*Critical* Artificial Intelligence (AI Stock) DANGER.

wow this was not of my bingo card

Bloomberg just moments ago released an

article bagging on artificial

intelligence's progress at all of them

almost open AI which obviously works

with Microsoft and Powers u a co-pilot

Google anthropic which Amazon keeps

throwing money into they talk about

Apple backing out of a deal they talk

about the uh artificial general

intelligence bubble bursting uh they

talk about failures of new models and

some of the cons strengths that are

slowing progress substantially this is

really interesting and in my opinion has

some really big implications and we

should talk about what some of these

challenges are so we kind of know okay

like what does this mean for us is this

really a big deal is this bad or is this

just some more clickbait let's take a

look at this together open AI Google and

anthropic struggle to build more

advanced AI three of the leading AI

companies are seeing de diminishing

returns from their costly efforts to

develop newer

models now this is interesting to me

because they specifically say with open

AI for example that they're working on

this model known as Orin uh orin's

apparently different from open AI 01

preview you can use now or 40 just keep

this in mind these are all slightly

different supposed to be improvements on

each other like 01 has sort of a

thinking model uh and when I say

thinking model basically if you give it

a generic prompt like here's just a

generic prompt I took a screenshot of

help me write a business plan for

eliminate stand blah blah blah right it

it thinks first basically what it's

really doing is breaking apart your

request making a better prompt and then

responding to its prompt which is good

because frankly it takes the whole

prompting phase away from you which is

great but is it really a huge leap

forward maybe not uh but Orin is

supposed to be that huge leap forward

but apparently according to two people

familiar with the matter who spoke on a

condition of anonymity it's always those

Anonymous leakers as of the late summer

Orin fell short when trying to answer

coding questions that it had not been

trained on which basically means it's

just trying to regurgitate stuff it's

already seen to you which is a little

problematic because the point of

artificial intelligence is to get us to

the point where it can respond to things

that it has not seen before you know

like I like this analogy of

if let's just say everybody tomorrow

wants to look up uh the history of John

F Kennedy's

life how often do you actually have to

run that compute through an Nvidia

h100 well the answer is once you process

it once and then you've basically

created or just stolen the Wikipedia

article and then you just from cash feed

that response to all of the people who

ask the same questions I mean how many

different questions could there possibly

be you know are you know is dietary fat

going to make me fat you know what are

bowling balls made of what's inside of a

bowling ball yeah tell me about JFK like

all these things could just be cashed

encyclopedic responses basically but

when it becomes a problem is when newer

models

fail to to help us solve problems that

they have not seen yet before because

that is kind of what we want them to do

that's what you need in order to get to

artificial general intelligence to be

humanlike and they're saying Orin so far

is not considered as big of a step up as

open ai's existing models were from

prior models so basically they're saying

hey look we had a big step up from 3.5

where it was like oh my gosh look this

is an explosion if this goes on forever

you know we're going to be on the moon

and really Orin is sort of like H maybe

it's you know tapering off a little and

so this is basically the definition of

diminishing turns I've actually

regularly on the channel talked about

how I think artificial intelligence is

going to follow this stair step pattern

where you'll have really big moments and

really big momentary breakthroughs for

example you know maybe uh uh 3 to all of

a sudden that 3.5 model maybe you could

even give 40 that stair step but what

happens is these bases become

substantially longer and longer and we

might not have another breakthrough a

huge breakthrough until 2030 which I

recognize sounds depressing I'm just

saying as an example it takes longer for

those next real breakthroughs to

actually come and what that actually

looks like uh is this

exponential uh cost increase compared to

uh the

logarithmic decay of progress now that

was like a really big word sentence

let's put it just basically I'm going to

redraw what Nvidia said so Nvidia said

which is true for Tesla and all these

other companies that your costs to train

and develop AI

Skyrocket uh more Talent more chips more

processing but at the same time as that

happens your progress actually looks a

whole lot more like this where you grow

a lot at the beginning and then you

taper out so you're spending a whole lot

more money to make a whole lot less

progress you know quick example if let's

just to be extreme let's say you know

your first $10 million of spend got you

to here you know which is 50% progress

and now all of a sudden you're spending

uh 10 oops now all of a sudden you're

spending 10

billion to make. 1% progress you know it

gets a little depressing in terms of how

fast you're actually getting anywhere

right but let's keep going with the

article because it's very interesting so

uh opening ey by the way just finished

their uh one of their initial training

rounds in September for Orin and they

fundraised for that round they actually

fundraised uh $6.6 billion this is

considered the largest Venture Capital

round ever led by Jared Kushner you know

Ivanka Trump's husband works with uh

Donald Trump they're all connected it

values open AI at $157 billion and for

an unknown reason we actually saw Apple

back out of the deal don't know if it

was because of valuation or apple

doesn't like the progress or what it is

uh but apparently now you're seeing some

stumbling blocks and these diminishing

returns which you're also seeing at

Google with Gemini not living up to

internal expectations according to three

people with knowledge of the matter

anthropic has not seen a timetable for

its 3.5 Opus release and they've

actually removed some references to

coming soon companies are also finding a

difficult to find untapped sources of

human-made training data and so you

don't actually have new data to train

with so this is problematic and now open

AI is putting Orin through a human

feedback process which in my opinion

brings me back to like the days Tesla

was talking about oh yeah we have all

these humans sitting in a room and they

they sit around and they label things

like stop signs and traffic lights and

and deal with Edge case scenarios which

I think a lot of AI companies don't want

you to realize iizz that there is a lot

of human intervention in AI training and

when AI hasn't seen a scenario before

it's kind of like

hm yeah I'm going to make a lot of

mistakes before I actually get the right

answer here which isn't great much of

the tech industry has B on the so-called

scaling laws that say more computing

power data and larger models will

inevitably pave the way for greater

leaps forward in the power of AI see

I've questioned that because that's

basically this sort of FO moment

mentality of oh things are just going to

go up forever and and then because

there's more scale with these data

centers you'll actually progress even

faster I don't think so I've been very

much the believer that you're seeing

stair steps of improvement the 3.5

moment you know maybe the 40 moment but

then what happens between those

different stairs is it actually takes

longer and longer and longer uh to

improve and you end up with this

logarithmic curve until you truly get

another leap forward which you know I

think is probably like a 2030 the age

bubble is bursting a little bit says a

chief ethic scientists at AI startup

hugging face oops uh it's become clear

she said that different training

approaches may need may be needed to

make AI models work it's basically way

saying like we can't just keep trying to

train with the same things we've been

doing it's just not getting us any

further which is

unfortunate Deep Mind from Google says

we're uh you know pleased with the

progress that we're making anthropic and

open AI declined to comment people call

them scaling laws it's a misnomer

they're not laws of the universe they're

empirical regularities I'm going to bet

in favor of them continuing but I'm not

certain of that it's also somebody in

the AI industry a lot of things that

could derail the process like we could

run out of data say some plateauing

performance is basically in line with my

uh stair stepper thesis not enough uh

data to create products uh uh now you

could spit out clever essays and poems

but are you going to get smarter than a

noble laurate training it off Wikipedia

posts and YouTube captions probably not

uh by the way I just want to do a quick

shout out if you haven't gotten free

stocks yet with an amazing trading

platform Weeble make sure to go to

metkevin.com

Weeble metkevin it's also in the

description you get some free stocks

pretty amazing and they've got a great

trading platform they keep updating the

software too so if you haven't checked

out Weeble yet do consider that

metkevin.com Weeble I use them almost

every day when I trade all right open a

in particular has ink deals with

Publishers to fill some of the need for

high quality data and also adapt to

Growing legal pressure from Publishers

and artists over data used to generate

or build these AI products some tech

companies are also hiring people with

graduate degrees that can label data

related to their own subject expertise

so now you're basically hiring humans

and phds to write encyclopedia articles

for you you're throwing money at people

I should spell that correctly but

whatever tech companies are also turning

into syn turning to synthetic data such

as computer generated images or text

meant to mimic content put together by

uh real people but there are limits to

that we can generate quality

synthetically yet we struggle to get

unique high quality data sets you know

in my opinion this is a little bit kind

of like the problem that Tesla runs into

with FSD is the you really need you

don't need more data of uh you know the

Tesla sitting in the left lane on the

highway or really anywhere on the

highway you need millions of more Edge

case scenario interactions and and that

the AI can see over and over and over

again uh so far it feels like you

regularly have this lately with the

latest FSD Updates this one step forward

one step back it's a little depressing

where it's kind of like oh okay like

oh that's fixed now oh but now that's

broken kind of sucks but anyway uh it's

still great like don't get me wrong like

those little setbacks suck but FSD in

general I love anyway still AI companies

cons continue to pursue a more is better

playbook in their quest to buil products

that level human intelligence increasing

the amount of computer spend spending

hundreds of millions of dollars to train

bleeding edge AI models as cost rise so

do the expectations for each model

people are get getting very excited uh

but anthropic has uh yet to uh you know

provide a date for when its Opus is

coming out uh you know they've removed

references that 3.5 Opus will be coming

later this year challenges to develop

3.5 Opus according to two other people

familiar with the matter uh Opus uh you

know some of the marketing in terms of

timetable was removed fine uh Google's

still struggling to restore the ability

to generate images after it gave us a

bunch of woke stuff more recently open

AI rolled out a a preview version of uh

01 that spends extra time Computing

that's the picture of uh you know what

what I uh what I showed you uh you know

incremental Improvement I would say they

have a limit in terms of how many times

you can actually use it and the downside

of that 01 is it doesn't actually give

you current results whereas like grock

does and 40 at GPT does also talking

about very good releases coming later

this year but that won't include J GPT

or GPT 5 yet uh and then now what you're

getting is like Google anthropic open AI

is now shifting attention from the size

of the models to newer use cases

including a crop of AI tools called

agents that can book flights or send

emails on a user's behalf we will have

better and better models but I think the

thing that will feel like the next big

breakthrough will be agents oo um I mean

AI is already making big leaps and

customer service I do actually think

that like you're going to have a massive

cycle of unemployment at some point in

the future because of how much

artificial intelligence uh is being used

I mean even if you just take 10% of all

customer service workers and replace

them with AI uh or you know make

existing people more productive with AI

it's kind of a GameChanger like you know

I was on a phone for like a stupid very

minor problem with QuickBooks

for 17 minutes and at the end of the 17

minutes the person told me yeah you're

just going to have to call back like

everything looks good but you're just

going to have to call back to actually

make the final change uh when the other

team is in they're not in right now and

you're just going to have to go through

everything that you explain now again

and then because they'll be there they

can help you and I'm like you know like

a good AI could have told me that in the

first 20 seconds like for this problem

instead of spending 17 minutes on the

phone now you should call back during

these

hours

bro it's just crazy so there's there are

definitely a lot of efficiency

improvements uh that can be made at all

walks of the world but uh yeah in terms

of these these moonshots and and then we

now we got to talk implications for a

moment it's worth considering there's

some limitations here

but then that also makes me concerned

because you know what is this going to

mean for a company like Nvidia you know

Wall Street is expecting Nvidia to

basically you know it's been doubling

its uh its revenues because you know

they've exploded oh my gosh sheap you

like graphic cards are really good for

for AI this is fantastic bye bye bye you

know so their revenue doubled basically

in 2024 it's expected to double again uh

by January 2025 so that's 23 double 24

double basically they have a weird

calendar year uh then it's expected to

grow by about 50% in 2026 down to 20% in

2017 and 28 uh and 16% in

2029 but what happens if all of a sudden

those growth rates go to zero uh you

know and you basically you just maintain

these high levels because think about it

you got to resell new chips this is why

Nvidia is trying to get on a one-year

refresh cycle yeah yeah yeah now buy

Blackwell yeah now buy the buy Blackwell

2 buy Blackwell 3 okay let's let's skip

a number here like the iPhone's done you

know we're going to skip iPhone 2 we're

going to go to I don't know Blackwell

five even though we didn't have a four

you know whatever right uh it and it's

all designed to sort of Market more

sales because and if you if people stop

buying the new product because the

current product is so good then your

growth rate is going to plummet at

Nvidia and I mean don't get me wrong I

love Nvidia but it's about 100 $50 per

share company divided by $284 you're

trading for about 53 times earnings your

expected growth over the years going

forward here the next four years

is uh about 25% per year which puts you

at about a two Peg you know price to

earnings growth ratio which actually

isn't that bad for a designing company

it actually feels kind of cheap it

should be probably closer to about a 2.6

Peg but to me that's a sign that some

people are starting to scratch their

heads going crap what if Nvidia doesn't

grow at 25% and what if they just like

even if they just keep revenue where it

is and it doesn't go negative which it

could their cash flow is still going to

be really really good but your price to

earnings growth ratio is going to be

infinite you know let's go down a little

bit and say they just grow at 5% per

year okay well now you have a 10 Peg

which is way over on the valuation and

now the valuation has to go down by five

V which means you know all of a sudden

you have this G you know AI related

crash where Nvidia goes down to $30 a

share I'm not saying that's going to

happen I'm just saying like on a quick

valuation basis that's what can happen

when people expect these insane growth

figures in hopium again don't get me

wrong I love Nvidia I think it's a great

company but this is a big red flag right

here and we saw what happened when you

know what was it uh arm and AMD they

missed by like 60 and 80 basis points on

guidance and then their stock dropped

like 8 to 15% over the next trading days

not great I mean look at AMD year to dat

AMD is up

3.6% year-to date look at arm year to

dat a double not bad go out a year 170

fantastic had a lot of growth but where

is it from Summer's Peak actually down

about 23% already since Summer's Peak

and arm you know might be a little

different because it did IPO uh you know

more recently when when did they IPO

again they ipoed in September of 23

after Nvidia wanted to buy them and the

deal didn't go through because of uh you

know good old antitrust but you know the

question is okay what what ends up

happening if growth starts slowing down

on some of these and this article

somewhat feeds us a little bit of

concern that some of the valuations and

expect expectations here could start

being being a little bit Rich we could

start seeing some oopsy dupsies now one

of the other things that we've

discovered is the labor market and

recession very important is that when

stocks go up you generally don't

generate more jobs or

hiring when stocks go down you tend to

lay off more so what happens if you now

have a dual Factor where stocks fall you

get lay offs but then because you have

ai because the AI bubbles person but

because you have ai you actually lay off

even more or you hire back less which

makes you know potential stimulus

efforts from the fed or otherwise even

harder kind of crazy it I don't know

where we're going in the world but I

just want you to know I'm going to be

here we're going to work on this

together uh and uh we'll we'll keep

paying attention to all this stuff

together so if you like this make sure

to subscribe get your free stocks with

Weeble if you want term life insurance

is the same thing Lauren and I use we

Apple pay for it it's on autopay every

month with Apple pay it's really great

go to metkevin.com lifee both of these

are paid sponsors of the channel Weeble

and uh life insurance are pretty great

go check them out uh and then if you

want a free Alpha report delivered to

you every morning at 6:15 a.m. make sure

you go to meetkevin.com

Alfa what's really cool about that is

two days ago I'm like I think you know

the volatility is really low on GameStop

I think this one might pop boom it's up

10% % on the day yesterday I'm like I

don't know I'm I'm feeling Tesla might

be topped Here Sell some calls or or you

know may maybe look at a certain put

boom things down 6% now no guarantees we

can keep that sort of stuff going uh

obviously there's risk in investing go

check that out over at meetkevin.com

Alfa at this point it's totally free so

you may as well sign up and get it and

see if you like it thanks so much for

watching folks and we'll see you in the

next one goodbye and good luck out there

why not advertise these things that you

told us here I feel like nobody else

knows about this we'll we'll try a

little advertising and see how it goes

congratulations man you have done so

much people love you people look up to

you Kevin PA there financial analyst and

YouTuber meet Kevin always great to get

your take