putting itright out there there's no there's no awkwardness there and I think I think that's incredible but I could focus on so many of the things that you said but I'm going to just take one because I was going to actually say it as well was the the barrier for entry right I think that is the key part of all of this and you can say that the like Chinese AI might help accelerate this in a massive way when people realize that cost at scale is not as much of a factor um but ultimately we need to always look atways that we can lower the barrier for entry so that you know AB people can use like a 40 pound phone to get the things that they need rather than having to have a high-end Apple 16 iPhone or whatever it is plus a a data subscription plus a chat GPT subscription or whatever your your L&M provider is that's a lot of cost for people that just can't necessarily afford that so when we talk about democratization that's highly tied in with affordability and economics as well and that's why I think these recentannouncements are very very exciting and as George mentioned in the data fix as well and I'll just finish with this one um in terms of like Stargate and the value that's been por portioned to that project how does all of that now get because it must be re-evaluated now that given these Revelations you can't just come out and say 500 billion was it George yeah allocated over four years that that you're playing in a different ballpark now so back to your back to your sports background as well there Curso there's so many things that we could just talk on and on about but I think it really is like a super pivotal time just given all of the ways that these things these inflections are hitting each other from different angles so yeah it's a lot and one final thought on this is just I think it'll be a while even as the technology catches up like to enable this um you know uh even at all uh there's still we're in the again the MVP or infancy stages of user experience for AI so far because we've cracked naturallanguage processing we've said everything is a conversation right and so so um you know I think we'll start to see a fundamental shift there because uh right now I still feel that um you have to have a certain level of like understanding and expertise beyond what you would expect from an everyday user to really truly maximize your use of AI like if you look at a chat GPT or or anthropic right like these models you know uh like clog right Sonet and all them it doesn't matter the they all uh you know it's law of averages right soif I just hit it with a question I'm going to get like a pretty average response like unless I start to learn how to fine-tune how I interact with this model I.