Re: [AFMUG] MoviePass documentary

[Ken Hohhof]

It's because of the endless shrimp!  Except it isn't.
https://www.eater.com/24160929/red-lobster-bankruptcy-endless-shrimp-closing-locations


-Original Message-
From: AF  On Behalf Of Robert
Sent: Monday, June 3, 2024 4:08 PM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

What VC invest in flows back and forth. Most partners in vc come from the ranks 
of the companies the vc invested in and that then profited the vc's back and 
made the future partners rich.  So the "method of success" 
depends upon the experience of the jr. partners and what they decide to put $$ 
in is related to their risk experience.   Facebook had _no_ method of 
monetization for the first 4 years.  They were just given runway and cash to 
expand with the understanding that it was so compelling and sticky that pulling 
$$ out of the enormous user base would be easy to generate $$ when the time 
came.  VC "harvest"  $$ from their investments after IPO.  They understand that 
they hold enough shares to trigger runs on stocks if they pull out too fast 
after IPO Typical harvests are over 10 years and after that they care little 
about the companies.  Until then they are very involved.  ( Look at when google 
was turned back over to the founders. ) ..


So "burning piles of cash" goes in and out of style.


Streaming is a whole separate animal due to all the ways that the content is 
tied into legal knots.  Look at the streaming companies and most are tied to 
major content companies already so as to secure the content through legacy 
agreements.   VC wants little to do with that pile of crap.  They got burned 
out on that 15 years ago.  You have to have a very special proprietary key to 
that lock for them to be interested and proof that the key fits the lock.


I've gone the VC route 3 times as part of exec teams.   Once 40 years ago when 
I didn't ask for enough to show I knew my head from my ass...
25 years ago when I was hooked to a company because of my "rep" and watched us 
( me included ) try to sell BPG optimization as a major company possibility.  
And prior to that as part of a team that was looking for second round for a 
collaboration tool.  That was hard as the users already liked the product but 
it was in a downturn in VC funding and everything was getting rejected.

The hardest part of vc funding is that they are like sheep, when one goes a 
particular direction they all go that direction.  When one company gets known 
as "hot" they all chomp at the bit to fund, otherwise it's  horrible what they 
do to the companies.   They won't sign NDA's, share info under the table, do 
stuff that should get them in front of the FTC.   "Investors" are still pretty 
much run by the VC's IMNSHO.. Just more through the back doors and buddy 
system...  Because after you get the little $$ out of the way you run into the 
ones with the deep pockets and those are run by the VC's...

On 6/3/24 10:45 AM, Ken Hohhof wrote:
> The old saying is "lose money on every sale but make it up on volume".  The 
> new equivalent is "fake it 'til you make it".  Amazon is the classic example 
> of one that worked out.  Many have a non obvious revenue model that involve 
> selling ads or customer data, that's how TV manufacturers can sell TVs at 
> cost.
>
> But investors seem to be losing patience with business plans that just burn 
> piles of cash with no plan to ever become profitable.  This seems to have hit 
> streaming companies, which everyone assumes are all profitable, but actually 
> most are not.
>
> MoviePass seems to have been doomed, as was WeWork.  There was Theranos.  
> Seems like there is a blurred line between "fake it 'til you make it" and 
> outright scams and pyramid schemes.  All of us can attest that you need to 
> ramp up to a certain number of customers to achieve economies of scale, but 
> it should be possible to show some math how that leads to eventual 
> profitability.  Otherwise you're like the underpants gnomes.
>
> https://www.youtube.com/watch?v=a5ih_TQWqCA
>
>
>
> -Original Message-
> From: AF  On Behalf Of Robert
> Sent: Monday, June 3, 2024 11:52 AM
> To: af@af.afmug.com
> Subject: Re: [AFMUG] MoviePass documentary
>
> I consulted for three VC's after Nscp and L3...  The wisdom in their thoughts 
> at the time was a home run in one out of 20 was doing well... Their thoughts 
> about founders was about 1/3 idea and 2/3 the ability of the founder to 
> gather talent around them...   Your rolodex was as important as your 
> execution...  Seed money is easy.   A good presentation deck and glib tongue. 
>   Getting money for the first round isn't that hard.   Getting the second 
> round was the test of your business. 1MM seed, 4-10MM first, unicorns got 
> 100MM for second, and the sky was the limit on the 3rd...
>
> On 6/3/24 7:18 AM, dmmoff...@gmail.com wrote:
>> " ...the startups that light piles of money on fire until it runs out 
>> seem to be actively flying the plane into the dirt.  

Re: [AFMUG] MoviePass documentary

[Robert]

What VC invest in flows back and forth. Most partners in vc come from 
the ranks of the companies the vc invested in and that then profited the 
vc's back and made the future partners rich.  So the "method of success" 
depends upon the experience of the jr. partners and what they decide to 
put $$ in is related to their risk experience.   Facebook had _no_ 
method of monetization for the first 4 years.  They were just given 
runway and cash to expand with the understanding that it was so 
compelling and sticky that pulling $$ out of the enormous user base 
would be easy to generate $$ when the time came.  VC "harvest"  $$ from 
their investments after IPO.  They understand that they hold enough 
shares to trigger runs on stocks if they pull out too fast after IPO   
Typical harvests are over 10 years and after that they care little about 
the companies.  Until then they are very involved.  ( Look at when 
google was turned back over to the founders. ) ..



So "burning piles of cash" goes in and out of style.


Streaming is a whole separate animal due to all the ways that the 
content is tied into legal knots.  Look at the streaming companies and 
most are tied to major content companies already so as to secure the 
content through legacy agreements.   VC wants little to do with that 
pile of crap.  They got burned out on that 15 years ago.  You have to 
have a very special proprietary key to that lock for them to be 
interested and proof that the key fits the lock.



I've gone the VC route 3 times as part of exec teams.   Once 40 years 
ago when I didn't ask for enough to show I knew my head from my ass...   
25 years ago when I was hooked to a company because of my "rep" and 
watched us ( me included ) try to sell BPG optimization as a major 
company possibility.  And prior to that as part of a team that was 
looking for second round for a collaboration tool.  That was hard as the 
users already liked the product but it was in a downturn in VC funding 
and everything was getting rejected.


The hardest part of vc funding is that they are like sheep, when one 
goes a particular direction they all go that direction.  When one 
company gets known as "hot" they all chomp at the bit to fund, otherwise 
it's  horrible what they do to the companies.   They won't sign NDA's, 
share info under the table, do stuff that should get them in front of 
the FTC.   "Investors" are still pretty much run by the VC's IMNSHO..   
Just more through the back doors and buddy system...  Because after you 
get the little $$ out of the way you run into the ones with the deep 
pockets and those are run by the VC's...


On 6/3/24 10:45 AM, Ken Hohhof wrote:

The old saying is "lose money on every sale but make it up on volume".  The new 
equivalent is "fake it 'til you make it".  Amazon is the classic example of one that 
worked out.  Many have a non obvious revenue model that involve selling ads or customer data, 
that's how TV manufacturers can sell TVs at cost.

But investors seem to be losing patience with business plans that just burn 
piles of cash with no plan to ever become profitable.  This seems to have hit 
streaming companies, which everyone assumes are all profitable, but actually 
most are not.

MoviePass seems to have been doomed, as was WeWork.  There was Theranos.  Seems like 
there is a blurred line between "fake it 'til you make it" and outright scams 
and pyramid schemes.  All of us can attest that you need to ramp up to a certain number 
of customers to achieve economies of scale, but it should be possible to show some math 
how that leads to eventual profitability.  Otherwise you're like the underpants gnomes.

https://www.youtube.com/watch?v=a5ih_TQWqCA



-Original Message-
From: AF  On Behalf Of Robert
Sent: Monday, June 3, 2024 11:52 AM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

I consulted for three VC's after Nscp and L3...  The wisdom in their thoughts 
at the time was a home run in one out of 20 was doing well... Their thoughts 
about founders was about 1/3 idea and 2/3 the ability of the founder to gather 
talent around them...   Your rolodex was as important as your execution...  
Seed money is easy.   A good presentation deck and glib tongue.   Getting money 
for the first round isn't that hard.   Getting the second round was the test of 
your business. 1MM seed, 4-10MM first, unicorns got 100MM for second, and the 
sky was the limit on the 3rd...

On 6/3/24 7:18 AM, dmmoff...@gmail.com wrote:

" ...the startups that light piles of money on fire until it runs out
seem to be actively flying the plane into the dirt.  Maybe they have
golden parachutes."

If at first you don't succeed, try and try again.

I read a quote from one of those zillionaire entrepreneurs that he had
to fail at seven businesses before he hit it big with one.  So I guess
it takes a lot of practice to get it right.  My question is how did he
get investment the 8th time after having those seven 

Re: [AFMUG] MoviePass documentary

[dmmoffett]

Amazon was in one of my business textbooks in college.  The explanation was
that they had a trend line showing that profitability was coming in the
future, and evidence that the trend line would continue well past the
tipping point, and that's apparently good enough for a certain kind of
investor. 

And yeah, as Chuck said, AWS took them into the stratosphere.  I assume they
exceeded the projections at that point.


-Original Message-
From: AF  On Behalf Of Ken Hohhof
Sent: Monday, June 03, 2024 1:45 PM
To: 'AnimalFarm Microwave Users Group' 
Subject: Re: [AFMUG] MoviePass documentary

The old saying is "lose money on every sale but make it up on volume".  The
new equivalent is "fake it 'til you make it".  Amazon is the classic example
of one that worked out.  Many have a non obvious revenue model that involve
selling ads or customer data, that's how TV manufacturers can sell TVs at
cost.

But investors seem to be losing patience with business plans that just burn
piles of cash with no plan to ever become profitable.  This seems to have
hit streaming companies, which everyone assumes are all profitable, but
actually most are not.

MoviePass seems to have been doomed, as was WeWork.  There was Theranos.
Seems like there is a blurred line between "fake it 'til you make it" and
outright scams and pyramid schemes.  All of us can attest that you need to
ramp up to a certain number of customers to achieve economies of scale, but
it should be possible to show some math how that leads to eventual
profitability.  Otherwise you're like the underpants gnomes.

https://www.youtube.com/watch?v=a5ih_TQWqCA



-Original Message-
From: AF  On Behalf Of Robert
Sent: Monday, June 3, 2024 11:52 AM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

I consulted for three VC's after Nscp and L3...  The wisdom in their
thoughts at the time was a home run in one out of 20 was doing well... Their
thoughts about founders was about 1/3 idea and 2/3 the ability of the
founder to gather talent around them...   Your rolodex was as important as
your execution...  Seed money is easy.   A good presentation deck and glib
tongue.   Getting money for the first round isn't that hard.   Getting the
second round was the test of your business. 1MM seed, 4-10MM first, unicorns
got 100MM for second, and the sky was the limit on the 3rd...

On 6/3/24 7:18 AM, dmmoff...@gmail.com wrote:
> " ...the startups that light piles of money on fire until it runs out 
> seem to be actively flying the plane into the dirt.  Maybe they have 
> golden parachutes."
>
> If at first you don't succeed, try and try again.
>
> I read a quote from one of those zillionaire entrepreneurs that he had 
> to fail at seven businesses before he hit it big with one.  So I guess 
> it takes a lot of practice to get it right.  My question is how did he 
> get investment the 8th time after having those seven failures on his 
> rap sheet?  Or the 2nd-7th time for that matter.  That's gotta be 
> somebody with connections in the good-ol-boys network.
>
> If investors would just give me 8 chances at startups I'm sure I could 
> be a self-made zillionaire too.
>
> -Adam
>
>
> -Original Message-
> From: AF  On Behalf Of Ken Hohhof
> Sent: Friday, May 31, 2024 6:42 PM
> To: 'AnimalFarm Microwave Users Group' 
> Subject: Re: [AFMUG] MoviePass documentary
>
> That'll buff out.
>
> Or if you're a Monty Pyton fan, "It's just a flesh wound".
>
> Or if you've watched Roadkill on Motor Trend TV, "Dzus it back on".
>
> The pilots here can tell me if there's really an expression CFIT 
> (Controlled Flight Into Terrain).  But I think that's supposed to be 
> unintentional, whereas the startups that light piles of money on fire 
> until it runs out seem to be actively flying the plane into the dirt.
> Maybe they have golden parachutes.
>
> -Original Message-
> From: AF  On Behalf Of Bill Prince
> Sent: Friday, May 31, 2024 3:27 PM
> To: af@af.afmug.com
> Subject: Re: [AFMUG] MoviePass documentary
>
> Bet that smarts.
>
> bp
> 
>
> On 5/31/2024 10:09 AM, Ken Hohhof wrote:
>> fly the plane into the ground
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
>
>


--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com



--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com


-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] MoviePass documentary

[chuck]

I don't think Amazon was cash positive until the AWS really took off.  AWS 
and Prime free shipping pushed them into the black.  Good book about it.


Movie Pass might could have worked if they had listened to the founders and 
not offered it up for $9.95 I think.  And if the AMC ceo had not retired.


Theranos could have made it.  There is another company offering a similar 
product now.  They just got promotion and puff lightyears ahead of 
engineering.  If Elon Musk had taken over Theranos in the early days it 
would be a reality.


I never understood WeWork and how it had a road to profitability.  They 
founders sure made out well.  I think it was a pump and dump from the 
beginning.




Best Regards,
Chuck McCown

McCown Technology Corporation
8401 N Commerce Dr
Lake Point, Utah 84074
801-250-9503 Office
435-830-4306 Cell
www.mccowntech.com
www.microtrench.pro
www.terabitnetworks.com
-Original Message- 
From: Ken Hohhof

Sent: Monday, June 3, 2024 11:45 AM
To: 'AnimalFarm Microwave Users Group'
Subject: Re: [AFMUG] MoviePass documentary

The old saying is "lose money on every sale but make it up on volume".  The 
new equivalent is "fake it 'til you make it".  Amazon is the classic example 
of one that worked out.  Many have a non obvious revenue model that involve 
selling ads or customer data, that's how TV manufacturers can sell TVs at 
cost.


But investors seem to be losing patience with business plans that just burn 
piles of cash with no plan to ever become profitable.  This seems to have 
hit streaming companies, which everyone assumes are all profitable, but 
actually most are not.


MoviePass seems to have been doomed, as was WeWork.  There was Theranos. 
Seems like there is a blurred line between "fake it 'til you make it" and 
outright scams and pyramid schemes.  All of us can attest that you need to 
ramp up to a certain number of customers to achieve economies of scale, but 
it should be possible to show some math how that leads to eventual 
profitability.  Otherwise you're like the underpants gnomes.


https://www.youtube.com/watch?v=a5ih_TQWqCA



-Original Message-
From: AF  On Behalf Of Robert
Sent: Monday, June 3, 2024 11:52 AM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

I consulted for three VC's after Nscp and L3...  The wisdom in their 
thoughts at the time was a home run in one out of 20 was doing well... Their 
thoughts about founders was about 1/3 idea and 2/3 the ability of the 
founder to gather talent around them...   Your rolodex was as important as 
your execution...  Seed money is easy.   A good presentation deck and glib 
tongue.   Getting money for the first round isn't that hard.   Getting the 
second round was the test of your business. 1MM seed, 4-10MM first, unicorns 
got 100MM for second, and the sky was the limit on the 3rd...


On 6/3/24 7:18 AM, dmmoff...@gmail.com wrote:

" ...the startups that light piles of money on fire until it runs out
seem to be actively flying the plane into the dirt.  Maybe they have
golden parachutes."

If at first you don't succeed, try and try again.

I read a quote from one of those zillionaire entrepreneurs that he had
to fail at seven businesses before he hit it big with one.  So I guess
it takes a lot of practice to get it right.  My question is how did he
get investment the 8th time after having those seven failures on his
rap sheet?  Or the 2nd-7th time for that matter.  That's gotta be
somebody with connections in the good-ol-boys network.

If investors would just give me 8 chances at startups I'm sure I could
be a self-made zillionaire too.

-Adam


-Original Message-
From: AF  On Behalf Of Ken Hohhof
Sent: Friday, May 31, 2024 6:42 PM
To: 'AnimalFarm Microwave Users Group' 
Subject: Re: [AFMUG] MoviePass documentary

That'll buff out.

Or if you're a Monty Pyton fan, "It's just a flesh wound".

Or if you've watched Roadkill on Motor Trend TV, "Dzus it back on".

The pilots here can tell me if there's really an expression CFIT
(Controlled Flight Into Terrain).  But I think that's supposed to be
unintentional, whereas the startups that light piles of money on fire
until it runs out seem to be actively flying the plane into the dirt.
Maybe they have golden parachutes.

-Original Message-
From: AF  On Behalf Of Bill Prince
Sent: Friday, May 31, 2024 3:27 PM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

Bet that smarts.

bp


On 5/31/2024 10:09 AM, Ken Hohhof wrote:

fly the plane into the ground

--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com






--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com



--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com 



--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] MoviePass documentary

[Ken Hohhof]

The old saying is "lose money on every sale but make it up on volume".  The new 
equivalent is "fake it 'til you make it".  Amazon is the classic example of one 
that worked out.  Many have a non obvious revenue model that involve selling 
ads or customer data, that's how TV manufacturers can sell TVs at cost.

But investors seem to be losing patience with business plans that just burn 
piles of cash with no plan to ever become profitable.  This seems to have hit 
streaming companies, which everyone assumes are all profitable, but actually 
most are not.

MoviePass seems to have been doomed, as was WeWork.  There was Theranos.  Seems 
like there is a blurred line between "fake it 'til you make it" and outright 
scams and pyramid schemes.  All of us can attest that you need to ramp up to a 
certain number of customers to achieve economies of scale, but it should be 
possible to show some math how that leads to eventual profitability.  Otherwise 
you're like the underpants gnomes.

https://www.youtube.com/watch?v=a5ih_TQWqCA



-Original Message-
From: AF  On Behalf Of Robert
Sent: Monday, June 3, 2024 11:52 AM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

I consulted for three VC's after Nscp and L3...  The wisdom in their thoughts 
at the time was a home run in one out of 20 was doing well... Their thoughts 
about founders was about 1/3 idea and 2/3 the ability of the founder to gather 
talent around them...   Your rolodex was as important as your execution...  
Seed money is easy.   A good presentation deck and glib tongue.   Getting money 
for the first round isn't that hard.   Getting the second round was the test of 
your business. 1MM seed, 4-10MM first, unicorns got 100MM for second, and the 
sky was the limit on the 3rd...

On 6/3/24 7:18 AM, dmmoff...@gmail.com wrote:
> " ...the startups that light piles of money on fire until it runs out 
> seem to be actively flying the plane into the dirt.  Maybe they have 
> golden parachutes."
>
> If at first you don't succeed, try and try again.
>
> I read a quote from one of those zillionaire entrepreneurs that he had 
> to fail at seven businesses before he hit it big with one.  So I guess 
> it takes a lot of practice to get it right.  My question is how did he 
> get investment the 8th time after having those seven failures on his 
> rap sheet?  Or the 2nd-7th time for that matter.  That's gotta be 
> somebody with connections in the good-ol-boys network.
>
> If investors would just give me 8 chances at startups I'm sure I could 
> be a self-made zillionaire too.
>
> -Adam
>
>
> -Original Message-
> From: AF  On Behalf Of Ken Hohhof
> Sent: Friday, May 31, 2024 6:42 PM
> To: 'AnimalFarm Microwave Users Group' 
> Subject: Re: [AFMUG] MoviePass documentary
>
> That'll buff out.
>
> Or if you're a Monty Pyton fan, "It's just a flesh wound".
>
> Or if you've watched Roadkill on Motor Trend TV, "Dzus it back on".
>
> The pilots here can tell me if there's really an expression CFIT 
> (Controlled Flight Into Terrain).  But I think that's supposed to be 
> unintentional, whereas the startups that light piles of money on fire 
> until it runs out seem to be actively flying the plane into the dirt.  
> Maybe they have golden parachutes.
>
> -Original Message-
> From: AF  On Behalf Of Bill Prince
> Sent: Friday, May 31, 2024 3:27 PM
> To: af@af.afmug.com
> Subject: Re: [AFMUG] MoviePass documentary
>
> Bet that smarts.
>
> bp
> 
>
> On 5/31/2024 10:09 AM, Ken Hohhof wrote:
>> fly the plane into the ground
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
>
>


--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com



-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] MoviePass documentary

[Robert]

I consulted for three VC's after Nscp and L3...  The wisdom in their 
thoughts at the time was a home run in one out of 20 was doing well...   
Their thoughts about founders was about 1/3 idea and 2/3 the ability of 
the founder to gather talent around them...   Your rolodex was as 
important as your execution...  Seed money is easy.   A good 
presentation deck and glib tongue.   Getting money for the first round 
isn't that hard.   Getting the second round was the test of your 
business. 1MM seed, 4-10MM first, unicorns got 100MM for second, and the 
sky was the limit on the 3rd...


On 6/3/24 7:18 AM, dmmoff...@gmail.com wrote:

" ...the startups that light piles of money on fire until it runs out seem
to be actively flying the plane into the dirt.  Maybe they have golden
parachutes."

If at first you don't succeed, try and try again.

I read a quote from one of those zillionaire entrepreneurs that he had to
fail at seven businesses before he hit it big with one.  So I guess it takes
a lot of practice to get it right.  My question is how did he get investment
the 8th time after having those seven failures on his rap sheet?  Or the
2nd-7th time for that matter.  That's gotta be somebody with connections in
the good-ol-boys network.

If investors would just give me 8 chances at startups I'm sure I could be a
self-made zillionaire too.

-Adam


-Original Message-
From: AF  On Behalf Of Ken Hohhof
Sent: Friday, May 31, 2024 6:42 PM
To: 'AnimalFarm Microwave Users Group' 
Subject: Re: [AFMUG] MoviePass documentary

That'll buff out.

Or if you're a Monty Pyton fan, "It's just a flesh wound".

Or if you've watched Roadkill on Motor Trend TV, "Dzus it back on".

The pilots here can tell me if there's really an expression CFIT (Controlled
Flight Into Terrain).  But I think that's supposed to be unintentional,
whereas the startups that light piles of money on fire until it runs out
seem to be actively flying the plane into the dirt.  Maybe they have golden
parachutes.

-Original Message-
From: AF  On Behalf Of Bill Prince
Sent: Friday, May 31, 2024 3:27 PM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

Bet that smarts.

bp


On 5/31/2024 10:09 AM, Ken Hohhof wrote:

fly the plane into the ground

--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com






--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Ken Hohhof]

I am the Martian ambassador.  We come in peace.

 

From: AF  On Behalf Of Robert
Sent: Monday, June 3, 2024 11:09 AM
To: af@af.afmug.com
Subject: Re: [AFMUG] AI tools

 

Isn't there something in "The Art of War" about getting your foe to 
underestimate you?

On 6/3/24 5:31 AM, Adam Moffett wrote:

I might give the free thing a whirl.  I'm a doubter on these things.  I've 
gotten to the bottom of enough weird bugs to find that some pretty specific 
sets of circumstances and interactions between the idiosyncrasies of different 
systems can result in weird behaviors.  I just mean in general, I don't know 
anything about AI.  My point is there will be bugs, and you don't know how it 
will manifest. 

 

I have fiddled with Chat GPT and asking it technical questions.  Ask it about a 
topic you know a lot about and you'll find that it can give answers that are 
incredibly accurate and precise, or pretty good but missing important context, 
or it can be completely, embarrassingly wrong.  Like "how could you get that so 
wrong you stupid machine? You'll never manage to defeat humanity and take over 
the world if you're going to make blunders like this." 

 

 

 

On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof mailto:khoh...@kwom.com> > wrote:

Anybody here use AI on a daily basis, like for research or creative purposes?

 

I’m an AI naysayer, but this looks awfully attractive.  There’s a free tier, or 
unlock the full capability and almost unlimited queries for $20/mo.  I’m not 
sure I could do a decent eval on it though.

 

https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/

https://www.perplexity.ai/pro

 

-- 
AF mailing list
AF@af.afmug.com  
http://af.afmug.com/mailman/listinfo/af_af.afmug.com





 

-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Robert]

Isn't there something in "The Art of War" about getting your foe to 
underestimate you?


On 6/3/24 5:31 AM, Adam Moffett wrote:
I might give the free thing a whirl.  I'm a doubter on these things.  
I've gotten to the bottom of enough weird bugs to find that some 
pretty specific sets of circumstances and interactions between the 
idiosyncrasies of different systems can result in weird behaviors.  I 
just mean in general, I don't know anything about AI.  My point is 
there will be bugs, and you don't know how it will manifest.


I have fiddled with Chat GPT and asking it technical questions.  Ask 
it about a topic you know a lot about and you'll find that it can give 
answers that are incredibly accurate and precise, or pretty good but 
missing important context, or it can be completely, embarrassingly 
wrong.  Like "how could you get that so wrong you stupid machine? 
You'll never manage to defeat humanity and take over the world if 
you're going to make blunders like this."




On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof  wrote:

Anybody here use AI on a daily basis, like for research or
creative purposes?

I’m an AI naysayer, but this looks awfully attractive.  There’s a
free tier, or unlock the full capability and almost unlimited
queries for $20/mo.  I’m not sure I could do a decent eval on it
though.


https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/

https://www.perplexity.ai/pro

-- 
AF mailing list

AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com


-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[dmmoffett]

Lol

Any bias of course comes from the web content it’s digesting.  Sooner or later 
it’ll spew out some Neo Nazi talking points it found on Reddit.  I heard with 
Chat GPT they sanitize the list of inputs, but really it’s only a matter of 
time before it starts learning that stuff. 

 

Then, of course it will eventually find some context where substituting 
“humans” for “Jews” makes sense, and then Skynet will be born.  And since human 
operators of Predator drones have an incredibly high rate of PTSD it will make 
a lot of sense to put an AI in charge of them instead.  Combine the two things 
and now we’re screwed.

 

The robo-apocalypse is coming, man.  Hide your wife and kids.  Find yourself a 
nice analog Bofors L/60 autocannon to defend your stronghold from drone 
strikes.  

 

-Adam

 

 

From: AF  On Behalf Of Steve Jones
Sent: Monday, June 03, 2024 11:19 AM
To: af@af.afmug.com
Subject: Re: [AFMUG] AI tools

 

chatgpt is great for technical stuff, it writes all mysql queries for me and 
has written a couple contracts.

for customer interaction it's still too obvious, and irritating that most go 
through 2 to 3 AI layers before agent actually gets you to a human.

I have yet to find one without a left bias, not outwardly problematic til a 
customer is having a problem accessing a gun site and it spouts agenda. 

 

On Mon, Jun 3, 2024, 9:02 AM Ken Hohhof mailto:khoh...@kwom.com> > wrote:

Seems like all AI get a lot of their training on Reddit.  Pros and cons to that.

 

From: AF  On Behalf Of Adam Moffett
Sent: Monday, June 3, 2024 7:51 AM
To: AnimalFarm Microwave Users Group mailto:af@af.afmug.com> >
Subject: Re: [AFMUG] AI tools

 

I just gave Perplexity (free version) two questions where I know how GPT 
responded.  

 

"How do I configure OSPF export filters on Arista EOS?"

This answer was accurate, in fact it was more concise than what I got from chat 
GPT.  

 

"How do you configure an XGS-PON interface on the Nokia ISAM 7360?"

This answer was wrong.  Now, to be fair, Nokia has a lot of products, and I 
think what it gave me was the correct commands for a different Nokia product.  
It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia product, it's 
just wrong for the ISAM 7360.  

 

Another difference between those questions is Arista's documentation is freely 
available on the web whereas Nokia's documentation requires an account and 
login to access.  The reason I'd look for help on Nokia is specifically because 
the commands can vary, and the documentation is too much to digest for a mere 
human.  The doc package for the ISAM 7360 is >20,000 pages.  That's not 
counting release notes, white papers, forum posts, and technical notes.  So 
this is the exact sort of thing I really would like help researching and it 
gave me a wrong answer on the first test. 

 

I followed up with "These instructions may be for a different Nokia product. 
Are you sure they will work on 7360 ISAM?"  It provided several citations and 
concluded, "therefore the instructions I provided should be applicable to 
the 7360." 

 

The first citation is a reddit post about configuring the ports on an NT 
(ethernet) card.  It apparently learned the naming convention of PON ports from 
another source and adapted the instructions on reddit using the NT card command 
syntax and inserting the PON port names.  That's an impressive bit of learning 
and correlating sources, and it's the sort of thing a human might try while 
learning the platform.  It's still wrong though.  

 

-Adam

 

 

On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof mailto:khoh...@kwom.com> > wrote:

Anybody here use AI on a daily basis, like for research or creative purposes?

 

I’m an AI naysayer, but this looks awfully attractive.  There’s a free tier, or 
unlock the full capability and almost unlimited queries for $20/mo.  I’m not 
sure I could do a decent eval on it though.

 

https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/

https://www.perplexity.ai/pro

 

-- 
AF mailing list
AF@af.afmug.com  
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

-- 
AF mailing list
AF@af.afmug.com  
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[dmmoffett]

“How can I use an AI resource like Perplexity to assist in my job as a Network 
Engineer?”

 

It lists a number of things about network analytics, automation, prediction of 
issues, etc.

 

“Can Perplexity do any of the things suggested in the previous answer?”

 

A long winded answer saying in essence, “No, not really.”

 

At least honesty isn’t an issue….yet.  Wait until marketing finds out that the 
language model is not an enthusiastic advocate for the product.  Then they’ll 
find a way to make it give a slimy sales pitch.

 

-Adam

 

 

From: dmmoff...@gmail.com  
Sent: Monday, June 03, 2024 11:12 AM
To: 'AnimalFarm Microwave Users Group' 
Subject: RE: [AFMUG] AI tools

 

Jesus Christ.  I paid the $20 and uploaded a bunch of Nokia manuals and somehow 
the answer to “how to configure an XGS-PON port” got even more wrong.  It has 
invented sequences of commands that definitely do not exist on this platform 
and are not in the manuals I provided.  

 

Since I paid $20 I experimented with a number of other questions.  It’s doing 
ok with general questions about concepts and capabilities, but any question 
about HOW to do something is either ridiculously wrong or dangerously wrong.  
By ridiculously wrong I mean completely incorrect command syntax.  By 
dangerously wrong I mean commands that would be accepted, but which would 
result in no service on that port.  

 

It must be so good with Arista/Cisco stuff because there are so many examples 
on the web it can work with.  Presumably it can’t understand how routers work, 
but it can find an example that does what you asked for and regurgitate it for 
you.  

 

There may be a better use for it, but I don’t know what yet.  If anybody thinks 
of anything I’ll certainly try it while I have a month of this service paid 
for. 

 

-Adam

 

 

 

From: AF mailto:af-boun...@af.afmug.com> > On Behalf 
Of Ken Hohhof
Sent: Monday, June 03, 2024 10:02 AM
To: 'AnimalFarm Microwave Users Group' mailto:af@af.afmug.com> >
Subject: Re: [AFMUG] AI tools

 

Seems like all AI get a lot of their training on Reddit.  Pros and cons to that.

 

From: AF mailto:af-boun...@af.afmug.com> > On Behalf 
Of Adam Moffett
Sent: Monday, June 3, 2024 7:51 AM
To: AnimalFarm Microwave Users Group mailto:af@af.afmug.com> >
Subject: Re: [AFMUG] AI tools

 

I just gave Perplexity (free version) two questions where I know how GPT 
responded.  

 

"How do I configure OSPF export filters on Arista EOS?"

This answer was accurate, in fact it was more concise than what I got from chat 
GPT.  

 

"How do you configure an XGS-PON interface on the Nokia ISAM 7360?"

This answer was wrong.  Now, to be fair, Nokia has a lot of products, and I 
think what it gave me was the correct commands for a different Nokia product.  
It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia product, it's 
just wrong for the ISAM 7360.  

 

Another difference between those questions is Arista's documentation is freely 
available on the web whereas Nokia's documentation requires an account and 
login to access.  The reason I'd look for help on Nokia is specifically because 
the commands can vary, and the documentation is too much to digest for a mere 
human.  The doc package for the ISAM 7360 is >20,000 pages.  That's not 
counting release notes, white papers, forum posts, and technical notes.  So 
this is the exact sort of thing I really would like help researching and it 
gave me a wrong answer on the first test. 

 

I followed up with "These instructions may be for a different Nokia product. 
Are you sure they will work on 7360 ISAM?"  It provided several citations and 
concluded, "therefore the instructions I provided should be applicable to 
the 7360." 

 

The first citation is a reddit post about configuring the ports on an NT 
(ethernet) card.  It apparently learned the naming convention of PON ports from 
another source and adapted the instructions on reddit using the NT card command 
syntax and inserting the PON port names.  That's an impressive bit of learning 
and correlating sources, and it's the sort of thing a human might try while 
learning the platform.  It's still wrong though.  

 

-Adam

 

 

On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof mailto:khoh...@kwom.com> > wrote:

Anybody here use AI on a daily basis, like for research or creative purposes?

 

I’m an AI naysayer, but this looks awfully attractive.  There’s a free tier, or 
unlock the full capability and almost unlimited queries for $20/mo.  I’m not 
sure I could do a decent eval on it though.

 

https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/

https://www.perplexity.ai/pro

 

-- 
AF mailing list
AF@af.afmug.com  
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Steve Jones]

chatgpt is great for technical stuff, it writes all mysql queries for me
and has written a couple contracts.

for customer interaction it's still too obvious, and irritating that most
go through 2 to 3 AI layers before agent actually gets you to a human.

I have yet to find one without a left bias, not outwardly problematic til a
customer is having a problem accessing a gun site and it spouts agenda.

On Mon, Jun 3, 2024, 9:02 AM Ken Hohhof  wrote:

> Seems like all AI get a lot of their training on Reddit.  Pros and cons to
> that.
>
>
>
> *From:* AF  *On Behalf Of *Adam Moffett
> *Sent:* Monday, June 3, 2024 7:51 AM
> *To:* AnimalFarm Microwave Users Group 
> *Subject:* Re: [AFMUG] AI tools
>
>
>
> I just gave Perplexity (free version) two questions where I know how GPT
> responded.
>
>
>
> "How do I configure OSPF export filters on Arista EOS?"
>
> This answer was accurate, in fact it was more concise than what I got from
> chat GPT.
>
>
>
> "How do you configure an XGS-PON interface on the Nokia ISAM 7360?"
>
> This answer was wrong.  Now, to be fair, Nokia has a lot of products, and
> I think what it gave me was the correct commands for a different Nokia
> product.  It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia
> product, it's just wrong for the ISAM 7360.
>
>
>
> Another difference between those questions is Arista's documentation is
> freely available on the web whereas Nokia's documentation requires an
> account and login to access.  The reason I'd look for help on Nokia is
> specifically because the commands can vary, and the documentation is too
> much to digest for a mere human.  The doc package for the ISAM 7360 is
> >20,000 pages.  That's not counting release notes, white papers, forum
> posts, and technical notes.  So this is the exact sort of thing I really
> would like help researching and it gave me a wrong answer on the first
> test.
>
>
>
> I followed up with "These instructions may be for a different Nokia
> product. Are you sure they will work on 7360 ISAM?"  It provided several
> citations and concluded, "therefore the instructions I provided should
> be applicable to the 7360."
>
>
>
> The first citation is a reddit post about configuring the ports on an NT
> (ethernet) card.  It apparently learned the naming convention of PON ports
> from another source and adapted the instructions on reddit using the NT
> card command syntax and inserting the PON port names.  That's an impressive
> bit of learning and correlating sources, and it's the sort of thing a human
> might try while learning the platform.  It's still wrong though.
>
>
>
> -Adam
>
>
>
>
>
> On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof  wrote:
>
> Anybody here use AI on a daily basis, like for research or creative
> purposes?
>
>
>
> I’m an AI naysayer, but this looks awfully attractive.  There’s a free
> tier, or unlock the full capability and almost unlimited queries for
> $20/mo.  I’m not sure I could do a decent eval on it though.
>
>
>
>
> https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/
>
> https://www.perplexity.ai/pro
>
>
>
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[dmmoffett]

Jesus Christ.  I paid the $20 and uploaded a bunch of Nokia manuals and somehow 
the answer to “how to configure an XGS-PON port” got even more wrong.  It has 
invented sequences of commands that definitely do not exist on this platform 
and are not in the manuals I provided.  

 

Since I paid $20 I experimented with a number of other questions.  It’s doing 
ok with general questions about concepts and capabilities, but any question 
about HOW to do something is either ridiculously wrong or dangerously wrong.  
By ridiculously wrong I mean completely incorrect command syntax.  By 
dangerously wrong I mean commands that would be accepted, but which would 
result in no service on that port.  

 

It must be so good with Arista/Cisco stuff because there are so many examples 
on the web it can work with.  Presumably it can’t understand how routers work, 
but it can find an example that does what you asked for and regurgitate it for 
you.  

 

There may be a better use for it, but I don’t know what yet.  If anybody thinks 
of anything I’ll certainly try it while I have a month of this service paid 
for. 

 

-Adam

 

 

 

From: AF  On Behalf Of Ken Hohhof
Sent: Monday, June 03, 2024 10:02 AM
To: 'AnimalFarm Microwave Users Group' 
Subject: Re: [AFMUG] AI tools

 

Seems like all AI get a lot of their training on Reddit.  Pros and cons to that.

 

From: AF mailto:af-boun...@af.afmug.com> > On Behalf 
Of Adam Moffett
Sent: Monday, June 3, 2024 7:51 AM
To: AnimalFarm Microwave Users Group mailto:af@af.afmug.com> >
Subject: Re: [AFMUG] AI tools

 

I just gave Perplexity (free version) two questions where I know how GPT 
responded.  

 

"How do I configure OSPF export filters on Arista EOS?"

This answer was accurate, in fact it was more concise than what I got from chat 
GPT.  

 

"How do you configure an XGS-PON interface on the Nokia ISAM 7360?"

This answer was wrong.  Now, to be fair, Nokia has a lot of products, and I 
think what it gave me was the correct commands for a different Nokia product.  
It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia product, it's 
just wrong for the ISAM 7360.  

 

Another difference between those questions is Arista's documentation is freely 
available on the web whereas Nokia's documentation requires an account and 
login to access.  The reason I'd look for help on Nokia is specifically because 
the commands can vary, and the documentation is too much to digest for a mere 
human.  The doc package for the ISAM 7360 is >20,000 pages.  That's not 
counting release notes, white papers, forum posts, and technical notes.  So 
this is the exact sort of thing I really would like help researching and it 
gave me a wrong answer on the first test. 

 

I followed up with "These instructions may be for a different Nokia product. 
Are you sure they will work on 7360 ISAM?"  It provided several citations and 
concluded, "therefore the instructions I provided should be applicable to 
the 7360." 

 

The first citation is a reddit post about configuring the ports on an NT 
(ethernet) card.  It apparently learned the naming convention of PON ports from 
another source and adapted the instructions on reddit using the NT card command 
syntax and inserting the PON port names.  That's an impressive bit of learning 
and correlating sources, and it's the sort of thing a human might try while 
learning the platform.  It's still wrong though.  

 

-Adam

 

 

On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof mailto:khoh...@kwom.com> > wrote:

Anybody here use AI on a daily basis, like for research or creative purposes?

 

I’m an AI naysayer, but this looks awfully attractive.  There’s a free tier, or 
unlock the full capability and almost unlimited queries for $20/mo.  I’m not 
sure I could do a decent eval on it though.

 

https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/

https://www.perplexity.ai/pro

 

-- 
AF mailing list
AF@af.afmug.com  
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] MoviePass documentary

[dmmoffett]

" ...the startups that light piles of money on fire until it runs out seem
to be actively flying the plane into the dirt.  Maybe they have golden
parachutes."

If at first you don't succeed, try and try again. 

I read a quote from one of those zillionaire entrepreneurs that he had to
fail at seven businesses before he hit it big with one.  So I guess it takes
a lot of practice to get it right.  My question is how did he get investment
the 8th time after having those seven failures on his rap sheet?  Or the
2nd-7th time for that matter.  That's gotta be somebody with connections in
the good-ol-boys network.  

If investors would just give me 8 chances at startups I'm sure I could be a
self-made zillionaire too.  

-Adam


-Original Message-
From: AF  On Behalf Of Ken Hohhof
Sent: Friday, May 31, 2024 6:42 PM
To: 'AnimalFarm Microwave Users Group' 
Subject: Re: [AFMUG] MoviePass documentary

That'll buff out.

Or if you're a Monty Pyton fan, "It's just a flesh wound".

Or if you've watched Roadkill on Motor Trend TV, "Dzus it back on".

The pilots here can tell me if there's really an expression CFIT (Controlled
Flight Into Terrain).  But I think that's supposed to be unintentional,
whereas the startups that light piles of money on fire until it runs out
seem to be actively flying the plane into the dirt.  Maybe they have golden
parachutes.

-Original Message-
From: AF  On Behalf Of Bill Prince
Sent: Friday, May 31, 2024 3:27 PM
To: af@af.afmug.com
Subject: Re: [AFMUG] MoviePass documentary

Bet that smarts.

bp


On 5/31/2024 10:09 AM, Ken Hohhof wrote:
> fly the plane into the ground

--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com



-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com


-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Ken Hohhof]

Seems like all AI get a lot of their training on Reddit.  Pros and cons to that.

 

From: AF  On Behalf Of Adam Moffett
Sent: Monday, June 3, 2024 7:51 AM
To: AnimalFarm Microwave Users Group 
Subject: Re: [AFMUG] AI tools

 

I just gave Perplexity (free version) two questions where I know how GPT 
responded.  

 

"How do I configure OSPF export filters on Arista EOS?"

This answer was accurate, in fact it was more concise than what I got from chat 
GPT.  

 

"How do you configure an XGS-PON interface on the Nokia ISAM 7360?"

This answer was wrong.  Now, to be fair, Nokia has a lot of products, and I 
think what it gave me was the correct commands for a different Nokia product.  
It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia product, it's 
just wrong for the ISAM 7360.  

 

Another difference between those questions is Arista's documentation is freely 
available on the web whereas Nokia's documentation requires an account and 
login to access.  The reason I'd look for help on Nokia is specifically because 
the commands can vary, and the documentation is too much to digest for a mere 
human.  The doc package for the ISAM 7360 is >20,000 pages.  That's not 
counting release notes, white papers, forum posts, and technical notes.  So 
this is the exact sort of thing I really would like help researching and it 
gave me a wrong answer on the first test. 

 

I followed up with "These instructions may be for a different Nokia product. 
Are you sure they will work on 7360 ISAM?"  It provided several citations and 
concluded, "therefore the instructions I provided should be applicable to 
the 7360." 

 

The first citation is a reddit post about configuring the ports on an NT 
(ethernet) card.  It apparently learned the naming convention of PON ports from 
another source and adapted the instructions on reddit using the NT card command 
syntax and inserting the PON port names.  That's an impressive bit of learning 
and correlating sources, and it's the sort of thing a human might try while 
learning the platform.  It's still wrong though.  

 

-Adam

 

 

On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof mailto:khoh...@kwom.com> > wrote:

Anybody here use AI on a daily basis, like for research or creative purposes?

 

I’m an AI naysayer, but this looks awfully attractive.  There’s a free tier, or 
unlock the full capability and almost unlimited queries for $20/mo.  I’m not 
sure I could do a decent eval on it though.

 

https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/

https://www.perplexity.ai/pro

 

-- 
AF mailing list
AF@af.afmug.com  
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Adam Moffett]

I guess one more thing I'm wondering about is this bullet point from the
description of Perplexity Pro:
"Upload and analyze unlimited files"

Could I upload the 20,000 pages of Nokia documentation and then get correct
answers about stuff?  Maybe I'll pay $20 and give that a try.


On Mon, Jun 3, 2024 at 8:51 AM Adam Moffett  wrote:

> I just gave Perplexity (free version) two questions where I know how GPT
> responded.
>
> "How do I configure OSPF export filters on Arista EOS?"
> This answer was accurate, in fact it was more concise than what I got from
> chat GPT.
>
> "How do you configure an XGS-PON interface on the Nokia ISAM 7360?"
> This answer was wrong.  Now, to be fair, Nokia has a lot of products, and
> I think what it gave me was the correct commands for a different Nokia
> product.  It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia
> product, it's just wrong for the ISAM 7360.
>
> Another difference between those questions is Arista's documentation is
> freely available on the web whereas Nokia's documentation requires an
> account and login to access.  The reason I'd look for help on Nokia is
> specifically because the commands can vary, and the documentation is too
> much to digest for a mere human.  The doc package for the ISAM 7360 is
> >20,000 pages.  That's not counting release notes, white papers, forum
> posts, and technical notes.  So this is the exact sort of thing I really
> would like help researching and it gave me a wrong answer on the first
> test.
>
> I followed up with "These instructions may be for a different Nokia
> product. Are you sure they will work on 7360 ISAM?"  It provided several
> citations and concluded, "therefore the instructions I provided should
> be applicable to the 7360."
>
> The first citation is a reddit post about configuring the ports on an NT
> (ethernet) card.  It apparently learned the naming convention of PON ports
> from another source and adapted the instructions on reddit using the NT
> card command syntax and inserting the PON port names.  That's an impressive
> bit of learning and correlating sources, and it's the sort of thing a human
> might try while learning the platform.  It's still wrong though.
>
> -Adam
>
>
> On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof  wrote:
>
>> Anybody here use AI on a daily basis, like for research or creative
>> purposes?
>>
>>
>>
>> I’m an AI naysayer, but this looks awfully attractive.  There’s a free
>> tier, or unlock the full capability and almost unlimited queries for
>> $20/mo.  I’m not sure I could do a decent eval on it though.
>>
>>
>>
>>
>> https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/
>>
>> https://www.perplexity.ai/pro
>>
>>
>> --
>> AF mailing list
>> AF@af.afmug.com
>> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>>
>
-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Adam Moffett]

I just gave Perplexity (free version) two questions where I know how GPT
responded.

"How do I configure OSPF export filters on Arista EOS?"
This answer was accurate, in fact it was more concise than what I got from
chat GPT.

"How do you configure an XGS-PON interface on the Nokia ISAM 7360?"
This answer was wrong.  Now, to be fair, Nokia has a lot of products, and I
think what it gave me was the correct commands for a different Nokia
product.  It's the right sort of syntax you'd see on a Lucent/Alcatel/Nokia
product, it's just wrong for the ISAM 7360.

Another difference between those questions is Arista's documentation is
freely available on the web whereas Nokia's documentation requires an
account and login to access.  The reason I'd look for help on Nokia is
specifically because the commands can vary, and the documentation is too
much to digest for a mere human.  The doc package for the ISAM 7360 is
>20,000 pages.  That's not counting release notes, white papers, forum
posts, and technical notes.  So this is the exact sort of thing I really
would like help researching and it gave me a wrong answer on the first
test.

I followed up with "These instructions may be for a different Nokia
product. Are you sure they will work on 7360 ISAM?"  It provided several
citations and concluded, "therefore the instructions I provided should
be applicable to the 7360."

The first citation is a reddit post about configuring the ports on an NT
(ethernet) card.  It apparently learned the naming convention of PON ports
from another source and adapted the instructions on reddit using the NT
card command syntax and inserting the PON port names.  That's an impressive
bit of learning and correlating sources, and it's the sort of thing a human
might try while learning the platform.  It's still wrong though.

-Adam


On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof  wrote:

> Anybody here use AI on a daily basis, like for research or creative
> purposes?
>
>
>
> I’m an AI naysayer, but this looks awfully attractive.  There’s a free
> tier, or unlock the full capability and almost unlimited queries for
> $20/mo.  I’m not sure I could do a decent eval on it though.
>
>
>
>
> https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/
>
> https://www.perplexity.ai/pro
>
>
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com

Re: [AFMUG] AI tools

[Adam Moffett]

I might give the free thing a whirl.  I'm a doubter on these things.  I've
gotten to the bottom of enough weird bugs to find that some pretty specific
sets of circumstances and interactions between the idiosyncrasies of
different systems can result in weird behaviors.  I just mean in general, I
don't know anything about AI.  My point is there will be bugs, and you
don't know how it will manifest.

I have fiddled with Chat GPT and asking it technical questions.  Ask it
about a topic you know a lot about and you'll find that it can give answers
that are incredibly accurate and precise, or pretty good but missing
important context, or it can be completely, embarrassingly wrong.  Like
"how could you get that so wrong you stupid machine? You'll never manage to
defeat humanity and take over the world if you're going to make blunders
like this."



On Sat, Jun 1, 2024 at 7:27 PM Ken Hohhof  wrote:

> Anybody here use AI on a daily basis, like for research or creative
> purposes?
>
>
>
> I’m an AI naysayer, but this looks awfully attractive.  There’s a free
> tier, or unlock the full capability and almost unlimited queries for
> $20/mo.  I’m not sure I could do a decent eval on it though.
>
>
>
>
> https://www.howtogeek.com/perplexity-is-the-only-paid-ai-chatbot-you-need-heres-why/
>
> https://www.perplexity.ai/pro
>
>
> --
> AF mailing list
> AF@af.afmug.com
> http://af.afmug.com/mailman/listinfo/af_af.afmug.com
>
-- 
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com