5/23/2023 – BuiltOnAir Live Podcast Full Show – S14-E07

Duration: 65 minutes

Be Sure to Subscribe to the podcast!

FULL EPISODE VIDEO

Watch the full video of the show. See below for segment details.

FULL EPISODE AUDIO

Listen to the full Audio podcast for this episode here. Or add to your favorite podcast player

Listen On: Apple | Overcast | Spotify

The BuiltOnAir Podcast is Sponsored by On2Air – Integrations and App extensions to run your business operations in Airtable.

In This Episode

Welcome to the BuiltOnAir Podcast, the live show.  The BuiltOnAir Podcast is a live weekly show highlighting everything happening in the Airtable world.

Check us out at BuiltOnAir.com. Join our community, join our Slack Channel, and meet your fellow Airtable fans.

Alli Alosa – Hi there! I’m Alli 🙂 I’m a fine artist turned “techie” with a passion for organization and automation. I’m also proud to be a Community Leader in the Airtable forum, and a co-host of the BuiltOnAir podcast. My favorite part about being an Airtable consultant and developer is that I get to talk with people from all sorts of industries, and each project is an opportunity to learn how a business works.

Kamille Parks – I am an Airtable Community Forums Leader and the developer behind the custom Airtable app “Scheduler”, one of the winning projects in the Airtable Custom Blocks Contest now widely available on the Marketplace. I focus on building simple scripts, automations, and custom apps for Airtable that streamline data entry and everyday workflows.

Dan Fellars – I am the Founder of Openside, On2Air, and BuiltOnAir. I love automation and software. When not coding the next feature of On2Air, I love spending time with my wife and kids and golfing.

Show Segments

Round The Bases – 00:04:25 –

Meet the Experts – 00:27:22 –

Meet Lorenzo Lee.

Lorenzo is the Founder

Visit them online

Meet the Experts – 00:30:56 –

Meet Bill French.

Visit them online

Industry View – 00:36:08 –

Learn all the ways you can use Airtable within the AI & Airtable Industry. From pre-made bases, apps, automations, and custom bases, we’ll show you how to set up bases and get your AI & Airtable organization established in Airtable.

Our panel of experts will dive deep into the world of AI and how it relates to Airtable.

Click Here

Full Segment Details

Segment: Round The Bases

Start Time: 00:04:25

Roundup of what’s happening in the Airtable communities – Airtable, BuiltOnAir, Reddit, Facebook, YouTube, and Twitter.

Segment: Meet the Experts

Start Time: 00:27:22

Lorenzo Lee –

Meet Lorenzo Lee.

Lorenzo is the Founder

Visit them online

Segment: Meet the Experts

Start Time: 00:30:56

Bill French –

Meet Bill French.

Visit them online

Segment: Industry View

Start Time: 00:36:08

Airtable for the AI & Airtable Industry

Learn all the ways you can use Airtable within the AI & Airtable Industry. From pre-made bases, apps, automations, and custom bases, we’ll show you how to set up bases and get your AI & Airtable organization established in Airtable.

Our panel of experts will dive deep into the world of AI and how it relates to Airtable.

Click Here

Full Transcription

The full transcription for the show can be found here:

[00:01:40] All right, welcome back to the
[00:01:43] BuiltOnAir Podcast. We are in season 14,
[00:01:45] episode seven.
[00:01:47] We are back from
[00:01:49] in live, at least four out of the five
[00:01:51] on this show were in live in person in
[00:01:53] San Francisco last week
[00:01:55] at the dare table conference. So
[00:01:58] if you did not check out our show,
[00:02:00] we did a live show, during the event,
[00:02:03] we pulled on
[00:02:04] how many people did we
[00:02:05] talk to Kamille and Alli? Probably, I,
[00:02:08] I think we had about 4,5 rounds
[00:02:11] and some of them were pairs.
[00:02:14] So
[00:02:16] a good 12, 13 people.
[00:02:20] And we could have gone longer,
[00:02:21] but the, the, the show was continuing on.
[00:02:24] So we had to cut it short,
[00:02:25] but there was lots of people
[00:02:26] we could have spoke to.
[00:02:27] There was,
[00:02:27] it was a great event.
[00:02:29] It was awesome to just be in person and,
[00:02:31] and, geek out on Airtable,
[00:02:34] any, any comments on Daretable,
[00:02:36] those that went?
[00:02:39] I loved it.
[00:02:40] I
[00:02:41] thought it was just so neat.
[00:02:45] Yeah.
[00:02:46] Yeah.
[00:02:47] So good times.
[00:02:49] Then we went out and,
[00:02:51] chatted more over dinner.
[00:02:53] So it was a good two day event. And
[00:02:56] I believe they're going to
[00:02:57] Oslo Norway next and then
[00:03:00] in the fall and then,
[00:03:01] New York next year.
[00:03:03] So,
[00:03:04] feel free to get those on your calendar.
[00:03:08] Ok. So let's, kick off the show.
[00:03:10] We've got some amazing
[00:03:11] guests with us today.
[00:03:12] We're gonna be diving into the
[00:03:14] world of A I chat GPT
[00:03:16] all the buzzwords you've
[00:03:17] been hearing about.
[00:03:18] We're gonna figure out what that
[00:03:20] means for Airtable.
[00:03:22] So I'll go through what
[00:03:23] the show will be for today.
[00:03:25] We'll always, we'll start with our
[00:03:26] Round the Bases,
[00:03:27] get you up to date with
[00:03:28] what's going on,
[00:03:29] what's new in the world of Airtable
[00:03:31] and then a spotlight on On2Air,
[00:03:33] our primary sponsor.
[00:03:35] Then we'll get some background
[00:03:37] on our two guests, Lorenzo
[00:03:39] and Bill learn a little bit about them
[00:03:42] and, and what they're up to
[00:03:44] and then we're gonna dive deep
[00:03:46] into the world of A I open it up for
[00:03:48] discussion. We might share some,
[00:03:51] some insights from our guests and
[00:03:54] things that they've done.
[00:03:56] And hopefully the goal is to, to
[00:03:58] make you
[00:04:00] a little bit more prepared of the
[00:04:02] changing world of A I and how it might
[00:04:04] impact you in using Airtable
[00:04:06] or in your business and your, in your job,
[00:04:09] what it might be.
[00:04:10] So hopefully you gain some
[00:04:12] insights that can help you
[00:04:13] be more effective
[00:04:15] and then we'll give a shout out to
[00:04:17] how to join our community
[00:04:19] and then wrap it up.
[00:04:20] So
[00:04:21] with that, Round the Bases,
[00:04:24] there's really only one thing
[00:04:26] I want to leave a lot of time for,
[00:04:27] for our panel discussion.
[00:04:29] So as far as Round the Bases,
[00:04:31] it's been two weeks
[00:04:33] since last week, we didn't do this.
[00:04:35] So prior to that,
[00:04:36] since the last time,
[00:04:37] we've, we've been on this format for a
[00:04:39] show
[00:04:40] Airtable had some pretty big
[00:04:42] product updates.
[00:04:44] So
[00:04:45] they announce their A I,
[00:04:47] let's save the conversation.
[00:04:49] We'll talk about A I's
[00:04:50] launch in the A I discussion
[00:04:52] or Airtables launch
[00:04:54] of or, or announcement of
[00:04:56] their A I product.
[00:04:57] We'll save that discussion but
[00:04:59] let's talk about some
[00:05:00] of their other stuff.
[00:05:01] So let's see.
[00:05:03] Besides A I,
[00:05:04] we've got
[00:05:06] pre configured apps, so apps by Airtable.
[00:05:10] So that includes,
[00:05:13] so anybody wanna give a summary
[00:05:15] of apps by Airtable? What do,
[00:05:17] what do they mean by that?
[00:05:19] I
[00:05:20] mean,
[00:05:21] it's really just a glorified template.
[00:05:23] If I'm understanding everything
[00:05:24] correctly, it's just, it's a template that
[00:05:28] has interfaces all built into it
[00:05:31] and they've built in all sorts of
[00:05:33] different tools that they think
[00:05:35] you might need based
[00:05:36] on what industry you're in.
[00:05:37] Like marketing
[00:05:40] or product I think is the other,
[00:05:42] I can't quite recall
[00:05:44] what the marketing.
[00:05:49] Yeah, the
[00:05:51] product, yeah.
[00:05:54] Interesting though,
[00:05:55] they also say that
[00:05:56] there's 10 that they're,
[00:05:56] they've released but
[00:05:58] they've only released two.
[00:05:59] Well, I think we're counting
[00:06:01] all of these subsets within products.
[00:06:05] So. Ok.
[00:06:06] Yeah,
[00:06:07] I mean,
[00:06:09] I'm not super
[00:06:10] excited
[00:06:10] about it.
[00:06:11] Yeah, six marketing ones
[00:06:13] and four product ones. So there
[00:06:15] are 10,
[00:06:17] all within the same two template
[00:06:21] apps. Right. Yeah.
[00:06:24] So it's cool. I mean,
[00:06:25] they're just more built out templates,
[00:06:27] really. Yeah.
[00:06:29] Yeah.
[00:06:30] It's decidedly biased towards marketing.
[00:06:33] Yeah.
[00:06:34] Yeah,
[00:06:35] I remember a day when
[00:06:37] a very large company biased
[00:06:39] themselves towards marketing,
[00:06:41] I think it was sales force,
[00:06:43] wasn't it, Dan?
[00:06:45] I would assume they're biased
[00:06:47] towards sales but maybe not.
[00:06:51] Then they branched into marketing.
[00:06:53] Yeah.
[00:06:55] Yeah.
[00:06:56] But yeah, so it's basically,
[00:06:58] yeah, similar to like the, the,
[00:07:00] the templates in the Airtable universe.
[00:07:03] This is kind of templates at the
[00:07:05] interface level.
[00:07:08] Yeah, that's probably the easiest way. So,
[00:07:11] so they had to take away
[00:07:12] if you remember what,
[00:07:14] a year ago, year and a half ago,
[00:07:15] maybe two years ago.
[00:07:17] When did they take apps away
[00:07:19] from the extensions?
[00:07:20] And
[00:07:21] they had to pull that name back
[00:07:23] because they had these plans to,
[00:07:25] to repurpose apps.
[00:07:29] So, do you, do you get the sense
[00:07:30] that they're going to
[00:07:32] expand this to beyond
[00:07:33] marketing to other
[00:07:34] like hr and?
[00:07:37] They said they were going to at the,
[00:07:40] at the dare table conference 2023.
[00:07:42] They had, this was the first opening
[00:07:47] channel, I guess.
[00:07:48] And they were talking about a lot of the
[00:07:51] recent announcements and announcements
[00:07:52] they were gonna make in the future.
[00:07:54] One of them
[00:07:55] was
[00:07:55] by
[00:07:55] Airtable and they are working on other,
[00:07:59] you know, branches of a company.
[00:08:01] It's just that marketing
[00:08:02] was like the biggest,
[00:08:04] I guess out of all of the companies
[00:08:06] that they were talking to,
[00:08:07] to try and figure out
[00:08:08] what should go in the templates.
[00:08:10] They're like we got the biggest
[00:08:11] response for marketing. So we're starting
[00:08:13] with marketing.
[00:08:14] I
[00:08:14] I wonder what percentage of
[00:08:17] customers, maybe new customers
[00:08:20] dive in and start in something like this.
[00:08:24] Because I know from
[00:08:26] my experience and watching other
[00:08:28] companies,
[00:08:29] there's only a one in three
[00:08:30] chance that they're gonna pick a template,
[00:08:33] they might pick a template to learn by,
[00:08:34] but they'll never pick that template to
[00:08:36] run by.
[00:08:37] And so I'm just wondering
[00:08:39] what their analytics show for that.
[00:08:45] And then I guess my question is,
[00:08:47] is you still can't use these,
[00:08:48] right? Like I don't see a click,
[00:08:51] a link to install or test,
[00:08:53] right? I think you have
[00:08:55] to sign up for like a wait,
[00:08:56] not a waitlist, but like
[00:08:58] you have to contact them in some way
[00:09:01] so they can do a demo of them maybe.
[00:09:03] But yeah, these are not yet
[00:09:05] usable templates
[00:09:09] by the masses. So
[00:09:12] I wonder, I'll be curious if they
[00:09:14] if they expand the universe
[00:09:16] to also support
[00:09:17] interface templates where
[00:09:19] anybody could submit.
[00:09:20] That would be nice as well.
[00:09:24] Yeah,
[00:09:25] they did say something about
[00:09:27] automations as well. So I wonder if
[00:09:30] that will
[00:09:34] you know, expand itself into
[00:09:35] regular templates and,
[00:09:36] and things in the universe
[00:09:37] because when you put something
[00:09:39] in the universe,
[00:09:40] automations don't come with it
[00:09:42] currently.
[00:09:43] Right. Right.
[00:09:46] Yeah. Yeah, they definitely need
[00:09:48] to enhance that with all of this.
[00:09:50] If, if these are shareable,
[00:09:51] then I would imagine they would come with
[00:09:54] automations, but
[00:09:57] yeah, so, so yeah, you can kind of see
[00:10:01] the main thing. The main difference is
[00:10:03] if you know this side bar,
[00:10:05] so we could jump to the,
[00:10:06] the big announcement that just well,
[00:10:09] it wasn't an announcement I don't believe,
[00:10:11] but they've started to roll out the
[00:10:15] the sidebar being collapsible.
[00:10:18] So Mecca is the first I saw it
[00:10:21] mentioned.
[00:10:22] He saw, he started seeing it this morning
[00:10:25] and you can see there's a,
[00:10:28] there's a gif of
[00:10:30] the collapsing.
[00:10:32] So now that was a big
[00:10:36] if you go to the,
[00:10:38] the post about the new features,
[00:10:41] it's just pages and pages
[00:10:43] of commentary, 12 pages. It like
[00:10:45] the one thread in the comment
[00:10:47] in the community,
[00:10:48] the Airtable community that has
[00:10:49] any discussion and it's
[00:10:51] mostly mostly very critical
[00:10:53] of the changes to the sidebar.
[00:10:57] It was a train wreck.
[00:10:59] Yeah. And, and now it
[00:11:01] has 111 cars on it.
[00:11:06] It's the only train wreck that
[00:11:08] actually grows in length as it unfolds
[00:11:12] it, it's, it, it was unfortunate.
[00:11:14] I mean, my takeaway to that is that
[00:11:16] 100% could be avoided
[00:11:19] and they found a way to not avoid it.
[00:11:22] If you have to work extra hard
[00:11:25] to get into this kind of a, a
[00:11:28] storm.
[00:11:31] It was frustrating.
[00:11:32] I'm glad that you can now
[00:11:34] collapse it. I think that was
[00:11:35] probably the number one
[00:11:37] complaint from people.
[00:11:38] I think
[00:11:39] my other complaint is like
[00:11:41] I created separate
[00:11:43] interfaces like it sets of interface
[00:11:46] pages for a reason. And now they're all
[00:11:48] lumped together on the same menu,
[00:11:50] on the left hand side bar
[00:11:52] with no control over
[00:11:53] what appears there or not.
[00:11:55] And that part is bothersome to me.
[00:11:57] So hopefully we get a little bit more
[00:12:00] control over what pages actually appear.
[00:12:05] I mean, it's almost like
[00:12:07] you know, you imagine
[00:12:09] you were making pottery
[00:12:11] and you shaped your pottery
[00:12:13] to be exactly the way,
[00:12:14] you know, your art form
[00:12:16] and your customers appreciate it
[00:12:19] and then
[00:12:20] you came into the store one morning
[00:12:23] and all of your shapes
[00:12:24] were changed.
[00:12:28] Ok. What, what would the pottery maker,
[00:12:31] how would they feel
[00:12:33] if that just magically happened?
[00:12:36] And,
[00:12:37] and customers looked at it and went,
[00:12:40] I'm not buying that,
[00:12:42] right? They just totally rejected it.
[00:12:44] But
[00:12:45] I mean, that,
[00:12:46] that's just complete insanity
[00:12:48] and that's why I always say
[00:12:50] if you're gonna monkey
[00:12:52] with the U I or the U X
[00:12:54] you make sure that you have a
[00:12:56] way to be backward compatible.
[00:12:59] The world of tech believes
[00:13:01] that backward compatibility
[00:13:03] relates to data
[00:13:05] and only data
[00:13:06] and that is a fundamental mistake to,
[00:13:09] to frame it that way
[00:13:10] it relates to backward
[00:13:13] compatibility of user experience.
[00:13:16] And if you don't do that,
[00:13:17] you end up with this kind
[00:13:19] of a train wreck.
[00:13:21] Yeah,
[00:13:22] that's what she said right here.
[00:13:23] That, that is correct.
[00:13:25] I think that if I died tomorrow,
[00:13:27] that would be the one
[00:13:28] thing
[00:13:29] I hope people would remember.
[00:13:31] Legacy U I behaviors
[00:13:32] cannot be underestimated
[00:13:34] every ounce of your success
[00:13:36] is predicated upon what happened before.
[00:13:39] Not what will happen next.
[00:13:42] Yeah.
[00:13:42] And, and they just totally
[00:13:44] like said, oh screw that.
[00:13:46] We, we know better.
[00:13:48] And that,
[00:13:49] that U
[00:13:50] is, is what got them into deep trouble.
[00:13:54] I like, I do one thing I like about
[00:13:57] like a thread like this is you get people
[00:13:59] out of the woodworks that aren't,
[00:14:01] you know, active on other communities.
[00:14:03] You're like, oh, there's,
[00:14:04] there's a lot of other people
[00:14:06] out here using this tool.
[00:14:07] Right.
[00:14:08] Right. And, and they
[00:14:09] have similar sentiments.
[00:14:10] Yeah.
[00:14:11] Yeah.
[00:14:11] It, it definitely pulls people in
[00:14:15] but probably not for good reasons.
[00:14:19] So yeah, so that, yeah,
[00:14:21] definitely lots. If, if you have
[00:14:23] an hour to read through
[00:14:24] the commentary and,
[00:14:26] and see the, the pulse of
[00:14:28] the user base, this is a good thread to,
[00:14:32] to get that pulse.
[00:14:33] They, the one guy there
[00:14:35] I think you're right on his message.
[00:14:38] What is his name?
[00:14:40] Yeah, he,
[00:14:41] this message is kind of fun because
[00:14:44] he likened it to
[00:14:46] a relationship.
[00:14:48] And, and it, it kind of,
[00:14:50] it hit a nerve because, you know,
[00:14:53] we've all had those
[00:14:54] kinds of relationships in
[00:14:55] the past where we're like,
[00:14:57] questioning
[00:14:58] should we keep going?
[00:14:59] But I thought he was really good
[00:15:02] in framing that narrative
[00:15:04] to help hopefully help Airtable
[00:15:06] people understand.
[00:15:07] Yeah, we're past the dating stage
[00:15:10] and, and his jab at Tim Cook.
[00:15:14] I, I'm so surprised Scott hasn't.
[00:15:19] Yeah,
[00:15:20] we, we can probably predict
[00:15:22] in fact if you took that
[00:15:23] message and put it into
[00:15:24] Chad GPT and said, what would Scott say?
[00:15:27] You'd probably,
[00:15:28] you'd probably get a really fun
[00:15:30] A I take on that.
[00:15:32] Yeah.
[00:15:32] Yeah.
[00:15:33] Good stuff. All right. A
[00:15:35] couple other announcements
[00:15:37] that they made.
[00:15:39] So now they have
[00:15:42] verified data, which is still
[00:15:44] so basically what happened is there,
[00:15:46] there was a leadership conference
[00:15:48] for Airtable in New York two weeks ago
[00:15:51] and they had to make these announcements
[00:15:54] as part of that, but it sounds like
[00:15:56] these features were not yet ready.
[00:15:58] So
[00:16:00] I don't believe I have access to
[00:16:02] some enterprise accounts
[00:16:03] and these are still
[00:16:04] not available.
[00:16:06] So they had to announce it
[00:16:07] to get it out there,
[00:16:09] but it sounds like developments
[00:16:10] not quite ready for some of these,
[00:16:12] the sidebar did go live.
[00:16:14] I think it's, you have to opt in,
[00:16:15] I'm in an enterprise
[00:16:16] environment and we have
[00:16:17] all of these.
[00:16:19] You do? Ok. So you've gotta,
[00:16:21] you've gotta ask for it. So
[00:16:23] if you scroll down a little bit Dan
[00:16:25] on this page a
[00:16:26] little bit further
[00:16:28] right there and, and italics verified data
[00:16:31] is available on enterprise plans.
[00:16:35] But that is not the case for me either.
[00:16:37] I have an enterprise account and I don't
[00:16:39] see
[00:16:40] anything about it. I wonder where we had
[00:16:42] it.
[00:16:43] We've had it before it was made
[00:16:45] available, you know,
[00:16:47] enterprise wide.
[00:16:48] So I think that's why we have it.
[00:16:50] But all of the features that are talked
[00:16:52] are implemented. So I'm wondering like
[00:16:54] for new enterprise
[00:16:56] customers and old people have to
[00:17:00] opt in, that'd be weird.
[00:17:01] You should say that because
[00:17:02] it
[00:17:02] is
[00:17:03] right. I'm
[00:17:05] looking,
[00:17:06] I know. So the other feature is
[00:17:08] the two way sync
[00:17:09] that that's been in beta for a
[00:17:11] while. I know that one.
[00:17:12] I, I've opted into that one
[00:17:14] with some clients.
[00:17:16] But I haven't seen
[00:17:18] the verified data opt in.
[00:17:20] Me either.
[00:17:21] two-way sync I am excited about, I think,
[00:17:24] and I haven't tested this
[00:17:26] because during the beta, it was not,
[00:17:27] you couldn't edit
[00:17:29] synced fields from an interface,
[00:17:32] but it appears that with the full
[00:17:34] fledged version now you can,
[00:17:36] I haven't tested this. But
[00:17:38] if that's possible, that's super exciting,
[00:17:41] there's still some limitations
[00:17:43] to it like you can't,
[00:17:45] adding records is still
[00:17:47] restricted from the sync
[00:17:49] down base.
[00:17:50] But,
[00:17:52] so far I think it's been pretty,
[00:17:54] it's a market improvement and I don't
[00:17:56] know if they counted as a separate
[00:17:57] feature, but the syncs are now
[00:18:00] more or less instantaneous
[00:18:01] instead of being,
[00:18:03] you know,
[00:18:04] you have to wait a couple
[00:18:05] of minutes. So that's nice.
[00:18:07] Yeah, that's nice too. Absolutely.
[00:18:10] Yeah.
[00:18:12] But yeah, this one,
[00:18:14] this one I definitely wanna
[00:18:15] start playing with.
[00:18:18] Let's see how this plays out.
[00:18:21] Yeah, even on this article,
[00:18:23] I don't see any link to
[00:18:24] opt in.
[00:18:28] So
[00:18:29] anyway, yeah, if you're
[00:18:31] so yeah, these are coming
[00:18:33] at the enterprise level only for now.
[00:18:35] They did say at Daretable
[00:18:37] that eventually they plan to
[00:18:39] make these available or either
[00:18:41] make them available
[00:18:43] at the pro plan or they're,
[00:18:44] they're trying to figure out how to allow
[00:18:47] you to get access to these
[00:18:49] without having to jump up
[00:18:50] to the enterprise or making
[00:18:51] enterprise,
[00:18:53] you know, more accessible.
[00:18:55] So those are some of the things
[00:18:57] that they're working on.
[00:19:00] We have a couple
[00:19:00] of
[00:19:01] comments from,
[00:19:03] sorry, go ahead.
[00:19:05] I was gonna say we have a couple
[00:19:06] of comments from chat
[00:19:08] Kuovonne says isn't sync instant
[00:19:09] sync available for everyone,
[00:19:11] not just enterprise?
[00:19:12] I believe it is.
[00:19:13] Brian Kenney says I feel there's
[00:19:16] a necessary transition for Airtable to
[00:19:18] transform from a new and shiny tool
[00:19:20] to a mature and stable platform.
[00:19:22] No, not clear that Airtable
[00:19:25] leadership agrees with me.
[00:19:27] I think probably all of us agree
[00:19:29] with you, Brian. I think they should
[00:19:31] focus on stability
[00:19:32] if they say that they are.
[00:19:34] But they make certain decisions
[00:19:35] that make it seem
[00:19:37] like they're trying to be
[00:19:38] shiny.
[00:19:40] How would you know
[00:19:42] what Airtable agrees with at all?
[00:19:45] Well, you'd have to, you know,
[00:19:46] when they, they speak,
[00:19:49] you know, if you take the,
[00:19:51] the words that they say at face value,
[00:19:54] then yeah, sometimes if they
[00:19:57] they'll say they're working on things
[00:19:59] like increasing the amount of records that
[00:20:02] you could have in a base.
[00:20:03] I would call that
[00:20:03] working on stability,
[00:20:05] things like instant sync,
[00:20:06] which they have now implemented
[00:20:08] that's working on stability.
[00:20:10] But then there's things like
[00:20:11] you know, apps by Airtable,
[00:20:13] which is not necessarily
[00:20:14] anything to do with
[00:20:15] stability. It's more about adding
[00:20:18] some features or, you know,
[00:20:19] ease and convenience.
[00:20:21] That's a different sort of
[00:20:22] concern, not necessarily a bad concern
[00:20:25] in general, but like
[00:20:27] at,
[00:20:27] at some point,
[00:20:29] you know, the users that are
[00:20:31] already using Airtable
[00:20:32] are gonna be looking more
[00:20:33] for, like,
[00:20:34] I wanna know that my automations
[00:20:36] aren't going to fail. I want to know that
[00:20:38] I can continue to add records
[00:20:40] without having to do a work around.
[00:20:41] And those are there gonna be
[00:20:43] their concerns, not necessarily
[00:20:45] whether or not their
[00:20:46] navigation is horizontal or vertical.
[00:20:50] Yeah,
[00:20:51] I
[00:20:51] get the sense that they're,
[00:20:52] they're making
[00:20:53] this entire enterprise versus
[00:20:56] non enterprise
[00:21:00] division more complicated
[00:21:02] than it needs to be.
[00:21:04] I mean, they're basically saying
[00:21:06] we're gonna divide up the features
[00:21:08] to delineate the difference
[00:21:10] between enterprise and non enterprise.
[00:21:12] Whereas another
[00:21:15] approach is to say
[00:21:16] we're gonna limit
[00:21:18] the capabilities
[00:21:20] and give everyone all features.
[00:21:22] Now before you go crazy on that,
[00:21:24] I do understand that there's a number of
[00:21:28] features that
[00:21:29] no one except enterprise would ever want
[00:21:32] so set, but setting those aside vastly,
[00:21:35] everybody wants every feature.
[00:21:38] So why not give everybody
[00:21:39] every feature
[00:21:40] and simply constrain on a different
[00:21:42] metric.
[00:21:43] One that is on the accommodation
[00:21:45] metric, you're accommodating people to use
[00:21:47] all features
[00:21:49] at certain certain levels.
[00:21:51] That to me, architecturally
[00:21:52] would be a lot simpler
[00:21:54] and you wouldn't have all these
[00:21:55] people just complaining constantly about,
[00:21:57] oh, well,
[00:21:58] you know, I gotta, I gotta pay three grand
[00:22:01] just to get this one little feature
[00:22:03] I need and that's just the silliness.
[00:22:05] Yeah.
[00:22:06] Yeah, I would, I would
[00:22:07] agree that
[00:22:08] I think there's up to your point Bill,
[00:22:10] there's only like a handful of
[00:22:11] features that only enterprise
[00:22:12] like single sign on like
[00:22:14] a bakery shop that runs
[00:22:15] their business off of Airtable
[00:22:17] is probably not going
[00:22:18] to need single sign on.
[00:22:20] That's just my opinion
[00:22:21] and even like
[00:22:22] data sets and you're probably
[00:22:24] not gonna have so many
[00:22:25] bases that you're going to
[00:22:26] need a verified data set
[00:22:28] whereas the enterprise customer might,
[00:22:31] but,
[00:22:32] you know, there are some single
[00:22:34] operator businesses
[00:22:35] that need just a whole bunch
[00:22:36] of records or a whole bunch
[00:22:38] of automation runs. And
[00:22:39] it, it's weird that they have
[00:22:41] to be in the same class as like Disney
[00:22:44] to get
[00:22:45] that feature.
[00:22:47] Correct. Correct.
[00:22:48] Yeah, that, I mean, that's my point.
[00:22:50] They are making life harder
[00:22:52] for themselves because under the covers,
[00:22:54] the code required to do all of this,
[00:22:56] the way they're doing it
[00:22:58] is massively complex
[00:23:00] and that is probably breeding
[00:23:02] a new breeding ground
[00:23:04] for these kinds of missteps.
[00:23:05] The likes of which we just saw.
[00:23:08] So
[00:23:09] I'm puzzled.
[00:23:10] I, I have no idea why they would
[00:23:12] use draw the lines through
[00:23:14] their product the way they did
[00:23:17] in order to segregate the two worlds
[00:23:19] of enterprise and non enterprise.
[00:23:21] Yeah,
[00:23:22] I do see
[00:23:23] speaking of verified data,
[00:23:25] I do see that as well
[00:23:27] in my enterprise instances.
[00:23:29] You do?
[00:23:30] Nice.
[00:23:32] The sections there, it's empty.
[00:23:34] You just have to create a data set
[00:23:36] first and then you'll see them.
[00:23:39] Got you.
[00:23:40] Ok. So it's rolling out.
[00:23:42] So, yeah, so lots of new
[00:23:44] and lots of new products,
[00:23:45] especially if you're at
[00:23:46] the enterprise
[00:23:48] play with, we'll start to
[00:23:50] showcase those as they roll out.
[00:23:52] We'll do some,
[00:23:53] some feature segments
[00:23:55] on those in future shows.
[00:23:57] If we can, it's tougher to show
[00:24:00] enterprise that's another downfall of
[00:24:02] putting it behind enterprise.
[00:24:04] Those aren't as easy to show here.
[00:24:08] A couple of comments before we move on.
[00:24:10] Brian says on the other side of
[00:24:12] accommodation, I look at something like
[00:24:14] Jira which is very configurable and
[00:24:15] complain that it is too complicated.
[00:24:20] Jen says it's the empty box
[00:24:22] that is both enticing and overwhelming.
[00:24:24] So they try to put together the
[00:24:26] Legos in some order
[00:24:27] to get people to jump in.
[00:24:28] But the need to evolve.
[00:24:33] Imagine you went into a movie theater
[00:24:35] and you bought a ticket
[00:24:37] and then you decided to buy popcorn
[00:24:39] and maybe a Coke and of course,
[00:24:41] you paid exorbitant prices.
[00:24:42] We won't get into
[00:24:43] the economics behind that.
[00:24:45] But then
[00:24:46] there was another ticket
[00:24:47] that allowed you to
[00:24:49] have hot dogs, right?
[00:24:51] No one would tolerate that because
[00:24:54] everybody wants different things.
[00:24:56] You want the full menu
[00:24:58] and you're willing to pay more.
[00:25:01] But you'll do it
[00:25:03] you know, for that accommodating level.
[00:25:05] It just seems odd that they
[00:25:07] would slice it up that way anyway.
[00:25:08] That's
[00:25:09] I'm puzzled. Maybe somebody
[00:25:11] can tell me why I shouldn't be.
[00:25:14] Yeah, lots of decision makers
[00:25:18] on that
[00:25:19] often. So,
[00:25:21] yeah.
[00:25:21] Ok, let's, let's move on.
[00:25:24] Got a lot to talk about before we do.
[00:25:27] Quick shout out to On2Air.
[00:25:29] Our primary sponsor.
[00:25:30] It's an all one tool kit
[00:25:32] to run your business on Airtable.
[00:25:33] It's a suite of apps
[00:25:34] that help a business do more
[00:25:36] with their Airtable
[00:25:37] environment.
[00:25:38] So if you are running your business
[00:25:40] on Airtable, you need to check out
[00:25:42] on2air.com
[00:25:43] for today's highlights,
[00:25:45] just a quick shout out On2Air Forms.
[00:25:49] One of the features that
[00:25:51] that is available in On2Air Forms
[00:25:53] is the ability to have multi
[00:25:54] tiered selections.
[00:25:56] So if you think of a car
[00:25:58] trying to choose a car
[00:26:01] you start with your make.
[00:26:03] So this is a linked record
[00:26:05] that shows you all of the selections
[00:26:08] in the makes.
[00:26:09] So there's a table of car manufacturers.
[00:26:14] So if you pick the one that you want,
[00:26:16] you can then go down to the model
[00:26:18] and now you can select from your model and
[00:26:20] this is only going to show you
[00:26:22] models from the Ford selection.
[00:26:25] So this automatically filters
[00:26:28] what models you can select
[00:26:30] based off of that parent selection.
[00:26:32] And then once you select that,
[00:26:35] then this third one
[00:26:36] will only show you cars available
[00:26:39] that are with that model.
[00:26:41] So this one does not have any
[00:26:44] find one that does.
[00:26:49] I believe a Mustang.
[00:26:51] I think there's Mustangs.
[00:26:56] So this checks the inventory
[00:26:58] and it finds an actual car
[00:26:59] that's a Mustang.
[00:27:00] So now we have three levels of
[00:27:03] parent child relationships
[00:27:05] and filtering your
[00:27:06] selection based off of those. So
[00:27:09] nice feature that's available
[00:27:11] in On2Air forms,
[00:27:12] check it out at on2air.com
[00:27:18] all Right.
[00:27:19] OK. We're gonna do some
[00:27:22] background on Lorenzo and Bill.
[00:27:26] Let's start with Lorenzo.
[00:27:28] I know you were on the show
[00:27:29] a year ago, I believe.
[00:27:31] So, that one, we did a full hour
[00:27:34] talking with Lorenzo
[00:27:35] and had some cool stuff
[00:27:37] to share but
[00:27:38] give us an update,
[00:27:39] give us some background on you
[00:27:41] and, and what you're up to.
[00:27:43] Sure.
[00:27:44] It was a while ago but
[00:27:46] Lorenzo founder at Lotic Digital
[00:27:49] we are
[00:27:51] a development agency
[00:27:53] that focuses on Airtable integrations
[00:27:57] and
[00:27:58] well Airtable
[00:27:59] and make integrations mainly.
[00:28:01] So mainly we deal with customers
[00:28:03] who have
[00:28:04] large scale Airtable
[00:28:06] integrations and they need to
[00:28:08] connect
[00:28:09] their Airtable to
[00:28:11] any number of platforms,
[00:28:14] whether it's quickbooks,
[00:28:15] dropbox, docusign,
[00:28:18] Google, Google stack,
[00:28:19] Microsoft stack.
[00:28:20] So we kind of focus on the integrations
[00:28:22] between Airtable
[00:28:23] and third party platforms.
[00:28:26] More recently with the rise of A I,
[00:28:29] we've, we have done a lot in terms of
[00:28:32] integrating
[00:28:33] Airtable bases to
[00:28:36] technologies like
[00:28:38] image generation and technologies
[00:28:41] like open A I. So lots of stuff around
[00:28:44] chat GPT and whether it's
[00:28:47] generating content or editing content or
[00:28:51] mining databases for semantic search.
[00:28:53] We've done a lot in that domain.
[00:28:55] So
[00:28:56] heavy focus on that. But in general
[00:28:59] in our domain, we get a lot of,
[00:29:02] a lot of that. So companies will come and
[00:29:04] they'll say, hey, we wanna
[00:29:06] integrate our existing solution with this
[00:29:08] next gen platform or stuff
[00:29:10] and sometimes it's not Airtable
[00:29:12] customers as well.
[00:29:13] So
[00:29:14] we have a lot of customers that use make.
[00:29:17] And maybe another third party
[00:29:19] system that they use. And
[00:29:20] in some cases, we will
[00:29:22] integrate Airtable
[00:29:25] as a as a database to
[00:29:27] kind of help the integration
[00:29:29] and sometimes it's just
[00:29:30] no Airtable and just integration between
[00:29:33] point to point solutions. So
[00:29:35] that is a
[00:29:37] quick intro on me and kind of
[00:29:39] what our company's been up to lately.
[00:29:41] Awesome.
[00:29:42] You're based in Minneapolis?
[00:29:45] Minneapolis, Minneapolis.
[00:29:47] We have some team members, most of our,
[00:29:49] our team is in the US. We do have some,
[00:29:52] some team members over in Europe as well.
[00:29:56] Great. So they can find you
[00:29:58] loticdigital.com
[00:30:00] loticdigital.com
[00:30:01] all right. Awesome. Thank you, Lorenzo.
[00:30:04] And you presented last week
[00:30:05] at Daretable on a lot
[00:30:07] of what you just talked
[00:30:07] about your A I integrations with Make
[00:30:11] I did an A I solution. So maybe we'll
[00:30:14] get into that later. We'll see.
[00:30:16] Yep.
[00:30:17] And I'll give you a shout out.
[00:30:18] You didn't mention this, but
[00:30:20] you were, you were selected
[00:30:22] as the O E M partner
[00:30:24] of the year with Make, I believe?
[00:30:27] Yes,
[00:30:29] yeah, we won two awards with make
[00:30:33] this last year and two years ago.
[00:30:36] Part I think O E M and Enterprise.
[00:30:39] So we do a lot of
[00:30:41] O E M
[00:30:42] support and integration work with make.
[00:30:45] Awesome,
[00:30:46] well done. Very cool.
[00:30:48] All right,
[00:30:49] let's move back to.
[00:30:53] OK, Bill.
[00:30:54] You've been on the show at least
[00:30:57] twice that I remember. Maybe more.
[00:30:59] So good to have you back.
[00:31:01] Give us, give us your
[00:31:03] quick story.
[00:31:06] Back story is very short.
[00:31:08] I'd rather leave the time for
[00:31:10] more important things. But,
[00:31:13] you know, I
[00:31:15] started in the eighties in computing.
[00:31:18] I guess the best way to describe
[00:31:20] what happened between then and now
[00:31:22] is I'm semi
[00:31:23] retired.
[00:31:25] I'm an old guy in the desert enjoying
[00:31:28] being an old guy because
[00:31:30] I don't have to filter anything.
[00:31:32] So my job, I see my job really
[00:31:35] now is stirring, stirring the mud.
[00:31:40] And, and then trying to reflect upon,
[00:31:41] you know, where we came from.
[00:31:43] So,
[00:31:43] you know, in the eighties,
[00:31:45] I worked with NASA and developed a
[00:31:47] kind of a cool little
[00:31:49] technology that moved data
[00:31:50] from space shuttle computers to desktops.
[00:31:54] Desktops were new.
[00:31:55] Nobody understood them, but
[00:31:57] even I B M couldn't get them
[00:31:59] to go faster than 19,200
[00:32:01] bits per second,
[00:32:03] which by the way is a crawl.
[00:32:04] We figured out how to
[00:32:06] do that six times that.
[00:32:08] And because of that,
[00:32:09] we ended up with a contract
[00:32:11] with NASA. Now that
[00:32:11] technology eventually became lap link,
[00:32:14] which is a crazy idea,
[00:32:16] laptops eventually needed
[00:32:18] to communicate with desktops.
[00:32:20] And so we made a killing
[00:32:22] in that market from 84
[00:32:23] till roughly 89.
[00:32:26] And so
[00:32:27] I kind of have a strong
[00:32:29] sense of where we've been,
[00:32:31] where the industry has
[00:32:32] kind of made mistakes
[00:32:34] where they've had successes
[00:32:36] and that kind of shines in
[00:32:37] some of the commentary that I
[00:32:39] cast about. So,
[00:32:41] you know, what do I do now?
[00:32:43] I'm trying to help the
[00:32:44] guys at Cyber Lander,
[00:32:45] make an invisible camper
[00:32:47] for the future of electrification.
[00:32:49] And I'm dabbling in
[00:32:52] A I, I spend a lot of time in A I
[00:32:55] people compel me to do stuff
[00:32:58] at a consulting level from time to time.
[00:33:00] But I'm very cautious
[00:33:01] about taking projects because,
[00:33:04] you know, I, I'm a specialist
[00:33:06] through and through.
[00:33:07] I don't build Airtable bases. In fact,
[00:33:10] most of the people on this call
[00:33:12] and everybody in the
[00:33:13] community can run circles
[00:33:14] around me
[00:33:14] when it comes to Airtable questions.
[00:33:17] But at the end of the day,
[00:33:18] there's certain fringe of those
[00:33:20] that need, you know, really deep
[00:33:23] consideration when it
[00:33:24] comes to real time data,
[00:33:26] very secure data find,
[00:33:28] I specialize in findability
[00:33:30] and I spend a lot of time thinking about,
[00:33:33] you know, some of the stupid
[00:33:34] things we do with databases
[00:33:36] and how to avoid them, how to simplify.
[00:33:39] So I'm always looking for parts
[00:33:40] that need to be ejected,
[00:33:41] need to be eliminated from a process.
[00:33:44] So that simplicity reins and
[00:33:46] systems are much more,
[00:33:49] reliant.
[00:33:51] Right. I mean, that, that's the key
[00:33:53] most of us want to go to the beach.
[00:33:55] We're always hacking on our
[00:33:56] keyboards at nine PM on Friday.
[00:33:59] And the reason we are is
[00:34:00] because we're not thinking ahead
[00:34:02] to eliminating parts, eliminating steps,
[00:34:05] simplifying work flows
[00:34:06] and automation and things like that.
[00:34:08] So I spent a lot of time
[00:34:10] thinking about that
[00:34:11] and I love to look for irony.
[00:34:12] There's plenty of it at Airtable,
[00:34:14] of course.
[00:34:15] So
[00:34:16] that,
[00:34:16] that's me in a nutshell.
[00:34:18] You know,
[00:34:19] if you're looking for an unvarnished
[00:34:22] kind of truth about the world,
[00:34:24] I'll give you one.
[00:34:25] It's not always accurate.
[00:34:27] I try to get it accurate,
[00:34:28] but it's always in, in kind of a flux too,
[00:34:31] you know. So
[00:34:33] you, you take, you take what you can
[00:34:35] from my, between my ears
[00:34:38] and some, some of it applies.
[00:34:40] Probably batting 3 22.
[00:34:42] I don't know
[00:34:44] which means I'm failing.
[00:34:47] What is it? 6 88. Right. Right.
[00:34:51] Yeah. I mean, if anybody who's
[00:34:53] been spent any time on
[00:34:54] any of the communities
[00:34:56] likely knows your name,
[00:34:57] knows your I think you've
[00:34:59] been using chat GP T for
[00:35:01] several years to write your essays.
[00:35:03] I don't know how you spit
[00:35:05] out so much content in
[00:35:07] the short amount of time that,
[00:35:09] that we have in the day. But
[00:35:10] you have produced volumes
[00:35:13] and volumes of insight
[00:35:15] and somebody asked me the other day
[00:35:17] they said, are you using
[00:35:18] A I to generate all
[00:35:19] this crap?
[00:35:20] And of course they were
[00:35:22] affectionately calling it crap.
[00:35:24] But
[00:35:24] at the end of the day,
[00:35:26] the answer is no, everything I write,
[00:35:27] I write myself.
[00:35:30] I've just kind of been that way.
[00:35:32] I'm an old guy,
[00:35:32] so I'm probably gonna do that,
[00:35:34] you know, until I die. But
[00:35:36] I'm kind of reluctant to allow a I
[00:35:38] to write things for me unless it is
[00:35:40] to help me understand something
[00:35:42] so that I can write it better.
[00:35:44] But that's kind of, you know,
[00:35:47] a personal thing.
[00:35:48] I think there's a lot of people
[00:35:50] that are doing really well
[00:35:51] leveraging their time
[00:35:52] even better than I do.
[00:35:54] Yeah, fair enough.
[00:35:56] Awesome. Thank you, Bill.
[00:35:58] Good to have you back on.
[00:36:00] Let's
[00:36:02] move on.
[00:36:04] Ok. So now for the remainder of the time
[00:36:07] we're gonna dive into
[00:36:09] the future of A I.
[00:36:11] Man, has it changed in the last year.
[00:36:14] So why don't we start with,
[00:36:16] with the Airtable?
[00:36:19] So Airtable is part of their announcements
[00:36:22] launched a new landing page
[00:36:25] and have some screenshots around
[00:36:28] how they're going to implement A I
[00:36:30] at least initially.
[00:36:32] Anybody wanna give a summary
[00:36:34] of what Airtable is planning to do?
[00:36:40] Lorenzo, you go first,
[00:36:43] you kind of know this better than I do.
[00:36:46] Well, I think it's right on there
[00:36:49] on their website. I kind of
[00:36:51] dabbled in
[00:36:54] viewing
[00:36:55] the features and it looks like
[00:36:57] at first glance, you're able to generate,
[00:37:01] generate content.
[00:37:02] So one of them is like pictures
[00:37:04] so it can look at pictures
[00:37:06] in your database and and generate
[00:37:08] descriptions.
[00:37:11] Now it will be interesting. So
[00:37:15] when you come from
[00:37:17] developing these solutions,
[00:37:19] you have all this extra metadata
[00:37:21] that you can apply
[00:37:22] to it, like which model it's using,
[00:37:24] what the tone is,
[00:37:27] how accurate it is,
[00:37:28] how creative it is.
[00:37:29] So it'll be interesting to see with
[00:37:30] these tools, how much they
[00:37:32] kind of black box that and what
[00:37:34] they give you as far as like
[00:37:36] what you're able to
[00:37:37] do with it because
[00:37:38] that's kind of how people are using A I
[00:37:40] now as we're using all that,
[00:37:42] right? So when we're generating content,
[00:37:44] we're saying in advance,
[00:37:45] hey, answer this question
[00:37:47] in context of this or
[00:37:49] I'm gonna feed you a description
[00:37:50] of my business.
[00:37:51] And then now that I've done
[00:37:52] that
[00:37:53] and you know that the furniture is
[00:37:56] for
[00:37:57] Children and it's,
[00:37:59] you know, supposed to use
[00:38:01] whatever from the environment
[00:38:03] and it's, it's
[00:38:04] ecofriendly or whatever.
[00:38:06] Now write my descriptions.
[00:38:08] I don't know how much of
[00:38:09] that's gonna be in here.
[00:38:10] So how generic it is. But maybe
[00:38:13] if you are using a lot of A I right now,
[00:38:16] this is probably less exciting
[00:38:18] than if you've never touched it,
[00:38:19] then maybe it'll be like, wow, I can
[00:38:22] generate,
[00:38:23] you know,
[00:38:24] descriptions from images and stuff. So
[00:38:26] it's kind of,
[00:38:28] it'll be interesting to see how
[00:38:30] underwhelming or overwhelming it is for
[00:38:33] their user base. So
[00:38:35] I just, I just see a lot of like
[00:38:37] image generation product descriptions.
[00:38:39] There is one where it's like
[00:38:41] it wasn't really clear if you scroll down.
[00:38:44] I think there's like a section where
[00:38:45] you can
[00:38:46] generate like the A I field.
[00:38:49] Yeah.
[00:38:50] Yeah, that was my understanding
[00:38:52] is there's gonna be like
[00:38:54] a new field type that's
[00:38:55] like an A I field type that will like auto
[00:38:57] generate with content.
[00:38:59] I don't think they,
[00:39:00] I don't think they're announcing images.
[00:39:03] I think like these images
[00:39:05] is, I think
[00:39:07] is just,
[00:39:10] is for product description
[00:39:13] is what they said.
[00:39:14] But one thing I would say is like,
[00:39:17] don't take these, these screenshots as,
[00:39:19] as what it's gonna look like.
[00:39:22] Like, I don't think they're
[00:39:24] coming out with a slider field.
[00:39:26] I think, you know, they, they
[00:39:27] take some liberty with these marketing
[00:39:30] gifs of what it could potentially do,
[00:39:32] but I don't think this is what it's
[00:39:35] gonna look like.
[00:39:36] Ironically,
[00:39:38] these marketing gifs
[00:39:40] look like something that
[00:39:41] chat GP T would come up with.
[00:39:45] That's the irony here, right?
[00:39:47] There's a lot of creative license
[00:39:49] in these and we've seen it time and time
[00:39:52] again
[00:39:52] and so it's gonna be bizarre, I think,
[00:39:56] well, it looks like it's the configuration
[00:39:59] panel before you get to a field.
[00:40:00] So I think it's
[00:40:02] more
[00:40:03] like a dialogue that pops up
[00:40:05] and they've had,
[00:40:07] you know, there's certain things
[00:40:08] that exist in current
[00:40:10] dialogues that aren't
[00:40:10] like field types you'll only see
[00:40:12] them in, in the sort of pop up.
[00:40:14] So, like,
[00:40:14] it wouldn't surprise me if there's a,
[00:40:17] a slider. I just don't think it's,
[00:40:19] you know, like
[00:40:20] the one that goes from general
[00:40:21] to expert,
[00:40:22] I don't know if that needs to be a
[00:40:23] slider. The first one that was like
[00:40:26] the number of words that makes sense.
[00:40:27] This one doesn't make sense,
[00:40:29] general to expert, make it.
[00:40:32] And I have kind of a problem
[00:40:34] with number of words.
[00:40:36] I don't think users
[00:40:37] give a crap
[00:40:39] about the number of words
[00:40:40] nor nor do they even
[00:40:41] think.
[00:40:42] But sometimes you do.
[00:40:43] But if you, if you have a listing,
[00:40:45] for instance,
[00:40:46] on a certain like platform,
[00:40:48] you're limited, you have a cap.
[00:40:50] So
[00:40:50] I think it, in some instances
[00:40:52] it might be good to say,
[00:40:54] do not go over 500 words
[00:40:55] or something like that.
[00:40:56] The,
[00:40:57] the,
[00:40:57] the issue here though is I think that,
[00:41:01] that they're trying to give you
[00:41:03] a means to control A I costs
[00:41:06] and that's very different than
[00:41:08] the output. So you may want
[00:41:10] the A I process to factor in
[00:41:13] many more words than you actually produce.
[00:41:16] And that is not apparent
[00:41:18] in their dialogue eventually,
[00:41:20] it will be apparent
[00:41:20] because
[00:41:21] it's a requirement.
[00:41:24] There are cases where
[00:41:25] you only want to generate five keywords,
[00:41:28] comma separated. Nothing more.
[00:41:30] No other information.
[00:41:32] How do you configure that?
[00:41:33] And by the way, the A I field, I think
[00:41:35] that they're referring to
[00:41:37] in here is exactly what Firebase
[00:41:39] launched a few months ago.
[00:41:41] So Firebase has these new fields
[00:41:44] that you can declare as A I fields.
[00:41:47] They're not different in any other
[00:41:49] regard except you're declaring
[00:41:51] any events occurring to that field
[00:41:54] is now going to do something
[00:41:56] to another field or at the same field.
[00:41:59] And I think that's kind of
[00:42:00] where they're hinting here.
[00:42:01] I've always been a,
[00:42:03] you know, an advocate
[00:42:04] of no more data types.
[00:42:06] You know, they're, they're making
[00:42:08] up data types
[00:42:09] just as the button was a made up
[00:42:10] data type.
[00:42:11] A formula field was a made up data type.
[00:42:14] Those are fields that really shouldn't
[00:42:16] be fields, they're attributes of fields,
[00:42:18] not fields themselves.
[00:42:20] And I think the A I
[00:42:23] field is also another misstep
[00:42:25] architecturally in, in the long run.
[00:42:29] Yeah. Interesting.
[00:42:31] Yeah, we'll see.
[00:42:32] I think it's still, yeah, early.
[00:42:35] Because this also implies that like
[00:42:37] the type of A I fields,
[00:42:39] like if they really
[00:42:39] implement it this way,
[00:42:41] like it's gonna be predefined,
[00:42:43] like the parameters
[00:42:45] that would go into it is
[00:42:46] based off of
[00:42:47] similar to like their dropdown
[00:42:49] of naming, it's gonna be a predefined list
[00:42:52] instead of customizable.
[00:42:54] So I hope they don't
[00:42:55] do that because that will,
[00:42:58] that, that's exactly why
[00:43:00] it should not be a field.
[00:43:01] Fields should react to A I not be A I,
[00:43:06] that's my opinion on that.
[00:43:08] It's a slippery slope
[00:43:09] and they're stepping right on
[00:43:11] it. Isn't that almost
[00:43:12] the same thing though?
[00:43:13] So if I have a new field
[00:43:15] and kind of like a look up field,
[00:43:17] I create an A I field.
[00:43:19] And so I'm selecting the field
[00:43:20] I want in context of, right?
[00:43:22] So there's one
[00:43:23] that's like
[00:43:24] something about tweet,
[00:43:25] like generate a tweet.
[00:43:27] So you're gonna have to select the field,
[00:43:29] kind of like a look up.
[00:43:30] I want to generate a
[00:43:31] tweet. I don't think it's gonna
[00:43:33] be specific like sales rep.
[00:43:35] I think you're gonna
[00:43:35] select the field
[00:43:37] that you want to generate content
[00:43:39] and then you're gonna set
[00:43:41] these parameters and
[00:43:41] then this new field is gonna display it
[00:43:44] kind of like,
[00:43:45] let,
[00:43:45] let's hope. I, I think,
[00:43:47] I think that's an appropriate way to go.
[00:43:49] It's
[00:43:50] the context of something,
[00:43:52] it's, it's like a look up.
[00:43:54] Yeah.
[00:43:55] Unless like if you go off of this,
[00:43:57] this looks like it's like
[00:43:59] predefined what the
[00:44:00] parameters are
[00:44:02] once you select.
[00:44:04] Yeah.
[00:44:04] Yeah.
[00:44:06] It, it could get ugly
[00:44:07] if they go this route.
[00:44:08] I just hope it doesn't slow down the bases
[00:44:13] if you think about like a base with like,
[00:44:15] you know,
[00:44:15] 8000 records and then you create
[00:44:17] this new A I field, right?
[00:44:19] And it's got to go
[00:44:20] and like generate,
[00:44:21] you know, all this content.
[00:44:24] Well, yeah, that's a good point.
[00:44:27] Like what triggers in
[00:44:29] A I field
[00:44:31] that look ups are easy.
[00:44:32] Roll ups are easy.
[00:44:33] It's whenever the original data
[00:44:35] changes, it Recalculates what the
[00:44:37] new sum total is or whatever.
[00:44:39] Formulas are easy. Whenever
[00:44:41] the input changes it Recalculates.
[00:44:43] So we have to point
[00:44:45] the A I field at something,
[00:44:47] right? It has to generate
[00:44:49] whatever it is informal voice with the,
[00:44:51] you know, level of expertise
[00:44:53] or whatever at something.
[00:44:55] Does it regenerate every time
[00:44:57] that something changes? Or is it,
[00:45:00] you know, you trigger it manually?
[00:45:03] That it doesn't seem clear
[00:45:05] from these screenshots when it
[00:45:08] pulls
[00:45:08] A I
[00:45:09] to do stuff.
[00:45:10] We've all had the experience of
[00:45:12] conversing with support
[00:45:15] and the support call ends with,
[00:45:18] well, you have too many formula fields
[00:45:21] and I predict in 2024 those conversations
[00:45:24] will end with while you have too
[00:45:27] many A I fields.
[00:45:30] It kind
[00:45:30] of
[00:45:30] just a matter of time.
[00:45:32] And so
[00:45:33] yeah, it's, it's a bad idea,
[00:45:35] you know, to just say no, this is a broad
[00:45:37] kind of experience and
[00:45:39] hopefully they'll, you know, they'll,
[00:45:41] they'll get the message soon.
[00:45:44] Yeah.
[00:45:44] It, it could get really crazy.
[00:45:46] Because every one of these
[00:45:48] has a cost, right?
[00:45:50] And if they're using
[00:45:52] you know, open A I,
[00:45:53] it has a bigger cost than
[00:45:55] say, with Google.
[00:45:58] Yeah, let's
[00:45:59] let's transition.
[00:46:00] So, yeah, lots of unknowns
[00:46:02] with Airtable in the direction
[00:46:04] they actually go in.
[00:46:05] So we'll see, they're saying summer
[00:46:08] is when we'll, we'll see something
[00:46:10] along these lines.
[00:46:11] There's a sign up
[00:46:12] to get on the waiting list.
[00:46:14] So I'm on the waiting list.
[00:46:15] We'll see when that,
[00:46:16] any news comes from that.
[00:46:18] But, let's transition,
[00:46:19] let's talk bigger picture.
[00:46:21] A I in general, trends.
[00:46:23] What the future is.
[00:46:26] Let's open up for discussion If,
[00:46:28] if you've got something to share
[00:46:30] Bill, why don't we start with you?
[00:46:33] If you've got
[00:46:35] what do you think?
[00:46:36] Give us, give us your high level
[00:46:37] view of where we're at and
[00:46:39] what, what to expect in the next
[00:46:41] 2 to 3 years.
[00:46:43] Yeah, I mean, I, I think
[00:46:46] A I in the world of data,
[00:46:49] I'm just
[00:46:49] speaking broadly, of course,
[00:46:52] A I in the world of data
[00:46:54] is going to lean very quickly
[00:46:58] you know, into
[00:47:01] hopefully I can share here.
[00:47:03] Am I sharing?
[00:47:05] Now you are, yeah.
[00:47:06] Yeah. So, you know,
[00:47:07] when we think of A I in the future,
[00:47:09] I, my feeling is that
[00:47:10] automation is going to play a big role.
[00:47:12] And so you guys
[00:47:13] have probably all seen the,
[00:47:15] you know, auto GP T it's,
[00:47:18] it's really getting a lot of traction.
[00:47:20] People seem to love it
[00:47:22] and I can understand why.
[00:47:25] I think it's over hyped.
[00:47:27] You know, their valuation at a
[00:47:29] billion 1.7 billion or whatever it is,
[00:47:31] is just amazing to me because
[00:47:34] they actually have a lot of flaws
[00:47:36] in their design and their architecture.
[00:47:38] But
[00:47:38] what they do have right is the idea that
[00:47:41] you could,
[00:47:43] you could set a goal
[00:47:46] and an initial task
[00:47:48] and the,
[00:47:49] and the system auto GP T
[00:47:51] will automatically go and execute
[00:47:54] toward meeting that objective.
[00:47:57] And so how does it do that? Well,
[00:47:59] it does it by simply
[00:48:02] using itself to,
[00:48:05] you know, if it's like
[00:48:06] the snake eating its own tail,
[00:48:07] it starts off by saying
[00:48:09] what's your objective?
[00:48:11] And what's the initial task?
[00:48:13] And from that initial task,
[00:48:14] are there other tasks
[00:48:15] that we could follow in order to reach
[00:48:18] the main objective?
[00:48:19] That involves a lot of,
[00:48:21] you know, kind of complicated
[00:48:23] code under the covers.
[00:48:25] But I built one in one day
[00:48:26] in Airtable.
[00:48:28] And
[00:48:29] and I actually this is kind of a
[00:48:31] hopefully you can see this,
[00:48:33] maybe I can make it larger,
[00:48:35] I guess I can't.
[00:48:36] Yeah.
[00:48:37] Well, if you, yeah,
[00:48:38] it'd be great if you could do that. But
[00:48:40] oh yay.
[00:48:41] That
[00:48:42] OK. Anyway,
[00:48:44] so and I'm gonna move
[00:48:46] the script aside here because
[00:48:47] really all we wanna do
[00:48:49] is focus on the objective.
[00:48:51] And, and in this case,
[00:48:52] I, I said something very broad and I said,
[00:48:54] OK,
[00:48:55] you know,
[00:48:57] auto A I, I, I call this Auto
[00:48:59] palm
[00:49:00] because it uses the Google
[00:49:03] L L MS not the GP T L L Ms
[00:49:05] or open A I.
[00:49:06] And I, I said, let's give it two things.
[00:49:09] We want to become an expert
[00:49:11] in Airtable
[00:49:11] and we're gonna learn all about Airtable
[00:49:13] in the process of becoming
[00:49:15] that expert. And so
[00:49:16] we had defined that as the initial task.
[00:49:19] Now,
[00:49:20] if you thought about trying
[00:49:21] to become an expert in Airtable,
[00:49:23] what task would make sense?
[00:49:25] And so under the covers,
[00:49:27] this system is designed to go and do that
[00:49:30] and rather than take a bunch
[00:49:32] of time to show you how it all runs,
[00:49:35] this is the output of that.
[00:49:38] And so it has created,
[00:49:40] what do we got here? 13 pages
[00:49:42] PDF
[00:49:44] how to become an Airtable expert.
[00:49:48] And it goes through and it divides this
[00:49:50] into all the logical tasks that wouldn't
[00:49:52] be necessary for you to pull that off.
[00:49:54] And it does a fairly good job
[00:49:57] of writing the pros,
[00:49:58] it does a fairly good job of
[00:50:00] understanding the various
[00:50:02] elements of Airtable
[00:50:03] and it does a fairly good job
[00:50:05] of learning how to use Airtable formulas.
[00:50:09] It fails miserably at this list.
[00:50:12] I'm not gonna get into
[00:50:14] we can go through here
[00:50:16] and cherry pick for hours on
[00:50:18] end, but that's not the point.
[00:50:19] The point is I have a template
[00:50:21] for learning how to become
[00:50:22] an expert in Airtable
[00:50:24] and I got it in one minute
[00:50:26] and I got it by doing only one thing,
[00:50:30] identify an objective
[00:50:32] and then
[00:50:34] give it an initial task.
[00:50:36] Now, these two sentences are critical,
[00:50:39] right? Because you're basically
[00:50:41] setting the tone for the A I
[00:50:43] automatron,
[00:50:43] if you will,
[00:50:45] to go off and do
[00:50:46] the job.
[00:50:47] And it will literally fire off
[00:50:50] based on this. And it will run
[00:50:52] dozens, sometimes hundreds
[00:50:55] of inquiries into the L L M
[00:50:58] to get the right text.
[00:51:00] Now, one of the challenges of course is
[00:51:03] you don't want that process to stray.
[00:51:06] And so how do I avoid
[00:51:07] letting it stray?
[00:51:09] And the answer is embeddings.
[00:51:11] So every time this system
[00:51:14] generates
[00:51:15] a particular result to a task
[00:51:19] I take and then reevaluate
[00:51:22] that in information,
[00:51:24] I establish, I break it into parts
[00:51:26] and then I set embeddings for each of the
[00:51:28] paragraphs
[00:51:29] and then I test the embeddings to see
[00:51:32] if they're on the compass
[00:51:33] heading of the goal.
[00:51:35] So that's yet another layer
[00:51:37] of iteration that's going on
[00:51:38] behind the scenes.
[00:51:39] Can you put more context
[00:51:41] around embeddings?
[00:51:42] What do you mean by that?
[00:51:43] Yeah. So an embedding is a way
[00:51:45] to take like this paragraph right here
[00:51:48] and send it over to
[00:51:50] in this case palm
[00:51:53] and say
[00:51:54] give me the vector for that clump of text
[00:51:59] in the L L M
[00:52:00] and then you take the goal
[00:52:03] and the initial task, you could kainate
[00:52:05] those two and do the same.
[00:52:07] You're saying give me the vector
[00:52:09] for that particular clump
[00:52:11] of text inside the L L M,
[00:52:13] and what that comes back with is an array.
[00:52:16] The vector is an array.
[00:52:17] It's a very large array. It's about 1500
[00:52:21] you know, array values
[00:52:23] and those two values
[00:52:25] can then be compared mathematically using
[00:52:28] cosine
[00:52:28] or, and some kind of
[00:52:30] you know, similarity differentiator
[00:52:33] such as dot product,
[00:52:34] you can then compare those
[00:52:36] and then try to understand
[00:52:37] how close are those two concepts?
[00:52:40] And if they're, are they far apart
[00:52:43] or are they near?
[00:52:44] Are they, you know,
[00:52:45] is it on the compass heading that I hope
[00:52:47] to achieve with my goal?
[00:52:49] And if it's not on the compass heading,
[00:52:52] repeat the process
[00:52:54] and continue to
[00:52:56] improve the process
[00:52:58] based on the fact that I want to get that
[00:53:01] cosine
[00:53:01] similarity to close,
[00:53:03] you know, the gap to close,
[00:53:05] you want them to be very
[00:53:06] similar in terms of the compass heading
[00:53:09] of the goal versus the compass heading
[00:53:11] of the text that
[00:53:12] the A I is generating.
[00:53:13] So it largely gets reduced to kind of
[00:53:17] a mathematical dance, if you will.
[00:53:19] It's an orchestration of math
[00:53:21] using embeddings which have
[00:53:22] nothing to do with
[00:53:24] going into chat GP T and typing
[00:53:25] something
[00:53:25] and getting an answer back.
[00:53:27] Although under the covers,
[00:53:29] it's using word vectors
[00:53:30] to try to understand what it should say,
[00:53:32] given what you tell it.
[00:53:34] This goes another level
[00:53:36] and I think this is where
[00:53:37] automation
[00:53:39] is gonna take A I for the benefit
[00:53:43] of, of customers.
[00:53:45] And so I,
[00:53:46] that's why I was really disappointed
[00:53:48] when I saw the Airtable announcement.
[00:53:50] They're not doing stuff like this.
[00:53:52] They're doing something that's very
[00:53:55] superficial.
[00:53:56] It's a veneer,
[00:53:58] it's a check mark for Wall Street
[00:54:00] checkmark for Gartner.
[00:54:01] And I've been very vocal about that.
[00:54:03] I think that they totally blew it.
[00:54:05] They could have done something
[00:54:07] that was, was much more powerful.
[00:54:09] And I think this is, is one example.
[00:54:12] So I got a question for you.
[00:54:15] Do you think
[00:54:17] so, they didn't do anything like this,
[00:54:20] but
[00:54:22] from the context of
[00:54:24] history of crawl walk run,
[00:54:27] right?
[00:54:27] Even if we NASA is building rocket ships,
[00:54:31] you usually start with a car,
[00:54:33] right? Or something more simple
[00:54:35] in context to get people warmed up to it.
[00:54:38] Do you feel that that's a, a
[00:54:40] bad strategy though?
[00:54:42] You know, because we're in,
[00:54:44] you know, we're more, we touch A I
[00:54:46] a lot more than the common user
[00:54:48] who has no context of what it is.
[00:54:49] So
[00:54:50] an
[00:54:51] average
[00:54:51] Airtable user
[00:54:52] might be blown away
[00:54:53] by the fact that they have this
[00:54:56] A I field that they can configure
[00:54:58] and then it'll create a tweet for them.
[00:54:59] That that would be, yeah.
[00:55:01] There, there's, there's no doubt
[00:55:03] that
[00:55:03] there's great benefit
[00:55:05] in what they're proposing.
[00:55:07] I think we can all agree that
[00:55:08] every company
[00:55:09] that has integrated A I in some
[00:55:11] fashion,
[00:55:12] has improved the lives of users,
[00:55:15] almost invariably.
[00:55:17] Coda though, has also kind
[00:55:19] of blown it and I've let them know
[00:55:21] and knowing certain words,
[00:55:23] but at the end of the day,
[00:55:26] this is a company that
[00:55:28] needs to sustain its
[00:55:30] edge in the marketplace
[00:55:32] and they, along with many others
[00:55:34] are doing the exact same thing
[00:55:36] that the others
[00:55:36] are.
[00:55:37] So they're not building a mode,
[00:55:39] they're not building
[00:55:40] any moat with regard to A I
[00:55:41] and I think that's the part
[00:55:42] that they really,
[00:55:43] they really blew on the, on that
[00:55:46] they could have
[00:55:47] with just a tiny bit more work.
[00:55:49] I mean, I built this in a day,
[00:55:51] this auto palm
[00:55:53] one day, with script.
[00:55:55] But have you error tested?
[00:55:58] Have, have I error tested? No.
[00:56:00] But
[00:56:01] I'm just
[00:56:02] so it would take a clue
[00:56:04] in
[00:56:04] the
[00:56:05] wild.
[00:56:05] I'm a clueless guy out in the wilderness
[00:56:07] who kind of got it right in one day.
[00:56:10] So you're telling me that
[00:56:12] an entire team
[00:56:13] with 1.6 billion in funding,
[00:56:15] right? They raised 1.6 billion in cash.
[00:56:18] They couldn't get it a heck
[00:56:19] of a lot more right
[00:56:20] than I would over the next six
[00:56:21] months? I don't know.
[00:56:23] I, I think they could
[00:56:25] and I think these are,
[00:56:26] these are well within reach
[00:56:28] the,
[00:56:29] can I just ask?
[00:56:30] So what database are you
[00:56:32] using to store these embeddings?
[00:56:34] Or is it in memory
[00:56:35] or how are you doing that part?
[00:56:37] I store them right in Airtable.
[00:56:39] OK. In a different table?
[00:56:41] No, in fact, the,
[00:56:43] the same, the same table.
[00:56:45] It depends on where
[00:56:46] the embedding needs to be.
[00:56:48] Most all of that embedding work
[00:56:51] for this auto palm process
[00:56:53] is in memory.
[00:56:54] We just manage it all in memory as
[00:56:57] got it.
[00:56:58] Yeah. So there's,
[00:56:59] there's really no need
[00:57:00] in that particular use case to do that.
[00:57:02] Now, in the case of survey responses
[00:57:05] where we,
[00:57:07] you know, run a script.
[00:57:08] And you know, I'll just run
[00:57:10] one of these examples.
[00:57:11] So what's going on here is
[00:57:13] that embeddings are used
[00:57:15] to try to figure out
[00:57:18] you know what
[00:57:19] percentage of customers
[00:57:21] own an R V in this data set?
[00:57:23] So there's not 250 responses
[00:57:24] or something like that.
[00:57:26] But at the end of the day,
[00:57:27] you really want to use
[00:57:29] embeddings in this case
[00:57:30] to understand,
[00:57:31] well, how near and far are
[00:57:33] these various categories of information?
[00:57:37] And then,
[00:57:38] you know, deduce a percentage
[00:57:40] of people who actually own an R V
[00:57:43] within that data set.
[00:57:46] That kind of goes back to
[00:57:47] that one post I think I made
[00:57:49] many months ago.
[00:57:50] Suddenly your Airtable
[00:57:52] records can have meaning.
[00:57:53] This is what I meant by meaning,
[00:57:55] the meaning is based on,
[00:57:56] on embeddings.
[00:57:57] But in these cases,
[00:57:59] the embeddings are literally stored within
[00:58:02] the Airtable database.
[00:58:04] Yeah, we had a
[00:58:07] during Daretable,
[00:58:10] a gentleman displayed,
[00:58:12] he had kind of a similar,
[00:58:13] well, he had like an Airtable
[00:58:14] embedding semantic search he was doing,
[00:58:18] now he stored all of his in memory,
[00:58:20] but he did get issues with like larger
[00:58:22] databases, like it worked great
[00:58:24] with smaller ones.
[00:58:25] But once your database gets
[00:58:26] larger,
[00:58:27] you can't run stuff in the script.
[00:58:30] Like he's, he was getting issues
[00:58:32] with memory of pulling all,
[00:58:34] you know, I think it was
[00:58:35] like he was limited to
[00:58:36] maybe
[00:58:37] two or 3000 records, but definitely
[00:58:41] it starts to break down
[00:58:42] if you have like really large databases.
[00:58:45] So I think that's right.
[00:58:47] You, you have to pick
[00:58:49] you have to pick your,
[00:58:51] you know, your poison
[00:58:52] on vector data storage because
[00:58:55] they are quite dense. And
[00:58:57] then you have to also
[00:58:59] pick your comparative
[00:59:00] process
[00:59:01] because you're not
[00:59:03] really searching for a vector,
[00:59:05] you're searching for
[00:59:07] the high score
[00:59:09] of two vectors compared.
[00:59:11] And that has to be done
[00:59:13] in line with an iterative process,
[00:59:15] which means
[00:59:16] you really got to know
[00:59:17] your hash indexes really well.
[00:59:19] Otherwise the performance
[00:59:20] is is gonna be shipped.
[00:59:21] And so
[00:59:22] um
[00:59:23] Airtable is not ideal as a vector
[00:59:25] database store. There's no question about
[00:59:27] that.
[00:59:28] Can it be used for certain cases,
[00:59:30] maybe roll ups and
[00:59:32] small data sets? Absolutely.
[00:59:34] Should vectors be stored in Airtable?
[00:59:37] Absolutely not.
[00:59:38] In every case, there's so many cases where
[00:59:41] you need to go to pine
[00:59:42] cone or some of the other,
[00:59:43] you know, vector solutions
[00:59:45] that are designed
[00:59:46] specifically for that activity.
[00:59:50] So anyway, it, it,
[00:59:51] it can be done
[00:59:52] and it can be done well, I mean,
[00:59:54] this little example which by the way,
[00:59:56] I think is up on,
[00:59:57] you know, table forums somewhere.
[01:00:00] And I think I exposed
[01:00:02] the underlying code that helps,
[01:00:04] helps me achieve
[01:00:05] this particular kind of chat
[01:00:07] environment with a data set
[01:00:08] that's very, very difficult, by the way,
[01:00:10] because you can't
[01:00:12] send a bazillion records over to
[01:00:15] chat GP T and get answers.
[01:00:17] Yeah, exactly.
[01:00:18] In your,
[01:00:19] in your auto palm tab or when
[01:00:21] you're first running
[01:00:23] through initative process,
[01:00:24] right? And you're generating like
[01:00:27] task two or task three
[01:00:29] against your main objective,
[01:00:33] have you played with the idea of
[01:00:36] chat GP T
[01:00:38] validating
[01:00:40] a specific task against or,
[01:00:42] well, the language model
[01:00:44] itself evaluating against
[01:00:45] the task versus like
[01:00:47] a vector database
[01:00:48] as far as like contextually.
[01:00:50] So like if you have a nuance of,
[01:00:52] of that initial task,
[01:00:54] learn the basics of Airtable
[01:00:56] as person X or person Y,
[01:00:59] how well does that
[01:01:00] vector store compare? Versus like
[01:01:03] just using the language
[01:01:04] model itself to say, hey,
[01:01:06] here's this first task. Is this in line
[01:01:09] with this? Display the output
[01:01:11] in a certain format.
[01:01:13] Well, I mean, that's exactly
[01:01:15] what happens first, right?
[01:01:16] We take this task and we submit it
[01:01:18] to the L L M and say, what do you think?
[01:01:20] Right. And the L L M comes
[01:01:22] back with a body of text
[01:01:23] and maybe some tips,
[01:01:25] right?
[01:01:25] Some other things
[01:01:26] that you should be doing.
[01:01:28] And so you already know what the
[01:01:30] L L M thinks about that task.
[01:01:33] You're not saying L L M give me
[01:01:36] all of the tasks related to the goal
[01:01:39] in one swipe.
[01:01:41] It's a separate swipe,
[01:01:42] it's a separate swipe
[01:01:44] at the L L M every time.
[01:01:45] And so that gives us the ability then
[01:01:47] to dissect each one of these thoughts
[01:01:51] and create a word vector
[01:01:53] or a vector point,
[01:01:56] and then compare the vector point
[01:01:58] with the entire body or the title or
[01:02:03] the goal plus the instruction.
[01:02:05] So
[01:02:06] that's already already happening here
[01:02:08] and it doesn't need a database to do that
[01:02:11] because remember all of this
[01:02:13] thinking and inferencing is all done
[01:02:15] in line in memory. All at the same time.
[01:02:18] Which it's pretty amazing.
[01:02:20] I I'm not sure that Airtable is up
[01:02:23] to the challenge of doing this in large
[01:02:25] volume.
[01:02:26] But certainly Google app
[01:02:28] is or Google cloud platform.
[01:02:31] You know, we say Google app script thing.
[01:02:32] You know, it's just a toy javascript
[01:02:34] language. It's not,
[01:02:35] it's, it's, you know, it's GCP
[01:02:37] and it's GCP running on some very,
[01:02:39] very significant hardware.
[01:02:41] So
[01:02:42] it,
[01:02:42] it's also important to note
[01:02:44] that while you can't take a
[01:02:46] large data table and send it over to,
[01:02:48] chat GP T
[01:02:49] and get answers about it,
[01:02:51] like straight away
[01:02:52] you can with
[01:02:54] with Bard and palm.
[01:02:57] There is no ceiling on what you send
[01:03:00] to those, to those models.
[01:03:02] I think there's a
[01:03:04] practical ceiling in there
[01:03:05] somewhere
[01:03:06] but it's probably like 50 K or maybe 100 k
[01:03:09] which is pretty high.
[01:03:10] And I think very soon
[01:03:11] Google will raise that
[01:03:13] to probably,
[01:03:14] you know,
[01:03:15] it, it's probably gonna be one
[01:03:17] megabyte, maybe five megabytes.
[01:03:18] We don't know
[01:03:20] and, and that will change,
[01:03:21] you know, the nature of saying I got data,
[01:03:23] give me some answers. Right?
[01:03:25] Right now you have to aggregate
[01:03:27] before you can go
[01:03:29] get answers to your data.
[01:03:31] That that
[01:03:31] seems
[01:03:31] to
[01:03:32] be,
[01:03:33] yeah, I got you.
[01:03:34] And, and I think some of this would be,
[01:03:36] you know, architecturally
[01:03:38] would be harder to for Airtable
[01:03:40] themselves to implement
[01:03:42] and it's very specific
[01:03:43] use cases, right? Airtable
[01:03:45] is more like, hey, we're Legos.
[01:03:46] You build on top of it what you want.
[01:03:49] Airtable doesn't
[01:03:50] really have too much outside of templates,
[01:03:54] contextual features that are specific
[01:03:58] to one industry or the next.
[01:04:00] And I think that's where
[01:04:01] it would be hard for them to build
[01:04:03] a moat around specific
[01:04:04] use cases because then
[01:04:05] they got,
[01:04:06] they gotta cosign the results
[01:04:08] and we know the results
[01:04:09] are pretty wild right now,
[01:04:10] just
[01:04:11] use cases. So then it's like,
[01:04:12] hey, I'm paying for this feature,
[01:04:14] but then when I execute the feature,
[01:04:16] the output
[01:04:18] doesn't match anything
[01:04:19] I can use professionally.
[01:04:20] I
[01:04:20] think that's, that's where there's still
[01:04:23] trepidation around
[01:04:25] deeper integrations.
[01:04:27] Yeah, de definitely there,
[01:04:29] there's a, a rough edge to all
[01:04:31] of this and it's not gonna get smooth
[01:04:33] for quite some time
[01:04:35] unless your, your developers
[01:04:38] of your app are
[01:04:39] concerned about the smoothness.
[01:04:42] That's one of the areas that I am
[01:04:44] very concerned.
[01:04:45] I think it makes a lot of sense
[01:04:47] for developers to really
[01:04:48] have a hands on approach to getting these,
[01:04:50] you know, things dialed in
[01:04:52] in a way that's going to be
[01:04:53] successful.
[01:04:55] And once again, embeddings
[01:04:56] come to the rescue because
[01:04:58] once you generate something
[01:05:01] you can use that as a backstop to say,
[01:05:04] is this accurate?
[01:05:06] Right? We, you know,
[01:05:07] given other things
[01:05:08] that we know are accurate,
[01:05:09] how close is it to that accuracy?
[01:05:12] So embeddings can actually
[01:05:14] be used to, to identify
[01:05:15] areas of your A I that are off the rails.
[01:05:18] It's no different than saying,
[01:05:21] let's use statistical
[01:05:22] measures to understand if somebody
[01:05:24] isn't cooking the books,
[01:05:26] those are the same things
[01:05:27] that kind of go on day after day.
[01:05:29] By the way, this little example,
[01:05:31] which I also put a link to
[01:05:33] is very simple.
[01:05:34] It's basically doing,
[01:05:35] I think what Airtable has already
[01:05:36] announced
[01:05:37] and that's why I said,
[01:05:38] I think in one of the writings I wrote,
[01:05:40] why are you waiting?
[01:05:41] Just get started.
[01:05:42] You know, find somebody that
[01:05:44] can help you with a little bit of script
[01:05:45] and before you know it,
[01:05:47] you know, you're able to
[01:05:48] say, yeah, give, give me a company name.
[01:05:51] Now summarize the company,
[01:05:52] find out how many employees there are.
[01:05:54] Tell me the market cap for the company.
[01:05:56] I mean, those are very rudimentary,
[01:05:58] very simple
[01:05:59] and they're easily validated too.
[01:06:00] So
[01:06:01] I, I think A A I is here.
[01:06:03] You don't have to wait for Airtable.
[01:06:05] Yeah,
[01:06:07] that's awesome. We're up,
[01:06:08] we're up against the hour. Lorenzo.
[01:06:09] I wanna make sure you get a chance.
[01:06:11] Anything you wanna add to this discussion?
[01:06:14] No, I, I think,
[01:06:17] you know, with a I,
[01:06:19] you could talk for hours right? On
[01:06:21] on any little nuance. So I think that's a
[01:06:24] good discussion. I think
[01:06:26] in the next few years,
[01:06:28] I do think we will have a,
[01:06:32] a separation.
[01:06:33] I believe that companies that
[01:06:35] learn how to adopt it,
[01:06:37] learn how to integrate it,
[01:06:39] teams and tools and use it
[01:06:40] every day
[01:06:41] will
[01:06:43] be really far ahead
[01:06:46] than the companies who don't adopt it.
[01:06:48] Right? Similar to the people who decide
[01:06:49] they don't need a website
[01:06:51] or they don't need,
[01:06:52] you know, a smartphone or a computer.
[01:06:54] I think the knowledge and the
[01:06:57] the skill and the ability
[01:07:00] of the individual person that uses
[01:07:02] AI to
[01:07:03] augment their existing knowledge,
[01:07:06] I think will be far ahead.
[01:07:07] You know,
[01:07:08] and I think people on the fringes of like,
[01:07:11] like no code movements
[01:07:13] and, and whatnot, I think
[01:07:15] it'll be a challenge for people
[01:07:17] like that to up their skills.
[01:07:19] Right? You know, where A I
[01:07:22] will replace some things that they do,
[01:07:24] but it will introduce
[01:07:26] new opportunities for them
[01:07:28] to do things they couldn't
[01:07:29] do before at faster pace.
[01:07:31] But they have to be up
[01:07:33] to the challenge, right? So it's like
[01:07:35] the farmer that's there
[01:07:36] and then the tractor rolls
[01:07:37] through and it's like,
[01:07:38] oh, my goodness. Well,
[01:07:39] you gotta learn how to drive a tractor,
[01:07:41] right? If, if you decide you don't
[01:07:43] want to or it's too advanced,
[01:07:44] you will be left behind,
[01:07:47] you know, in a, in a way
[01:07:48] because farming is a really good one too.
[01:07:50] To think about
[01:07:51] it. In the 19th century,
[01:07:53] almost everybody farmed.
[01:07:55] And in this, in the,
[01:07:57] the current century, almost nobody farms
[01:08:01] are we all standing around without jobs?
[01:08:03] No.
[01:08:05] We're not.
[01:08:06] You know, farming,
[01:08:08] farming completely
[01:08:09] obliterated an industry,
[01:08:10] you know, automated farming
[01:08:12] and the same will happen
[01:08:14] for information. We will not be
[01:08:16] standing around
[01:08:17] without jobs because of A I.
[01:08:19] Yeah,
[01:08:19] but the other thing,
[01:08:21] right, the people who learn
[01:08:23] how to drive the tractors
[01:08:25] and do that are gonna be,
[01:08:26] you know, they're gonna be
[01:08:27] able to be more efficient,
[01:08:28] they're gonna be able to
[01:08:29] grow food faster.
[01:08:30] So I think it's the same thing, but
[01:08:32] I think you can't be scared to, to learn
[01:08:35] code and script and some
[01:08:36] of these newer concepts, I think
[01:08:39] thinking that you're gonna
[01:08:40] be in this world of no code.
[01:08:42] I, I, I think you're gonna have to
[01:08:44] and Airtable is a great tool for that
[01:08:46] because it does introduce
[01:08:48] concepts of scripting
[01:08:50] easier than most.
[01:08:51] So I think people are gonna,
[01:08:52] you're gonna wanna step your game
[01:08:54] up for sure. A, a little bit.
[01:08:56] That would be
[01:08:57] my advice because three years
[01:08:59] from now it'll be a different world.
[01:09:01] Yeah,
[01:09:02] almost three months from now.
[01:09:05] Three
[01:09:06] months,
[01:09:06] very fast. Bill can't wait three years.
[01:09:09] Yeah,
[01:09:10] I'm gonna be dead.
[01:09:11] Let's get on with it.
[01:09:12] We'll make it happen. Basketball.
[01:09:15] Very good. Thank you.
[01:09:17] This is an awesome discussion.
[01:09:19] We need to do this again.
[01:09:20] Maybe in three months
[01:09:22] we'll plan on that
[01:09:23] and see if Bill's correct.
[01:09:24] If it's all changed,
[01:09:26] what's changed.
[01:09:28] Sunny days, we'll have a whole
[01:09:30] new list of bullet points to talk about.
[01:09:31] Yeah.
[01:09:32] Yeah. Very good.
[01:09:33] Thank you both for coming on
[01:09:35] for the, for the insight and
[01:09:37] always good to have you on.
[01:09:38] We'll have you back.
[01:09:39] We'll see everybody next week.
[01:09:41] Bye.