#9: Using AI to Fast-Track Your Company's Knowledge Base

Welcome back to the junction Today we are

going to be talking about using AI to

fast track your company's knowledge base.

I mean, as long as I've been working, there's

always been some sort of, like, company intranet.

Various companies I've worked at SharePoint, maybe

it's just like one document that lives

on the server, the Wiki.

But there's like a place that

hopefully people are collecting company policies.

It's written on paper in the basement in that

file storage thing that people used to use.

You know what?

That used to be a thing, like printing out.

You would document your process. You print it.

Is that the sink?

No, people put it in a binder, probably.

In a binder.

In a binder.

A three hole punch binding the three hole punch

at the office used to be a nightmare. That was a thing.

That was a thing.

It's like the stapler from the office. Right.

And you had to print out for the spine

of the binder so that once you stacked it

in the book case, you could date yourself here.

I just want to make sure people know.

I remember the days I just didn't ever install

the drivers for the printer, so I could say,

oh, I can't print anything, I'm sorry.

Oh, I've been there. Yeah.

Okay, so let's talk about, I guess, where

I was going with that is having a

knowledge base is not like a foreign concept.

There's probably companies that are

doing it really well.

I think Slack actually just rolled out

within the last year, six months, like

an intranet sort of functionality.

We've not, to my knowledge, used it here at Venn,

but we have an intranet It's on our own know.

Is it google drive? Is that what it is?

You use it all the time. Apparently.

I have it bookmarked, and I know

it's intranet Ven.com, something like that.

Should I not say that?

Can anyone get to have to you have to be logged in.

Okay, you could try, but it won't work.

All right, so we've got this sort of centralized

knowledge base, but then you have teams, and each

team has a different way of working where marketing

lives and works and spends our time.

I mean, we're all doing stuff in Slack, but as

far as, like, documenting processes, I might be documenting my

processes in a Google Doc, which doesn't necessarily get sent

over to the I don't I don't know. That's not what I use.

I don't have access to go upload or edit. Right.

Nobody's told me at the company.

That's where I need to put my process or policies.

Maybe not yet after this recording, though. Sure.

No, I'm a firm believer we need to have

a single source of truth in some capacity.

But I know in talking to other department

heads at our company that our knowledge base

is kept in different kind of corners.

And so there is this, in some sense,

a challenge of like, how do I do

this thing if I'm not in that department?

Or let's say I'm a new hire and I go in to

join the customer support team and I'm resolving issues or bugs, and

I want to know if that's ever been solved before.

Is there an easy way to go find that?

Maybe jump into slack. Right?

Do a quick search?

So this whole idea of being able to use

AI tools to ask these questions, like natural language,

that's what I think we should talk about today.

This Use case of this knowledge

base, we're calling it Ask Bjorn. Right.

And if you don't know, Bjorn is our favorite yeti.

Yeah, well, I think it's the world's favorite yeti.

Naturally. Yeah.

We can ask social media.

Let's take a poll yeah. Here.

Answer our survey questions after this show.

So let's talk through the Use case a little

bit, and maybe if there's other I think other

people would be interested in doing this at their

company if they don't already have some capability to

and I'm not talking a search bar. Okay.

Like you type in right?

No, it's a beyond that.

Yeah, it's beyond that. Right. No.

Too bad.

Knowledge base.

What's really unique here isn't that we don't

document it isn't that we do document.

It's that we really struggle to disperse

information at the time that it's needed.

I think every company, oh, this is not just us.

This is pervasive.

Well, you've got a dominant companies

that haven't mastered even documenting it. Right.

But the moment that you do, you have to

be able to disperse that information at the time

that it's needed, because for the most part, it's

typically information that is really monotonous or boring.

Right.

And we don't want or need

people necessarily to memorize it.

And if they do, they might memorize it incorrectly.

And so where we're struggling or where we have struggled

with this and why we're trying to solve this problem

is that these questions pop up in the middle of

phone calls with customers, with prospects, right.

And they're asking, hey, are you SoC two compliant?

And that sales guy that just started

yesterday is, like that moment of pause.

Now confidence is gone.

This guy doesn't even know.

And so being able to answer those questions

quickly, even without having the individual typing anything,

it's almost like AI is listening to the

call and preemptively answering the questions while they're

being asked, hey, does Intact have an API?

And in your little panel, maybe in Slack, I don't know.

It pops up, and it says here's

the question, does Intact have an API? Yes.

Here's the link to it.

It's that kind of idea that we could

very quickly give people the answers to the

questions that they're looking for, and not just

like, hey, here's, like, ten different documents. Right.

Or here's the top ten things.

It's the answer to that specific question

at the moment that you need it.

And that's kind of the idea

that we're trying to tackle.

And so ask Bjorn. Right?

Is this kind of idea that we've got a ton of

different unique things about ven the way we do things.

We've got a project methodology, we've got

six different phases of that, right?

And we've got this hypercare thing and then we've got

kind of these managed services and we just deal within

so many different platforms that we have to have a

wide diverse knowledge base on how things are.

The terminology, right? The nomenclature.

HubSpot calls it companies and

Salesforce calls it accounts.

There's just so much more that we have to digest and take

in to be able to speak about it in an expert manner.

One of the things I think any

small and growing scaling company struggles with

is training and that institutional knowledge.

You're kind of teetering on that, right.

So it's one thing to have here's our policy on this thing

and it's another thing to this is our way of work.

This is our methodology.

It is documented. Right.

But usually when we've onboarded people, it's a person

delivering that to say, okay, I'm training you up

on these applications and here's the fields, right?

And a lot of this happens on the job.

Or let's say they already are familiar with it

and they're qualified when they come in the door.

But these are the types of things that it's

not just a communication problem, it's not just a

training problem, it's not just a sales.

It's multifaceted.

And so that's one of the things that it's one use case.

But I think we could argue

that apply it across multiple disciplines.

I think everybody resonates with this.

Well, maybe I don't because I

don't really care what I eat.

I'll just eat anything.

But there's a lot of questions of like when

you're at a restaurant, hey, is this gluten free? Right?

And this guy just started he doesn't know,

but he can go figure it out. Right.

Any kind of like that?

No, it's huge.

Like giving equipping people with I was just

at the Container Store over the weekend, okay?

I picked up these. You would? Yeah.

I did mixing bowls, but they had these

wood tops and I was like, I don't

think these are probably dishwasher safe.

I'm a hand wash kind of gal. I'm not afraid of that.

I'm more concerned if they're cedar or if they're oak.

What kind of varnish do they have on them?

No, they looked like acacia.

Oh my gosh.

They looked fancy.

They were on sale. Okay.

But I wanted to know so that I can then

communicate to my household these can't go in the dishwasher.

So I asked the associate, said, do

you happen to know by chance?

I didn't see it on the label.

Can I wash these?

She goes, you know, I don't know.

She looks at it just like I did. Yeah.

And I was like, I didn't see it there.

She scans the barcode and goes, the bowl is

dishwasher safe and the lid hand wash only.

And she literally after says that.

She goes, I love technology.

That's awesome.

She goes, I before would have had to

go ask somebody in the back room.

That was its own little case study right there.

And so granted that we're not talking

about AI, probably not even talking about

automation in that sense, right?

But it's still powerful to be able to surface that information

for someone who is new or doesn't have to walk to

the back office to get a manager for information that they

are in a position that they should know.

Well, that same kind of idea is

what we want to tackle, and not

just from the traditional knowledge base, right?

Like, okay, here's the question, okay,

here's the link to the answer.

And not to make that sound like overly boring, but

people have been doing that for a long time.

What we want to do is above and

beyond that, we're going to have multiple we

do have multiple types of data sets, right?

We've talked the heck out of call transcripts, right?

We've got all of these slack channels, we've got a

ton of email, all of these different repositories of data.

And so this idea of a knowledge base kind

of extends above and beyond the traditional sense, right?

Askborn is this idea that we can ask of

all of the data that we have, right?

I want to ask this a question like, hey, what were

the topics from we do this thing called Famoli on Monday.

What were the high level points from Famoli on Monday?

Traditionally. What?

Does Intact have an API?

It's like kind of this general knowledge base assistant

that has access to multiple types of data and

for the most part, it's almost real time.

You know what I like about it too?

You're not dependent on another human to answer it.

And let's say because we all do it, you

don't want to ask for fear of looking like

I should know the answer to this.

Yeah, we do that.

We were doing that in GPT, right?

Like we're asking a little bit more finer detail

or like what does this acronym mean again? Right?

So it takes away the layer of now I have to bug Chase.

Yeah.

And he's also going to think

that's a really dumb question.

I trained you on that, right?

So it takes that away.

But let's go to number three.

What I love about it on the back end,

we can see what questions employees are asking so

that we can better improve our training up front

and our onboarding and we can tweak those areas.

We can surface that information on

the Internet in this knowledge base.

We can go, gosh.

Everybody keeps asking where our vacation policy is.

Is it not linked out?

Why does everyone oh, I think

everyone just needs a vacation.

But we can see trends in the questions people are

asking and figure out where we have gaps in information

and maybe the things that we need to be better

about communicating or training on up front.

Yeah, I'm just going to start asking it.

Why is Mel so great at what she

does and then that'll start surfacing up.

How is she so good?

What does she do?

She must get up at 430 in the

morning and go work out every day.

That's not true.

You do that.

Okay, so back to this.

What are our requirements?

Right, so going back to the Use case, we've

identified the problem, we know the input is definitely

going to be our Venn intranet at minimum.

You mentioned Slack, you mentioned email.

I'd be curious how you actually operationalize that

because you have to probably have a lot

of filters and gates on that.

You don't want to feed it everything.

No, that is something that we will need to

focus on from a security standpoint and that's not

necessarily maybe it's in the automation realm, right.

But giving the security clearance, right.

The top secret information kind of thing

is hot in the news these days.

But it's like the need to know,

do you need to know this?

And that level of automation is something

that we will pipe in it's.

Well, what is the email about kind of thing. Right.

Is it relevant to what you're asking?

Should you have access to this kind of thing?

And those are controls, security controls

that we need to build in. Right?

I mean, I guess I should say this isn't

something that we're not already to some level syncing.

We have the HubSpot Gmail connector

turned on for everybody's inbox.

So we're pulling emails and email

history into accounts, contacts, opportunities.

We're syncing those between salesforce and HubSpot.

So there is already some level of that happening.

You do have, well, you have some level

of just hierarchy of automation and data access

that we already have access to. Right.

I'm working on a deal. Right.

And all of the emails with that

individual are associated with an opportunity.

Well that probably means I should naturally

have access to those emails, right?

But maybe Scott sends an email to somebody at

the same company, but that person isn't associated with

my opportunity, so I shouldn't be able to what

is, what are Scott and Mel talking about?

Those are things where that security and that

quick automation is going to preemptively decide.

Chase can access this and he cannot access this.

So when he asks questions, this is the

data set that he should have access to.

So in practice I'm on the team here at

Venn and I have access to Ask Bjorn.

I go into Slack and I typed in a prompt. Yeah.

Is that it?

Yeah, that's cool. Boom.

So then I would only get access to answers or

it would provide an answer appropriate to the question based

on what we've trained it or fed it?

Yes, it goes back to that context. Right.

And this is where? Above and beyond?

Or maybe below and beyond.

AI right, we're doing that automation.

We're taking in the context.

Right, Mel?

This is Mel asking, right?

She's a director level.

She's in charge of marketing, right? Yeah.

Mel isn't associated with any opportunities, but this

type of question is associated with marketing.

And therefore we're going to surface up this data set.

Right.

I remember when I first started for Ven, I asked,

can I get access to all of our sales calls?

All of them?

It wasn't as simple as getting access to all the

sales calls because I wanted to mine them for insights.

And we were doing some really cool stuff that

we've talked about in an earlier episode around taking

the zoom recording, putting it into salesforce and then

also backing it up on Google Drive.

But I don't think we were differentiating between this

was a sales call and this was a call

between a manager and their direct report.

So it wasn't as simple as slicing

and dicing the calls that way.

And because there are very valid reasons and concerns

around giving someone access to all recordings, it became

sort of this, how do we solve that?

And then it was like, well, okay,

just send me the good ones. Easy.

Yeah, okay.

Come Friday, I didn't get any calls.

I know they had calls.

Hey, guys, you have any good sales calls?

Do you have any bad ones?

I just want to know what they asked. Right.

I was trying to get understanding, right?

Especially on those first one, two calls.

What questions are they asking?

As you say, that based on somebody's basically permission set,

we would be able to surface that information and maybe

need to know is a good general statement.

But in the same idea that we want

to have security and control, maybe we have

1000 emails with this one individual.

You don't want to surface 1000 emails. Right.

You're looking for a very specific question as it

pertains to a very specific set of data.

So taking in the entire context of the question where

this person sits in the organization, what type of function

are they trying to provide to the company?

We can take all of that.

Grab the right info that that person should

have access to, to pop up like this

kind of knowledge base, this instantaneous knowledge base

that answers the questions as fast as possible.

Okay, so we keep doing this, we get really excited, but we

got to bring it back to how do we implement it?

What does it take? Right?

So for that person out there listening,

nodding along, they're like, this sounds great.

I want to go do this.

Yeah, hook me up.

What is our next so have we scoped it in terms

of the number of hours it would take the resources.

What does that look like?

This is probably one of the more challenging things

to achieve unless you start out really simple.

And what we typically end up doing

is breaking it down into the smallest,

maybe you've heard minimum viable product.

Let's start with the smallest thing and iterate on it.

We were not going to give it all of the data right

out of the gate because that will take way more time.

So we'll probably start out with the intranet, right?

Digest the internet, give it access to that data.

There is some more advanced things that you

can do where basically you can give it

kind of like the Dewey Decimal System, right?

Old school. That's been a minute. Yeah.

Middle school librarian. Right.

Basically give the AI the location to the

answer, depending on what the question is.

That is probably the first step

that we're going to tackle.

And there's a specific way you do that

within the AI systems specifically to OpenAI.

You do it within this idea of called Embedding.

Embedding equals Dewey Decimal System. Right?

Basically, hey, these are where the answers

are to these types of questions.

It's way more complex than what I'm making

it out to be, but that's probably the

first thing we're going to tackle.

So get your data right.

Then figure out the best way to give

this model access to that data, which is

probably the most challenging thing to do here.

And if you're not technically inclined, reach out.

Love to help you, give you

some high level answers here.

Or maybe we can help you do it altogether after that.

That's where you got to provide that level of UI

and that's where we're going to pull in slack.

We don't need to go build out our own chat bot.

What about the permission sets?

Yeah, great question.

That's going to be within salesforce. So salesforce.

We've already got all of these permissions.

We already know where people are in the hierarchy.

What are they connected to?

Are they connected to that opportunity?

Is this even their account?

That level of security we're going

to let another system control?

The whole goal isn't to try

to build things from scratch.

It's to build upon the things that we already have.

This thing that we keep going back

to layering the AI on, right?

Yeah.

That's really interesting.

Should we be concerned about hallucination?

We've talked about it.

It wants to come up with an answer.

It has high confidence even though

you're in your own system.

So let me give you an example.

I ask how many days of PTO do I have at Venn?

1100. Right.

Well, I don't know.

I mean, I've seen crazier things that GPT is done.

So I'm just wondering, how do you put safeguards?

This probably goes into training is probably one

take that you could training the model.

Training the model, right.

You could pull in simple rules, like

when they ask this type of question. Here's the prompt.

How many days of PTO do I have?

The answer to the question is refer.

Like the refer to the policy?

Refer to your Gusto account we'll use. Right? Right.

Okay.

And you could go about it that way.

You could also love that you're

jumping them out to another system.

Chase come on.

You're trying to game it, right?

Can't you just pull in that like you've

got access to the Gusto API or something? I don't know.

We totally could. Okay.

We talked about accessing all

these different data sets.

We could totally do that. Okay.

But if you want to put in rules in place, right?

Like, I don't know.

I'm thinking bad actor, right?

Malicious.

You can put in kind of

preemptive type deals like rules.

And I think this is what

OpenAI is doing to some extent.

When you ask really? I don't know.

You ask it to do something illegal, right?

How do I make a bomb?

Or how do I do some drugs?

Or whatever these illicit questions are.

You can basically rate the question to determine if it

is something that we should or should not be answering.

And if it is not, you can kind of flag,

and you can preempt and send a response that says,

hey, Ask Bjorn loves your question, but unfortunately, this is

something that we're not able to answer.

And you don't even send that over to your

large language model, and you just stop the deal,

right, and just send a nice, polite response.

Or it could be rude polite or

rude, whichever tone you want to take.

And then it sends a slack

message to their hiring manager. Yeah. Red flag.

Red flag.

There's a couple of ways that you go about

doing that, and that is a good concern.

Hallucinations right.

Are going to be something that

you should be concerned about.

And this probably goes back into something that

we try to focus on really hard.

And that's testing.

You got to test this thing.

Ask it all types of questions.

Always be testing.

How many days of PTO does Mel have?

And if it says a hundred, right.

And you probably don't want to give

Mel access to that right away.

Mel's gone.

Mel's on vacation by that point. She's in Cosmel.

It's fine.

There's Internet down there.

I think that makes sense.

Cost I don't think we

really answered the question there.

So are we talking 10 hours of someone who's technically,

like you said, if you don't know where to start,

there are smart people out there that can assist.

But if someone has the resources or the

technical know how in house to do this,

we're talking 10 hours, 50 hours.

It probably depends on the size of your data

and how many data but can you ballpark it?

How long is it going to take us?

Yeah, well, this really depends

on where you're starting.

With OpenAI, you can build and train a model off

of an existing model that has already been trained, right.

And that kind of ideas where you're providing

it existing, sorry, a small data set.

We talked about kind of like 100 prompts or

100 examples is kind of like the bare minimum.

Well, if you start out with the bare minimum,

we're talking like a couple of bucks, right?

Literally a couple of bucks.

Obviously there's a lot more involved to

get it all connected and touching all

your data and all that stuff, right.

And then we talked about in an

earlier episode as far as usage.

Okay, so a couple of bucks to set it up,

but the cost, ongoing cost, would probably be fairly minimal.

Yeah.

And I know people are going to be like,

oh my gosh, it's way more expensive than that.

I'm saying at the bare minimum, right, the

lowest, the least amount of effort, right.

We're talking about, I don't know, call

it less than $100 to get it

trained and running some basic prompts, right.

There are certainly use cases where you're going to have

to send in hundreds of thousands of examples and you're

talking about a decent chunk of change, right.

Ten grand, 50 grand plus

depending on what you're doing.

How much context are you putting in the prompts, right?

Basically it goes back to the tokens.

If you send it 100 prompts

worth of tokens, little cost, right?

If you're sending it 100 million tokens, well, you're

talking a lot, a lot of money, right.

Same thing on the prompt side, if they're asking simple

questions like how many days of PTO do I have?

It's probably like 1015 tokens.

But if they're asking really complex questions that

are very lengthy, like paragraphs worth and then

everybody's doing it over and over again, well,

now your costs start to add up.

And for the most part, I think this is somewhat

linear based on the number of people that you have

because they're all probably going to be asking similar questions.

And if they're not, again, the insights that you'll get

from the types of questions might be able to help

inform where you can conduct some additional training.

Oh, for sure.

Or even surface some new capabilities of team members.

They're asking questions.

It's not about PTO, it's kind of about

your customer segments and things like that.

You might go, wow, maybe

it's not something I've considered.

Well, I would advise you against just unleashing something

and just like waiting for the bill to come.

But I love your idea of like, hey,

let's watch what people are putting in.

Not in a big brother way.

Yeah, but like, hey, they are asking about PTO a lot.

Maybe we do need know surface this info.

Maybe we should automate it and tell you every

paycheck what's your PTO balance in slack automatically.

You could in theory reduce your cost by

answering those questions in advance in some faster

format than them logging into Gusto.

Well, and I do think it's important.

As with any investment in technology, we

run into this a lot with clients.

A lot of times people don't know the

cost of an integration, especially if there's a

level of service to do the work.

In the work that we're doing, especially custom integrations,

they're not just turning it on and it runs.

There's also ongoing cost to maintain it. Oh, for sure.

But when you look at the time savings and you

actually quantify the hours saved on your team or full

time resource, the cost of having a salesforce administrator on,

and we could go on about that.

You have to do the same exercise for these use cases.

So how much time is spent today where team

members are asking their managers and their HR coordinator,

their HR business partner questions of this nature?

First of all, who's the subject matter expert,

what data sources, who maintains the data source,

and how often are they surfacing this information,

sharing this information with people?

And that's just like rudimentary just time savings.

But that time translates into money

because they're not doing something else. Well.

Certainly don't take my, hey, it's

a couple of bucks, right?

And be like, oh my gosh, we're going to make bank.

But I also don't think that just because

something costs $10,000 or $50,000 doesn't mean that

it's not worth the investment over time.

Like you said, it's that kind of ROI

and calculation that you need to do.

And if what you're trying to achieve is $100,000 worth

in cost or investment right on the upfront side, what

are you going to gain on the back end?

And these are all questions that you

should be asking naturally in anything that

you do from a business standpoint.

I mean, we've been doing it this is

a little bit different use case, but we've

been doing it from our last episode.

We talked about Zoom transcripts.

One of the questions I know that first came to

my mind as we were talking about these things is

let's not go replicate something that a salesforce or a

HubSpot or a zoom is already going to go do. Right.

Granted, it's in one system. Right.

We're kind of talking about bridging the gap between

systems, but we don't want to go replicate something

with AI that another tool is already probably going

to go do they haven't done already.

That's also just another consideration.

I think it's important for anyone listening.

I do need to address this headline that you put in

for this episode, five Ways AI Might Destroy the World.

Quote, Everyone on Earth could fall over dead

in the same second from The Guardian. Wow.

In the same exact like, even if that

was going to be a thing, I feel

like that just wouldn't theoretically be possible.

Even if a meteor hit Earth, it's going to take

more than 1 second for the whole it's got. To be. Yeah.

I get Terminator style that these things might be

problems, but I think there are some general fears

and people just don't know what they don't know.

And when you don't know what you don't know, right.

That kind of worries you. Yeah.

And I can kind of get it.

I fell into that category early on where

I was like, I don't really understand it.

I'm not sure everybody thinks this way.

How's this going to affect me?

Yeah, well, I got to tell you, I

don't really have much commentary on that one.

That one kind of says it all and

it feels a little looking through these things.

If AI systems wanted to push humans out, they

would have a lot of levers to pull.

You know what, that's only if you

give it access to those levers.

I said those earlier, it all runs on electricity.

So if you unplug it, guess what?

It's not going to work. Yeah.

I think for the most part here's, my biggest concern

is that we will achieve some level of AGI, right.

It will have some ability to self improve, to act on

its own, and if we don't do it right, or if

there is a bad actor in play, that thing will run

off and they're not going to pull the electricity because it's

probably benefiting them in some form or fashion.

And that's probably where we're

not going to be prepared.

The one that sticks out the most for me is

you can trade on the market instantaneously through API, right.

So this AGI could figure out how to make money

very quickly and it could crash markets very quickly.

Like the instant boom, everything's dropping and that

could set off some level of economic disaster.

That's my biggest concern.

I think we are still quite a ways

off and thankfully you can't just like willy

nilly go trade money on the Internet.

There are some know your customer kind of deals, right.

You have to put in your Social Security number. Right.

You have to pump some level of

money into it to get money back.

There are some safeguards. Right.

But that is one of my general fears.

I do think, as with any of these things, when you

start to look at why am I afraid of this?

Trying to have a better understanding of it oh,

for sure is the first step you can take.

So listening to this podcast, I don't know, we're

kind of like working our way through it, right.

We're trying to understand and unpack these

questions that other people have ourselves, right.

We're not AI.

We're not out there speaking

at IBM Gartner conferences, right.

But I do think that the more we open

up the conversation and share info, share knowledge, share

here's how we're using these tools for good.

Here's how we're using these tools to run our business.

Here's how we're using these tools to

be better professionals in our domains.

That's where I think taking that first

step to understand what's possible, we'll be

more equipped to when that happens. Hopefully never.

But as things do happen, because they will, that

are not positive, we at least have kind of

a baseline or a foundation of naturally.

Let's hit up the second one.

This one is really interesting.

Ukraine gained advantage in a war against Putin

with custom built AI unprecedented testing ground.

So this particular article is out today.

It's basically talking about how they're using AI

to compete against their number one enemy.

Obviously, lots of crazy events

unfolding over there, drones.

I mean, basically this is next gen warfare.

And what they're doing with AI is basically playing out

all of these different scenarios and finding the best.

What is the best next thing I should do?

It always goes back to giving it a lot of context,

a lot of data predicting what I should do next.

And this is what's going to be really interesting.

It doesn't necessarily talk about this in the article, but

what gets me slightly concerned is when we are giving

the AI control over the trigger, it can shoot, it

can pull that trigger, and it can deploy the bomb.

Right.

It's one thing to identify

and categorize events or people.

It's another thing to take life.

We're programming these things to take life.

That's probably my biggest concern now.

Do I believe Ukraine is doing everything they can

to push back this enemy, to push back Russia?

Absolutely.

They're going to take this to

the extreme because they're being attacked. Right.

They're being invaded by a foreign invader, and they're going

to do whatever it takes to get rid of them.

That's interesting. Yeah.

I mean, there really is it feels I know there are

still we talked at one point what AI can and cannot

do, and since then, this has sort of kept me up

at night, like making a list of I can't do that.

I have a hair appointment on Saturday. Right.

It's not going to cut my hair.

Is it going to color my hair?

I mean, I guess there's like,

robots one day that'll do that. Right.

I hope that's not a thing, but there are absolutely go

make a list because it'll help you I say it's keeping

me up at night, but it also help you sleep better

at night because it's easy to go down this rabbit hole

of doom and gloom with this stuff if you're not careful.

Right.

It can be used for bad.

Oh, for sure.

Scenario.

Well, you can always at least, at

the very least, go back to it.

Can't do it unless somebody does it.

Like some individual is going to have to

cross some if it's an ethical boundary, right.

They're going to have to cross that boundary.

Then they're going to have to program it.

Then they're going to have to test it.

And so it's just not going to do that.

Out of the gate.

So you do have some time, right?

You can still sleep?

Well, knowing that that isn't

going to happen overnight.

Maybe in five or ten years. Okay.

Maybe you should, I don't know, take some

meds, wear your face mask, have good dreams. Not bad.

Self care? Yeah, self care.

I believe in a good self care routine.

Or just don't go to sleep and work out. I don't know.

I guess we'll be talking about it at some point.

Hopefully in the future, we'll have an episode on it.

Episode 345.

Hey, remember back in 2023?

They were so young.

We would still be young, of course. Yeah.

I don't know about you.

I don't know what you're trying to say.

I don't have any white hairs.

All right, well, that's all we

got, so thanks for joining us.

We covered a lot of ground, so again,

please let us know what you think.

Send us an email thejunction@ventechnology.com until

next time, keep it automated.