Sidecar Sync

ASAE Annual Conference, Figure’s Humanoid Robot, & Segment Anything with Meta Ai’s SAM2 | 42

• Amith Nagarajan and Mallory Mejias • Episode 42

Share episode

Send us a text

In this episode of Sidecar Sync, Amith and Mallory gear up for the ASAE Annual Conference with a detailed discussion on how to navigate the event effectively. They delve into the world of AI vendors, exploring how to distinguish valuable tools from the hype, and discuss the latest advancements in humanoid robots and visual segmentation models. Whether you're attending ASAE Annual or evaluating AI tools from afar, this episode offers valuable insights and practical tips to stay ahead in the association world.

🔗 ASAE 2024 Annual Meeting & Exposition:
https://annual.asaecenter.org/expo.cfm

🛠 AI Tools and Resources Mentioned in This Episode:

Perplexity ➡ https://perplexity.ai
Figure 2 by Figure AI Inc. ➡ https://figure.ai
📽 WATCH:  https://youtu.be/42Uemd92b3E?si=c1MLSTzsdLPyfoXe
SAM2 by Meta AI ➡ https://ai.meta.com/blog/segment-anything-2/

Chapters:

00:00 - Introduction
04:48 - ASAE Annual Experience and Tips
09:46 - Evaluating AI Vendors at Conferences
21:28 - Importance of AI Safety & Privacy
28:12 - Humanoid Robots: Figure 2
38:18 - Potential Applications of Humanoid Robots
43:44 - Visual Segmentation with SAM2
51:40 - Further Applications for Associations
55:41 - Upcoming Events and Wrap-Up

🚀 Follow Sidecar on LinkedIn
https://linkedin.com/sidecar-global

👍 Please Like & Subscribe!
https://twitter.com/sidecarglobal
https://www.youtube.com/@SidecarSync
https://sidecarglobal.com

More about Your Hosts:

Amith Nagarajan is the Chairman of Blue Cypress 🔗 https://BlueCypress.io, a family of purpose-driven companies and proud practitioners of Conscious Capitalism. The Blue Cypress companies focus on helping associations, non-profits, and other purpose-driven organizations achieve long-term success. Amith is also an active early-stage investor in B2B SaaS companies. He’s had the good fortune of nearly three decades of success as an entrepreneur and enjoys helping others in their journey.

📣 Follow Amith on LinkedIn:
https://linkedin.com/amithnagarajan

Mallory Mejias is the Manager at Sidecar, and she's passionate about creating opportunities for association professionals to learn, grow, and better serve their members using artificial intelligence. She enjoys blending creativity and innovation to produce fresh, meaningful content for the association space.

📣 Follow Mallory on Linkedin:
https://linkedin.com/mallorymejias

Speaker 1:

I think understanding this stuff is so critical. The closer you are to where this technology will displace people Again. I think there will be that, and there will also be places where there's so much demand for additional service that I think people will actually do better. Welcome to Sidecar Sync, your weekly dose of innovation. If you're looking for the latest news, insights and developments in the association world, especially those driven by artificial intelligence, you're in the right place. We cut through the noise to bring you the most relevant updates, with a keen focus on how AI and other emerging technologies are shaping the future. No fluff, just facts and informed discussions.

Speaker 1:

I'm Amit Nagarajan, chairman of Blue Cypress, and I'm your host. Greetings everyone and welcome back to another episode of the Sidecar Sync. We are doing this episode today, a few days ahead of the ASAE Annual Conference, which is always a really fun time in the world of associations. Can't wait to get out there. We've got a lot of interesting topics, as usual at the intersection of all things association and AI, so we're going to get into that in a minute. My name is Amit Nagarajan.

Speaker 2:

And my name is Mallory Mejias.

Speaker 1:

And we are your hosts on this journey at the Sidecar Sync. Before we get into our really exciting topics for today, let's take a moment to hear a quick word from our sponsor.

Speaker 2:

Today's sponsor is Sidecar's AI Learning Hub. The Learning Hub is your go-to place to sharpen your AI skills, ensuring you're keeping up with the latest in the AI space. With the AI Learning Hub, you'll get access to a library of lessons designed to the unique challenges and opportunities within associations. Lessons designed to the unique challenges and opportunities within associations, weekly live office hours with AI experts and a community of fellow AI enthusiasts who are just as excited about learning AI as you are. Are you ready to future-proof your career? You can purchase 12 month access to the AI Learning Hub for $399. For more information, go to sidecarGlobalcom. Slash hub. Amit, you're back in New Orleans. I think I already know how you feel about that, but why don't you share with our listeners?

Speaker 1:

How do you feel? Well, it is hot, it's very hot.

Speaker 2:

And we just made Amit turn off his air conditioning so we could get the sound just right for this episode.

Speaker 1:

So he's sitting in a room with no AC in New Orleans. Well, it is probably. It's rising quickly, but I think it's still in the low 70s, so we'll see. If, by the end of this episode, I'm sweating profusely, then you'll know why.

Speaker 2:

Exactly.

Speaker 1:

So but yeah, New Orleans is a wonderful city in so many ways. Our weather in the summertime is not something I'm a big fan of. I prefer to be in California or Utah or Canada, or basically anywhere except for New Orleans, around this time of the year. But my kids are about to go back to school, so here I am.

Speaker 2:

Absolutely, and you mentioned at the top of this episode, asae Annual is this weekend in Cleveland and you and I will both be going along with a few others from the Blue Cypress family of companies.

Speaker 1:

How many ASAE Annuals have you been to? Amit? That's a great question. I really have no idea. I mean I don't go every year. I try to get there every year because there's just so many people there I want to reconnect with, but I'd say probably every other year, on average over 25-ish years, so a lot, and actually I probably used to go every year pretty consistently, so maybe 20. I don't know. It's been quite a few.

Speaker 1:

They're great events, highly recommended. It's a great place to meet a ton of people in the association community. I think it's roughly 50-50 in terms of association staff and vendors who are trying to sell stuff to associations, which of course, we would fall into that category. Sometimes people look at that and they go. It's kind of not the greatest ratio because there's so many people selling stuff and I think of course there's there's. That's a perspective I think has merit, but I also think that, you know, the vendor community has a lot of interesting perspectives to share as well. So I think that there's a good mix and I like connecting with people from both sides of the aisles in terms of people that are coming up with creative, innovative solutions that associations can benefit from and, of course, the association staff themselves.

Speaker 2:

For sure this will be my third ASA annual yeah, I guess my third. That's kind of crazy. That was one of the events I went to right when I started working at Blue Cypress and I was so impressed. That was actually the first industry conference I'd ever been to, so it was a lot of firsts for me, but it was about 5,000 people and then, as you said, kind of a good mix of vendors and association leaders and I feel like, even just going into this third year, I've been able to see kind of the evolution of more and more AI sessions, which has been really interesting to watch.

Speaker 1:

Was the first one. You went to the one in Nashville.

Speaker 2:

It was. It was in Nashville, Atlanta was last year, Cleveland this year and then Los Angeles next year. I don't know if you knew that, but LA, that seems like a crazy place for ASA annual.

Speaker 1:

Well, you know, I think going west is always a challenge for this conference because so many association folks are Central Time Zone or Eastern Time Zone. But you know California is a massive market. There's thousands of associations there. You know California, I think, is the sixth largest economy in the world, if it was its own country, so it's a big deal. So I think bringing it west from time to time is a good move for ASAE.

Speaker 1:

I think attendance probably won't be as strong in terms of folks coming from the East Coast, but then again it's an attractive destination, particularly in the summer. I mean, you know California weather in August is certainly more desirable than pretty much anywhere else. So hopefully it'll do well. But yeah, I think it's a great touch point for the sector at large to come together at that scale every year and, just you know, kind of get a feel of what people are thinking about. I think the content can be quite good and really the networking is what I go for. I just like to reconnect with so many people that I don't see throughout the year. But for that event. So that's an exciting thing for sure. I think in Nashville your very first time at ASAE you guys had a little bit of a snafu with the booth right, if I remember correctly.

Speaker 2:

Oh, a meath. It was my worst nightmare. I'm glad we can all laugh about it now. But again, my first industry conference and I was responsible for ordering all the booth materials and I had no, I did. I thought I did such a good job, worked with our graphic designer, was so ready, had everything in order, get to the expo hall, realize there's a height requirement on booths, and so ours was nice, 10 feet tall In fact. I remember thinking, wow, ours is taller than everyone's, this looks great.

Speaker 2:

Until someone rode by on a what must've been a scooter it's just seared into my memory and this woman rode by and said that's too tall, have to take it down. And I remember begging this woman please, please. I just started working here. This is, please, don't do this. And we worked with her and talked to her and begged her so much that they put us in the corner. So they didn't put us with everyone else in the main expo hall, but up against the wall so that we wouldn't detract from everyone else's booth, which was totally fair. And it all worked out in the end of me, but that was, um, it wasn't my favorite day, for sure.

Speaker 1:

Yeah, I remember. Uh, actually, I thought it worked out pretty well because the corner actually was, I think, pretty close to the food, if I remember correctly, or maybe it was close to the bathrooms or something, but it was close to something people needed access to regularly.

Speaker 1:

So it's kind of like the real estate play. You know primary place is go where people eat and secondary is go where they need to go to the bathroom. You'll find traffic in either place. So I'm glad it worked out. But it's one of those things that does sear itself into your memory and you know, sometimes breaking the rules does work out. I guess this was one of those times.

Speaker 2:

A hundred percent.

Speaker 1:

You did it on purpose, but that's super fun.

Speaker 2:

It's my first tidbit of advice I share with new hires. Pretty much is like oh, by the way, there are height requirements in expo halls, Just like. Don't make that mistake and nobody else has made it thus far, so at least that's a positive to come out. But everyone, if you will be attending ASA Annual Sidecar, will be in spot 939. I will be at that booth. I'm sure Amit will be stopping by at times and we will have hard copy versions of Ascend 2nd Edition. So please stop by and visit us 939 and get a copy Booth 939.

Speaker 1:

I'm going to have to remember that.

Speaker 2:

Yeah, you didn't't even know me. Yeah, you might be there.

Speaker 1:

I'll be there too and I'm looking forward to uh seeing folks and, yeah, the copies of ascend second edition, I think, will be literally flying off the table. A lot of people have expressed excitement about getting their hands on the uh fresh off the press new edition that has over a hundred pages, I think, of new content and pretty much every single existing chapter in that book has been reworked and updated and retools and, in fact, changed the entire order, like we talked about previously on the pod, to really reflect a better flow, I think, for our readers. So super excited to start getting feedback on the new edition of Ascend and if you are not familiar with Ascend, it is our AI book for associations. It's all about unlocking the full potential and full power of AI for the association sector specifically.

Speaker 2:

Today we've got an exciting episode lined up. Our first topic is evaluating the sea of AI vendors out there, and that was actually inspired by this very event, asa Annual. We will be talking about Figure 2, the latest humanoid robot, which Amit and I both agree is pretty creepy, and then we'll be talking about SAM 2, the latest humanoid robot, which Amit and I both agree is pretty creepy, and then we'll be talking about SAM2, which is the Segment Anything Model 2 by Meta AI. So, first and foremost, amit and I have been talking about ASAE Annual, but we might have some listeners that are not at all familiar with what that acronym means. So it's the American Society of Association Executives, and this is their annual event that they host.

Speaker 2:

Every year, as I mentioned, it brings about 5,000 people together, a mix of vendors and association leaders and I'm going to set the stage for you all if you haven't been but it's typically a mix of keynote sessions, breakout sessions, networking events, happy hours, lots of socializing, and then you've got this big, giant expo hall with technology and business services on one side and hospitality services on the other. The expo hall is really fun, especially the hospitality side. A lot of these cities and chambers of commerce, create booths and give out ice cream and coffee and even raw oysters. Last year that one was a little odd in Atlanta but on the business and technology side, it's mostly vendors offering their products and services to associations. So this year I'm sure many vendors will be promoting AI offerings, and it got us thinking about the importance of evaluating companies with AI products and services.

Speaker 2:

And I want to zoom out a little bit, because even if you're listening to this podcast and you're not attending ASA annual or you don't typically get your solutions from in-person events like this, I bet you've probably still felt lost, at least at your desk, in terms of sorting through all these AI vendors. So we want to explore where do you start, how do you decide if you really need a tool or a product or if it's just something that's fun to play with, and how can you ensure that you're buying from companies that are legit? So, amit, my first question for you and I really don't know how you're going to answer this one, so I'm curious but if you walked into an in-person expo hall or even a theoretical, a digital expo hall with full of AI offerings, would you walk through totally open-minded, just willing to talk with any company that caught your eye? Or would you kind of explore the vendors beforehand and go in with a game plan?

Speaker 1:

I think both approaches are great and if you have a specific set of needs in mind and you're looking at hey, who offers an AI solution for accounting or something like that, you have a very specific idea of what you're interested in. By all means, doing a little bit of pre-research is quite helpful because it can narrow down where you spend your time in the expo hall, so that can make a lot of sense. If you're more where I think a lot of people are right now where they're still curious, they're thinking through like, what are the use cases for AI in the association market? How can I benefit from this?

Speaker 1:

Wandering is sometimes super valuable, I think, even if you have, you know, a set of ideas of who you want to go visit. Doing a little bit of wandering, especially like off the beaten path a little bit, some of the side aisles might be interesting, where you know you don't necessarily have the biggest vendor with the biggest booths, but you might find some really innovative smaller startups that have some interesting things to offer. So I do that every year at this event and all the other trade shows that I go to, I look up, I just walk up and down the aisles and I kind of like, you know, just look at what they're doing and see if I see something that's interesting to me and then I'll just have brief conversations with people. And you know, I don't invest a ton of time at each booth but I'll listen to their pitch for 30 seconds and choose to move on or stay put if something compelling comes up.

Speaker 2:

Sounds very intimidating to me, thinking of giving a myth a 30-second pitch.

Speaker 1:

Well, you know, I think the way I look at it is there's this opportunity in a trade show to very quickly convey to someone something that's different than what they've heard. You know, if they've heard the same thing over and over, it's like hey, you know, come here because you want the free t-shirt or whatever other booth giveaway. That's really uninteresting, right? I don't want another bag full of plastic crap that I'm going to have to throw away or recycle. It's not fun. I'm more interested in, like, who's doing something that's different, and it's hard to find that because, you know, of course, in any it's a standard deviation thing. Like most people are going to be pretty close to the center and have the stuff that you've already heard of.

Speaker 1:

I do think this year will be very interesting because I'm anticipating a number of vendors we haven't heard of. I'm anticipating software vendors and other other folks with updates to their offerings that are AI powered. There really should be a lot. I'm hoping to see, you know, a very large array of AI and AI powered offerings, both services and software products. So I'm pumped about it. I'm really can't wait to explore what's out there. So I think there's a lot.

Speaker 1:

I've got a lot of ideas on how to go about thinking through like what's a legit use case and what's not, and we can get into that. But I think being open-minded is super important at the early stages of any new general purpose technology. It's kind of like back in the early days of the internet. You know what were the killer apps going to be. They're incredibly obvious now, but that's in the rear view mirror. So I think having a bit of an open-mindedness to it is always helpful. So I think having a bit of an open mindedness to it is always helpful, but particularly in an era of a new technology like this.

Speaker 2:

Yep, I definitely want to dive into the use case piece. I would say that a lot of the AI tools we talk about on this podcast are incredibly useful but also fun to play with, and one I'm thinking of is the text to music model Suno, which we covered in a previous episode, which does have business use cases, but again, it's just kind of fun. But I'm wondering how do you think people can ensure that they are investing in products and services that solve a current problem? And then I think what's even more difficult is how do you ensure that an AI product or service is going to help you innovate or create something new when it's never been tested? So kind of like solving a problem you have but also being open to new possibilities.

Speaker 1:

Sure, well, I think there's what you cannot do currently, but you might be able to do with AI. That's like a category of like really interesting innovation and there's products that fit into that category, like AI powered chat that has knowledge assistant or agenda capabilities We've talked about stuff like that a bunch on this pod and those are capabilities you don't have currently. Right To have infinite scale and infinite knowledge available instantly to your members through voice, through text Pretty cool concept. How that applies to your particular organization, does it apply or not? What's the best use case those are all things to explore. But thinking about that, it's essentially new ground. It's basically Greenfield. You know, nothing has happened yet in that space. Or some people would refer to it as Blue Ocean. Right, it's basically uncharted territory. There's not a lot going on, and that's super exciting to an entrepreneur like me. I think that's fun. I think that's fun. I think that's something that could change the world. I think it could change the business model for the association and there's definitely room to. There's tons of opportunities like that.

Speaker 1:

The other side of the coin is to look for AI solutions and offerings that can dramatically improve the efficiency of existing processes. So if you say okay, you know, the earlier example category was that which you cannot do without AI, and the second side of it is well, that you already do the tasks you currently perform but are perhaps mind-numbing or prone to error or just take a large percentage of your time. So, for example, if you have an IT team of six people and half of those people spend all of their day, every day, writing reports with the database right, and all they're doing is writing reports and solving things like that, well, what if an AI existed that could solve that for you? That would be very interesting if an AI could automatically create reports against any database for any of your end users. Hint hint, there is such a product available now, called Skip, but that kind of thing is like a play on efficiency. It also opens up the door to doing a lot more of the same, because there's essentially an infinite demand for like reports on your database. If you think about what would users actually ask for if they could get instant responses or near instance responses, you'd have a backlog like even longer than your current backlog, and those of you in IT who've been in this position in the past, you probably already have a very long backlog months and months of reports that need to be created. So that's an example of taking a process that's inefficient and probably not anyone's favorite and helping automate a large portion of it.

Speaker 1:

Another example is the people who are responsible for tagging content. Where, oh, we want to take our website content or our journal content, we want to tag it based on author information, topical information, other specialized attributes, depending on the nature of the publications that we produce. That, too, is a very manual, painstaking process that AI can automate completely. So you're going to see solutions like this in various places, not just from our group of companies. Of course, we have bunches of cool things to show off, but the idea is, can you find a pain point in your organization where you can alleviate that pain point significantly?

Speaker 1:

Because if you can do that, now what we're talking about is a whole bunch of people that you've freed up and that might sound euphemistic in terms of, like people being let go, but most associations aren't looking to just thin out their ranks. They're looking to eliminate redundancies from an efficiency perspective in their process and then get those people working on things that really are higher value. So I think that's a really exciting thing. Of course, that comes to the different set of risks and concerns people have, because, of course, the people who currently do those things, the first thing they think about is do I still have a job?

Speaker 1:

But anyway, I digress somewhat, but the idea would be in my mind is to quickly bucket these solutions into the two categories either, like it's a brand new opportunity, or it's something that we can make more efficient, and then, from there, think about like, does that actually affect your organization, right? Like, is it a process like the content, auto-tagging, auto-tagging, for example. Well, if you don't have a lot of content which I haven't yet met an association that doesn't have a lot of content but if you didn't have a lot of content or tagging didn't matter, then that solution is probably not very interesting. Move on, go to the next booth, go to the next vendor, see what else is out there.

Speaker 2:

I like the idea of two buckets. So for listeners who are attending ASA Annual this weekend or potentially just evaluating AI vendors online, thinking about that which you cannot do, that you could do with AI, and then that what you're currently doing that could be better or more efficient with AI Is that right? Yep? Or more efficient with AI Is that right? Yep? Do you think it would be helpful, within those buckets, to kind of come up with three things in your current organization for each that you keep in mind as you walk through the expo hall?

Speaker 1:

Yeah, I think one of the things I'd ask myself going in is you know what? Are we always stuck on right when we talk about like, hey, we know we want to get a whole bunch of really cool things done, but we're always stuck with X. And a lot of people will have what is X? And for some people it's well. We spend three quarters of our year getting ready for our annual meeting and that's like everyone in the organization is focused on that. Well, maybe break that down and say, well, what is it that's taking so much time? How can we potentially improve our efficiency there?

Speaker 1:

Or the other side of it is what do people complain about your customers, your members, what are they complaining about? Oftentimes, they'll complain about it's hard to find stuff on your website. Your website's too slow, you don't have the content that I'm interested in, or the content you have that I'm interested in is slightly off because it's not helpful in my role or in my position. It's a little bit different than what I need. Are there opportunities to leverage AI to address some of these customer concerns and customer complaints?

Speaker 1:

And I think there's tons of opportunities around that. So I think doing a little bit of self-reflection before you walk in to kind of re-acclimate yourself, especially if you're a fairly high level exec and you're not super detailed in the operations, to kind of re-familiarize yourself with some of the day-to-day challenges your team is having, as well as the challenges that your members are having, could be instructive in terms of where you home in on what could be useful to your organization Because your own specific role you might find something that's helpful to you personally or you might not. But to think a little bit more broadly about it from those two lenses I think could be very useful.

Speaker 2:

I read an article recently that quoted the Securities and Exchange Commission Chair, Gary Gensler, who explained earlier this year that people have to be aware of AI washing, which he defined in his own realm as false claims to investors by those purporting to use these new technologies, aka AI. This is something listeners of the podcast should be aware of, not saying particularly at ASAE Annual, but just in general this idea of AI washing, slapping the word or the phrase AI onto any new offering.

Speaker 1:

Yeah, totally. I mean, this pattern plays out in every tech boom. So back in the 90s and early 00s, like any business you can imagine, they slapped on a dot com at the end of the name and, bam, they became an internet company. So you take, like you know, whatever. Just imagine any business where calling it a dot-com is the most ridiculous thing you can imagine. It was happening. People would say, hey, we're a whatever dot-com, and then they'd go raise a bunch of money and then they'd fairly quickly thereafter blow up. So there is a lot of that. That is something that obviously is misguided from the perspective of investors, also from the perspective of consumers.

Speaker 1:

I think in the world of AI it is also happening. It's been happening for years. I think you have to look at it and say, well, where is the AI relevant, why does it matter and how frequently is the AI used? So there's tons of business applications that are out there that said, oh yeah, we've got AI, we're an AI-powered solution. You go, okay, well, show me where the AI is. And then they say, well, here it is, and it's like some component of their application where you know, you'll see like a pop-up box in the screen where someone types in some text and it's a prompt and it's generating some piece of content. So for marketing tools it might be like oh, let's generate an email to my customer, and it's basically like a very thin wrapper around ChatGPT or Clod or one of those other tools, and while on the one hand it provides some degree of utility because you don't have to go to the external tool, you can use it within the said tool, like, say, in an AMS. You might have that for, like, creating a marketing campaign or creating a meeting description or things like that, where you're typing stuff in. I'm not saying there isn't value in that type of a feature. I would just argue that it's probably kind of an edge case. It's not like an everyday efficiency thing. If you can, you know if you are doing this constantly or if you have lots of people doing this constantly and there's an AI accelerator in there, that's awesome. But a lot of times the features you'll see will be in these kind of side cases as opposed to mainstream cases in the software. So I'd ask for, like give me an example of the AI you're talking about. And I'd ask the vendor like okay, well, give me an example of the AI you're talking about, and I'd ask the vendor like OK, well, give me some ideas of how this actually works, explain to me what's actually happening.

Speaker 1:

One of the most important things you have to think about, which we talk about here on the Sidecar Sync regularly, is safety. So when you think about AI, you have to think about where your data is going. You have to think about that with any software tool you use In the world of AI. The reason people are so concerned about data safety and data security is they're worried about the model providers having access to that data and then training future models with your data and then those future models, either intentionally or inadvertently, somehow like using your information to benefit others, leaking confidential information, et cetera. Now, there are absolutely ways to address this. There are terms of service and enterprise agreements that you can have with model providers that protect you from that. There's also ways to deploy models in a private manner, like we talked about, like taking an open source model and doing a totally secure private deployment.

Speaker 1:

There's lots of ways to deal with this, but you have to ask the question. You have to ask your software vendor what is your approach to AI safety and AI data privacy, and I think that's a good question to ask and you'll probably get some blank stares, unfortunately. But that's a really critical question to ask and the goal isn't to trip up the poor salesperson who's trying to get you excited about their AI feature, but rather to make sure that you realize there's actually some depth to what they have to offer, because if it's really part of the core of what the company does, they're probably pretty conversational in this topic. If it's more of like a little bit of gloss, they've put on a legacy product, which is unfortunately often the case, that's more of the AI washing scenario that the SEC chair was talking about.

Speaker 2:

That's really helpful, really helpful context. I hope we're not scaring too many sales reps in their boots right now of all these questions that we're going to get people to ask them at ASA Annual this weekend, but I think it's a really relevant conversation. And you're right, there are these edge cases and lots of platforms that wrap, which is WRAP, right, amith.

Speaker 1:

Yes, indeed, I mean with Suno, maybe it is RAP.

Speaker 2:

It's all in semantics, so I think it is important. But yeah, I mean.

Speaker 1:

I would just say one more thing on that topic, and that is you know, if really what's happening is like in the case I provided, where you know someone's providing, like the ability to type in a prompt and get a response and that automatically populates into some field in their system Again, it's not that it's not valuable, and if you do that a hundred times a day, of course it's extremely efficiency gaining. So that's great. But a lot of times it's something that you do like once, or you do it once a month, and then you have to ask the question well, is there really any additional value from doing it in the system versus if I just pull up ChatGPT in another tab, do it there and then copy and paste? So if the value is, is I'm eliminating copy and paste? There's really not much there. There has to be some meaningful use of AI.

Speaker 1:

Like I look for something that says, hey, what could I not have done prior to AI that I can now do?

Speaker 1:

You know, in this category of content, auto-tagging like I'm talking about earlier I'm super pumped about it because you know the machine is capable of automatically categorizing, tagging, classifying all of this unstructured data in such an incredibly powerful, accurate way for basically no dollars in basically zero time that you know. That's a game changer for a lot of organizations who don't either don't currently do a good job with that which is most organizations, or invest a tremendous amount of resources in that process, right, so that's just an example of something where there is a major shift in both capability and efficiency. But a lot of the thin wrapper apps that are out there, like we're talking about, and that's not even really like AI washing in the sense of what the SEC chair was describing, but rather just really like kind of trivial use cases, the AI washing is more of just basically like it's a BS layer saying, oh yeah, it's an AI powered, blah, blah, blah, blah, blah, blah. And the reality is is there's like zero substance behind that and unfortunately, most people don't know how to evaluate that and it just sounds really good. It's an AI powered widget. It's like sounds awesome.

Speaker 2:

Well now, all our listeners do know how to evaluate that. Moving on to topic two, figure two, the latest humanoid robot developed by Figure AI Inc. Which is an AI robotics company based in California. The second generation robot builds on the capabilities of its predecessor, figure 1, with advancements in design, functionality and AI integration. Functionality and AI integration.

Speaker 2:

Diving into some key specs here, I want to make the note that normally when I talk about key specs of a new large language model, I'm talking about parameters and context windows, but the specs here are height and weight, which I thought was kind of funny. But just so you can get an image, this humanoid robot is 5'6". It weighs 70 kilograms or about 154 pounds. It weighs 70 kilograms or about 154 pounds. It exceeds 1.2 meters per second in speed, just so you know. Average walking speed for a healthy adult is about 1.42 meters per second and it's got an impressive range of motion in the knee 135 degrees in terms of range of motion, and in the waist, 195 degrees. Amit sent me a video of this robot where you see it kind of folding over itself as you would in a horror movie. So I highly recommend you check that out. We'll link it in the show notes, but some other design and feature facts.

Speaker 2:

Figure two includes a new design with integrated cabling within the limbs, enhancing durability and preparing the robot for real world industrial applications. The hands have 16 degrees of freedom and human equivalent strength, allowing it to perform complex tasks requiring precision and delicate handling. It's equipped with six RGB cameras which capture light in red, green and blue wavelengths, and it can perform common sense, visual reasoning and obstacle avoidance. It also has three times the computational power and AI inference capabilities compared to Figure 1, enabling fully autonomous task execution. Through a partnership with OpenAI, figure 2 can engage in speech-to-speech conversations with humans using onboard microphones and speakers. It's been tested at a BMW plant, where it successfully performed tasks like placing car parts accurately. It's also designed to assist in various sectors like manufacturing, logistics, warehousing and retail.

Speaker 2:

It seeks to address labor shortages and improve workplace safety by taking on physically demanding and unsafe tasks. With backing from prominent investors like Jeff Bezos, nvidia, microsoft and OpenAI, the company is well-positioned to achieve its goal of creating robots that support human work and fill labor shortages. Amit, this might just be in my world, but I feel like a lot of people are talking about large language models. I feel like less people are talking about humanoid robots. I don't know if you agree with that, but if you do, why do you think that's the case?

Speaker 1:

I think fewer people are talking about it for sure. The way I think about it is. You know, large language models and AI models in general have really taken the world by storm in the popular consciousness in the last year and a half, two years. For good reason, because they introduced a capability that previously was really not a capability, right? So really tremendous advancements there.

Speaker 1:

And we talk about the world of bits versus the world of atoms, aka the digital world versus the physical world, all the time, and you know, I think, that the two are intersecting in really interesting ways. We talk about multiple exponential curves in terms of advancements in price, relative to performance in various fields, and so we've talked about that in the context of material science previously. We've talked about that in the context, of course, about AI and raw compute. We've even touched a little bit on quantum computing in the past, and all of these exponential curves are moving at a very rapid pace, obviously, and keep going faster and faster. It's the nature of exponentials. Well, the convergence of these exponentials is resulting in new innovations in the physical world, such as the one we're talking about now, which is humanoid robots. Of course, robots will have many form factors. Humanoid robots are interesting for reasons we can get into in a second, but I think that this is a really, really big deal because it's showing you that the hardware world is advancing just as rapidly as the software world. Hardware normally works pretty slowly, you know it's on a relative basis. Software can advance quite quickly. Hardware tends to be a little bit slower.

Speaker 1:

This is a startup that's only been in existence for about three years, a startup that's only been in existence for about three years. So you compare them to a company like you know, Boston Robotics or you know any of the other ones that have been around for a long time. It's really quite stunning what they've been able to do with limited resources and very little time. So I think it's a testament, not so much to whether or not their product will be successful, though it looks pretty impressive. I don't know what the price is and what the applications initially will be, but I think it's more about like, look, these things are real, like as creepy as they may seem, and to me they're super creepy. They're going to be here and they're going to be in industrial settings. They're going to be in probably warehouses. They will be in probably retail settings, maybe in food service and maybe at some point probably retail settings, maybe in food service and maybe at some point even in the home.

Speaker 2:

Probably not in my home. I say that now, but I feel like you would have one.

Speaker 1:

You know, if the utility is high enough, I might, although the humanoid form factor in that context would be really creepy. I mean, imagine waking up in the middle of the night, you're kind of groggy and there's this silhouette of a human form factor. No, thank you. You know I love computers and technology, but not that much so in any event, though. So why is the humanoid form factor so interesting? It's because think about things in the context of the world we have constructed.

Speaker 1:

We have built a world over thousands of years that is designed for the human form factor because we are the users of the technologies.

Speaker 1:

We drive cars, we turn on light switches, we open doors, we go up and down stairs.

Speaker 1:

So if we design a robot, that is a humanoid form factor, that means it can interface with the real world, the physical world, just like we can, and so it's a very complex thing, because, you know, evolution has given us hundreds of millions of years of time to create what we have today in our species, and it's a pretty amazing machine, but like, for example, just the one advancement you talked about briefly in terms of the hand, that has been one of the areas of robotics that has been something of a grand challenge in that, you know, having both strength and sensitivity to be able to deal with both gripping strength like moving a box, or gripping a cable or something like that and at the same time, having enough sensitivity and speed and adjusting the pressure levels and all that so you can both, like break an egg and fry it, but also, you know, move a box that weighs 80 pounds.

Speaker 1:

So we have some pretty amazing hardware ourselves. So I think that these advancements are really quite interesting and they all compound each other right. The material science makes it possible for this thing to weigh 150 pounds, not 1500 pounds. That in turn allows the. It's electric, so it allows battery power to be sufficient. It's not hydraulic based, and the fact that it's not hydraulic based means that it also is lower maintenance and lower cost and more and more software controllable. So all of these things compounding each other in terms of the innovation cycle.

Speaker 2:

Yeah, there's something about this that's just off-putting to me. It's really exciting, but the video mixed with thinking of all the potential applications I'm also thinking potentially with defense in the future using robots like this it's just, it's definitely you can get on a spiral and go down a dark path. Thinking about it, do you think our listeners would have any use for humanoid robots in the workforce, like in their organizations?

Speaker 1:

I'm not sure about within the office work of an association. I think there are places where you know having some support in a physical form could be useful, particularly heading to a conference, if you had some robots helping you move objects around. Of course, you know, depending on the city you're in the union, labor may have a problem with that, but you know, I think, that there will be solutions for that as well where the robots might actually join the union. I don't know how that's going to work out, but that will be a thing at some point. My point would be this that these things you can't really fight the evolution of technology. It's too powerful of an economic force, so it's going to be there. I think that in the context of the physical world, associations will have variable levels of use for this stuff.

Speaker 1:

I do think in the fields that associations represent, there will be tremendous use in many examples. So you know, you think about, for example, places where we have labor shortages, where we're not talking about displacing workers, which is obviously a problem at so many levels, but you're talking about places where there aren't enough workers to fulfill the need. A good example is senior care. We have more and more people aging and we have not enough today in terms of qualified medical professionals, and you know people that are around the medical field to care for the elderly and to do things like help them use the bathroom or whatever the case may be. So those functions are very difficult and it's also, you know, very much heavily skewed based on income and assets. So you know people that don't have the resources to hire in-home care end up in homes. Robotics like this, if made available enough and inexpensive enough, could help improve safety for seniors at home. To live independently for a longer period of time could help people be safer in terms of what they do take care of a lot of things. So I think there's some interesting applications that I don't think people would argue with.

Speaker 1:

In terms of the ethics, there may still be somewhat of a creepy factor to it, of course, but if you imagine that and say, hey, you know, let's just say this robot cost $2,000 or something like that, I think it's dramatically more than that. At the moment it's probably 100 times that. But if it costs $2,000, $3,000, $4,000, something that's, you know, expensive but like, not outside of the realm of possibility for a fairly wide range of people and it opens up the door to more freedom for people as they age. Just going back to that one example, that's really interesting. It also solves for a major labor shortage problem.

Speaker 1:

So, getting back to your question, I think it's totally dependent upon the sector and association represents and the work that they do. So the other thing, too, I would say is this is far outside of my area of expertise and I think that there are probably hundreds of use cases for this technology that other people would come up with. But even like in the world of food service, like I mentioned, I know there's, there are robots in that environment and there's, you know, a continual labor shortage in that world too. So you know, I think there's some interesting possibilities in that arena as well.

Speaker 2:

Mm-hmm. So I mentioned with the topic intro that they're designed to assist with manufacturing, logistics, warehousing and retail and, like you said, that's probably just a few of many potential industries that will see these humanoid robots. If you were running an association of members that were in those respective industries, would you be afraid? What would you be doing to prepare for that?

Speaker 1:

Well, I think it's like everything else we talk about in this pod. I mean, you know, fear generally comes from a lack of understanding, sometimes like the more you understand something, the more afraid you should be, because something is just damn scary. I don't know if this is that for some people, but I think that most fear comes from a lack of understanding, and you can solve for that. You can learn about these things, you can experience these things, you can figure out, like, how they impact you. So the first thing I would do if I was the CEO of an association that was something deeply in the physical world and something industrial or something warehousing or something like that, is I would really go deep in understanding this technology. I would try to think in an exponential way, say, ok, well, this is where this is today. What's it going to be like in the next one, two, three, four years? Chart that out based upon what's happening in the space, and then say what impact will that have on the workforce that we are here to help? How can we prepare our workforce to adapt to this environment? What can we do to better partner with employers so that the people that we represent you know whether it's in a labor capacity or if it's a professional education capacity. How can we better like help our people move up the value chain if some of the lower level tasks are in fact going to be displaced? Certainly there are ways you can try to fight these things right, and they may have some degree of temporary relief, and I think that might be the right tactic for some people, depending on where they're at and what they're doing. But I think understanding this stuff is so critical the closer you are to where this technology will displace people. But again, I think there will be that and there will also be places where there's so much demand for additional service that I think people will actually do better.

Speaker 1:

You know, talking about that home health care category again, that's a really tough job. People who care for the elderly or for the disabled in their homes are subject to extremely long hours, backbreaking labor. Sometimes they're, you know, under duress, where the people they're caring for, whether intentionally or not, are, you know, quite negative, to put it mildly, to the people who are caring for them, and on and on and on. And so if you had some assistance for some of those folks, I think that could be interesting, right? Rather not necessarily displacing them entirely, but helping them care for people while you know, having more dignity and having a greater level of safety as well, both physically and psychologically.

Speaker 1:

So I think there's just there are potential elements to this that open up markets that currently are closed.

Speaker 1:

Right, it's the best way to probably frame it all up is we're going from a world of scarcity to a world of abundance, because if we're talking about intellectual labor in the context of AI models that are purely encapsulated in the digital realm as bits, right, so that's purely intellectual labor, it's the white collar world of thinking, decision-making, writing, coding, whatever.

Speaker 1:

And then, in the physical world, we're talking about also going from a world of scarcity to abundance, where labor is very expensive and a lot of times the skills you need are not available, and this could change that right, so it could bring abundance to the world of physical labor, and that is again just like with white collar labor, both super exciting and terrifying in some ways. So to me, the solution is always the same thing is we have to dig into it, because if you want to try to stop this stuff, you know, good luck to you if that's your approach, but I think you stand zero chance against the tidal wave. So to me it's about educating yourself and really thinking critically and as objectively as you can in terms of what's going to happen. Don't don't be sitting here going, oh yeah, no, it's not going to affect our sector that much, because you know we're just immune to it and that's just BS. So you got to dig into this stuff.

Speaker 2:

So step one being, as we always say, educate yourself, and then it sounds like step two this is what's a little less tangible is think critically. But a bit of guesswork as well to imagine in three years what this could look like. Is that right?

Speaker 1:

I think so. I mean, all of us are guessing at this point. Exponential curves are super hard to figure out. It's really hard to say. You know people who are in. The deeper you are actually, the more you realize that I think. But you know people say, oh well, you know, we got this last year, we're getting this this year, so next year it should be the same progression. And it's like no, actually it's going to be significantly more than that and we have so much, you know, empirical evidence that suggests that that is going to continue for the foreseeable future. So yeah, I mean I think.

Speaker 2:

I think it's everything you just said, mm hmm, in topic three today we're talking about SAM2, or a segment anything model two, which is the latest advancement from Meta AI in the field of visual segmentation, building upon the original segment anything model. Now I will say for all you listeners, I normally use perplexity to do a lot of the research for this podcast. Perplexity is a tool that I love and for this one in particular, I'll admit I wasn't fully grasping what a segment anything model did, and so I actually asked perplexity like can you dumb it down a little bit for me? Can you talk to me like I don't know anything about this? And it gave me some really good insights. So I think that goes back to your point, amit, about transforming content for your level of education and understanding. It's not something I considered using perplexity for before, but I just wanted to share that note with you all. So here are some of those easier to understand tidbits from perplexity Sam can automatically find and highlight objects in a photo.

Speaker 2:

For example, if you have a picture of a dog in a park, sam can outline the dog precisely, separating it from the background. Doesn't matter if it's a dog or a tree or a car or a person. Sam can handle all kinds of objects without needing to be specifically trained for each one. You can give Sam simple instructions, like pointing to a part of an image or drawing a box around an object, and it will know what to highlight. You can also use other types of instructions, like rough sketches or even text descriptions. Though text prompts were not initially available, they now are.

Speaker 2:

And why exactly is this kind of model useful? Well, if you're a graphic designer or photographer, sam can save you a lot of time by automatically cutting out objects from your photos, which you would otherwise have to do manually. For researchers or developers who need to label objects and images for their projects, sam can do this quickly and accurately, reducing the need for tedious manual work. Whether you're making a collage or creating special effects, working on an art project, sam can help by easily isolating objects from images. In healthcare, sam can help doctors by highlighting specific parts of medical images, like organs or tumors, making it easier to diagnose conditions. And for autonomous vehicles, sam can help the car's system understand its surroundings by identifying and tracking objects like pedestrians, other vehicles and road signs. And I'll also add after I did all my research, I realized do you have an iPhone, amit?

Speaker 1:

Yes.

Speaker 2:

I don't know if you know that you can actually do that in your photos, and so the whole time I was researching this, I realized well, wait, I think I can do this exact thing on my phone, and I looked it up and there's machine learning behind that and I had no idea that was AI. So I thought that was an interesting point to share with listeners too.

Speaker 1:

Yeah, you've seen bits and pieces of these concepts come to fruition over roughly the last decade, a little bit longer than that. You know, a lot of the image, visual kind of stuff that you're hearing about now came from some work in the deep neural net space through this contest, this academic contest from you know, 15 years ago it started called ImageNet, which led to really exponential growth in the capabilities to recognize objects and images, and back then it was simply being able to, you know, generate some small text tags saying oh, this is a cat, this is a dog, this is a cat in a park, this is a dog on a sofa, or things like that. And obviously this is a generation, multiple generations further along. And it's also more functional in that what you're talking about when you're identifying specific objects in a photo or a video, by the way, it's both still and live. You know it could be live, real time video. It could be.

Speaker 1:

You know, a recorded video is really interesting because it opens up a lot of potential applications. You've covered some of them in terms of creative work. The other thing, too, is when you're thinking about your content as an association. A lot of associations have as part of their content tons of images, tons of charts, in some cases lots of videos, a lot of different content assets and these are just stored pretty much just as big chunks. They're files, right, and there might be some tags on them and descriptions. But if we can utilize this technology to fully automate the segmentation or object recognition and object tracking, it opens up a lot of interesting potential applications to better highlight and annotate interesting things and videos from an education perspective and a fully automated basis to be able to update content automatically as things are changing. So let's just say, for example, we had a video and we have a new version of a product, but we have some videos that we've previously captured that shows the demonstration of that product. Like I'm thinking of a physical good. We can update that object in the video pretty much automatically even with current technology, not have to reshoot the whole thing. So there's use cases like that. There's a whole bunch of different things you can do and really the broader category here of computer vision is an interesting thing because it says, okay, well, computers can make sense of what they see.

Speaker 1:

We've talked in the past about the idea of AI having a world model, which they don't currently have generally. So these AI models that have been trained primarily on a corpus of text have a theoretical understanding but have never interacted with the real world. So they have read everything that has been written, perhaps they've listened to some stuff and maybe they've also watched some videos, but they've never actually been in the real world. So they have read everything that has been written, perhaps they've listened to some stuff and maybe they've also watched some videos, but they've never actually been in the real world. They've never experienced life in three dimensions, right? They never experienced touch and reactive to touch. They've never smelled anything or felt the breeze or any of these other things that are phenomenon in the physical world. So what computer vision does, as computer vision becomes far more sophisticated, is gives AI models more high fidelity insights into what they're experiencing. So, rather than just looking at a stream of pixels over time and then trying to kind of make sense of that based on pattern recognition, which is the statistical approach, and how a lot of the models are working, you know, if you think about like video generation models, they're not based upon so much like okay, I know exactly what's happening on a per object basis in each frame. They're more based on patterns and that's starting to change.

Speaker 1:

So this type of technology is basically like a way higher fidelity visual model and that will make its way into everything, notably SAM, which is from, as you mentioned, the FAIR lab at Meta is an open source innovation, so they put that out there as a totally free thing. I mean there are some license restrictions, primarily saying you can't compete with them using that tool, which I think is probably pretty reasonable, but it is an interesting contribution to the research community. I think you're going to see this appear in a ton of different projects. So Apple, as you said, has had I think Android has this too has had, you know, object recognition of some sort, primarily individual recognizing people and being able to group. You know all the photos of certain people or in certain locations, and the location part's partly the GPS tagging that most phones do, but it's also based on notable objects. But it's been much more primitive in terms of how that's been done. So if you have a photo with the Eiffel Tower in the background, they have a database of well-known public images and so they'll be able to match on that.

Speaker 1:

But I think this type of technology is far, far more sophisticated than that. So I find it really fascinating. I think it's an interesting thing to be aware of because, in parallel with what's happening in the world of language models, you have these innovations occurring and they're going back and forth. As you said, it didn't originally have text-based prompting, it was more point and click and say, hey, this is a box, I'm going to drag around this object now track the object, which is cool. But you can say, hey, I want you to track all the dogs, and so if there's five dogs in the video, it will track all five dogs as individual objects moving around in the video. So it's a pretty impressive piece of technology and, by the way, the whole thing's real time and runs on a very limited amount of hardware, so it's a really impressive advancement.

Speaker 2:

So this is something altogether different from a large language model.

Speaker 1:

Yes.

Speaker 2:

Okay.

Speaker 1:

But there's a language component because of the text prompts yeah, I mean, these things are all starting to weave back and forth in terms of their modalities. So this is primarily a vision advancement, right, but it has some language capabilities, and those language capabilities I don't know if that's part of the model training itself or if it's an engineering thing where they did an overlay on top of it where the first thing, or if it's, you know, an engineering thing where they did an overlay on top of it where the first thing you know. So if you put a prompt that says track all the dogs, they might have a language model, like probably Lama 3.1, that says okay, you know from that what does this mean. And then there's an identification of objects which might be in the model or might not be, and then that gets translated to hey, sam, these are the points in the picture as of this frame that I want you to track, and then the model, the visual model, tracks it from there. So I don't know how they actually built it, but it doesn't really matter actually, for as a practical concern, because whether these are multiple models working in unison or if it's like one big integrated multimodal model, as we say, they don't really matter.

Speaker 1:

The capability is what matters, and so it means as an association in your own content, which a lot of times is educational or informative for your community. If you have images and videos, this could be interesting. Right, it's like knowing the tool exists that you can make sense of your catalog of images and videos, possibly extracting more specific, you know, detailed insights from those images which could be interesting, like sometimes people say, well, you know on average how many people attend our events. You know, or on average, how many people go to our parties. You can say, okay, well, you know, sam, how many people are in this picture, and then you know to that across every photo you've ever taken at all the parties at your prior annual conference. I mean, it's probably a silly example, but that would be a highly manual task before this.

Speaker 2:

Anything else you want to cover here?

Speaker 1:

So you know, we talk about training AI models on video and images, and we've talked about that in the past, about how multimodal models will be trained on not just text but also video, audio images, et cetera, and they'll also be able to interact with their users through all those modalities. Through now, what we've been saying essentially is we're just going to feed these images and videos and audio clips into the models pre-training, which means that they're just getting, kind of, you know, processed by the training algorithm which is really looking for pattern recognition. That's how the neural net's been trained. What's happening with something like SAM is you're getting additional information, so you're taking that same image but you're breaking it down into its objects and in the context of a video, you're saying, hey, object A is moving in this direction, or this is what's happening, and that's really interesting because, beyond the raw image and pixel data right, that's coming in that the model can train on this higher order insight that comes from a segment anything type model is super interesting because that could lead to smarter models with world models built into them, because they'll have better understanding of the physical world.

Speaker 1:

So I think this becomes a component of future infrastructure that associations may use. I don't know that any association may use this directly, but I think it's one of these innovations like when we talked about advancements in AlphaFold3, or when we talk about advancements in material science and the GNOME project or anything like that those are not likely to be used by the association community directly. But aside from being just interesting intellectually, there are yet additional milestone markers on this path to more generally intelligent machines, and so that's why I find this particular topic really interesting. It represents a new, nearly free capability that you will find in every model, and so not only on phones but like literally everywhere. You will assume the computer has a very high level understanding of what's going on in the images. It's not just a picture, it's going to be an interactive thing. That will be something you can pick apart, put back together, and the computer will have a comprehensive understanding of what it is.

Speaker 2:

I think that's a great way to wrap up the episode of me. Everyone. Thank you for tuning in to today's episode of the Side Car Sync. If you are attending ASAE Annual reminder, we will be at booth 939, if Amit remembers it, and we would love to see you there.

Speaker 1:

I might show up at a different booth, but I think we also have an event on Sunday evening, if I remember correctly. Maybe let everybody know what that is.

Speaker 2:

We do, we do. How did I forget we have a sidecar? Asa annual happy hour. Well, it's our happy hour at ASA annuals event. It's Sunday at 4.30 PM at Punchbowl Social, which is about a 10, 12 minute walk from the convention center, so I hope you all will join us there. We'll be having some tacos, drinks, more books, more Ascend books and some other swag that we're handing out and some great conversations. So please stop by and see us Sunday, 4.30 PM, punchbowl Social.

Speaker 1:

Sunday afternoon at 4.30 at ASAE. Tacos and beer Sounds good.

Speaker 2:

I'll see you then, Amit.

Speaker 1:

All right, sounds good. Thanks for tuning into Sidecar Sync this week. Looking to dive deeper? Download your free copy of our new book Ascend Unlocking the Power of AI for Associations at ascendbookorg. It's packed with insights to power your association's journey with AI. And remember Sidecar is here with more resources, from webinars to bootcamps, to help you stay ahead in the association world. We'll catch you in the next episode. Until then, keep learning, keep growing and keep disrupting.