WEBVTT

00:00:00.599 --> 00:00:07.160
AI is everywhere and everyone is talking

00:00:03.800 --> 00:00:08.679
about it but as it turns out the vast

00:00:07.160 --> 00:00:14.599
majority of what they're saying is misleading at best and at worst an

00:00:11.960 --> 00:00:21.640
outright deception while the feature on your Smartwatch or your new co-pilot PC

00:00:17.359 --> 00:00:23.640
is called AI it's not the AI that you're

00:00:21.640 --> 00:00:28.800
probably thinking of so we're going to take a reality Hammer to this hype

00:00:25.519 --> 00:00:31.119
machine and break down what AI can do

00:00:28.800 --> 00:00:37.480
what it can't do and why the tech industry is so eager to slap the AI

00:00:33.960 --> 00:00:39.760
label on absolutely everything and guys

00:00:37.480 --> 00:00:45.680
this Rabbit Hole goes way deeper than you think okay predictably it ends at

00:00:42.960 --> 00:00:52.199
money but the path to get there it turns out is really confusing on purpose which

00:00:49.520 --> 00:00:58.440
makes it very interesting like this interesting message from our sponsor MSI

00:00:55.320 --> 00:01:00.760
their AIS rs2 is a pre-built PC touting

00:00:58.440 --> 00:01:08.280
some impressive specs like an Intel Core I74 1400 KF 16 gigs of ddr5 RAM and a

00:01:05.000 --> 00:01:10.439
GeForce RTX 4070 super check it out in

00:01:08.280 --> 00:01:10.439
the

00:01:17.520 --> 00:01:24.200
description the classic definition of AI is probably best Illustrated with

00:01:21.920 --> 00:01:30.200
fictional examples it's what you see in sci-fi Creations like Commander Data

00:01:26.840 --> 00:01:32.680
Hell 9000 and GLaDOS these are are

00:01:30.200 --> 00:01:38.000
computers or machines that demonstrate a capacity for reason however naive

00:01:35.520 --> 00:01:41.759
twisted or alien it might seem to us meatbags now you'd be forgiven for

00:01:40.119 --> 00:01:46.640
thinking that that's still the definition of AI a lot of people seem to

00:01:44.280 --> 00:01:51.640
think that it is but in reality the meaning of words is ever shifting and we

00:01:49.399 --> 00:01:58.119
would now refer to these characters as having a gii or artificial general

00:01:55.640 --> 00:02:06.000
intelligence what you're referring to as AI then is in fact narrow AI or as I've

00:02:01.640 --> 00:02:08.360
taken to calling it AI Ani is not a

00:02:06.000 --> 00:02:12.680
general intelligence unto itself but rather another component of a fully

00:02:10.679 --> 00:02:16.879
functioning system made useful by specialized algorithms and data

00:02:14.720 --> 00:02:21.120
processing utilities forming a complete artificial intelligence system didn't

00:02:19.760 --> 00:02:27.160
think I could make that point in the style of Richard stallman's famous

00:02:23.760 --> 00:02:29.000
interjection well I could haha but I

00:02:27.160 --> 00:02:34.040
also didn't have to that previous paragraph was actually written by gp4

00:02:31.640 --> 00:02:39.120
Omni and this is exactly the sort of thing that modern AI does very well and

00:02:37.319 --> 00:02:44.720
that's because most of the time when we hear the term AI we're actually

00:02:41.400 --> 00:02:47.080
referring to machine learning a subset

00:02:44.720 --> 00:02:52.319
of AI involving algorithms that can analyze patterns in data they get

00:02:49.840 --> 00:02:57.599
trained on things like text multimedia or even just raw number outputs and

00:02:54.959 --> 00:03:01.680
using this training data they identify patterns through statistical probability

00:03:00.159 --> 00:03:06.599
they can be further trained through reinforcement learning then by rewarding

00:03:04.040 --> 00:03:11.680
correct outputs and punishing incorrect outputs kind of like training a hamster

00:03:09.319 --> 00:03:17.120
the results allow these algorithms to summarize predict or even generate

00:03:14.400 --> 00:03:21.879
something seemingly new and in many cases they are so impressive that a good

00:03:20.120 --> 00:03:28.480
machine learning system can be indistinguishable from classic AI or a

00:03:25.640 --> 00:03:34.480
gii well then minus if it looks like an AI and it quacks like in AI what's the

00:03:31.560 --> 00:03:40.560
difference well artificial narrow intelligence is limited to specialized

00:03:36.760 --> 00:03:42.280
tasks gp4 Omni specifically is a large

00:03:40.560 --> 00:03:46.080
language model which means that it is trained to understand and generate

00:03:44.319 --> 00:03:51.040
natural language like the words I'm speaking now it's basically an

00:03:48.239 --> 00:03:55.280
autocomplete on steroids what sets it apart from your phone's keyboard though

00:03:52.720 --> 00:03:59.360
is that it can also process information based on patterns that are learned

00:03:56.760 --> 00:04:03.480
during training including definitions mathem iCal formula and so on and so

00:04:01.799 --> 00:04:08.400
forth that makes it capable of generating unique output that wasn't

00:04:05.799 --> 00:04:13.480
part of its training data GPT has traditionally been incapable of image

00:04:11.079 --> 00:04:18.120
video or audio generation there are other types of generative models like

00:04:15.439 --> 00:04:22.560
Sora sunno or dolly that feature their own specific talents but most of them

00:04:20.199 --> 00:04:26.320
are incapable of operating outside of their specific Niche and all of them are

00:04:24.759 --> 00:04:30.479
limited by their training data in a similar Manner and because they're

00:04:28.400 --> 00:04:34.800
limited by their training data in many cases the answers that they give

00:04:32.360 --> 00:04:39.360
resemble their training data which if you're an artist or a photographer and

00:04:36.720 --> 00:04:45.080
your work gets added to a model is probably not your idea of fair use much

00:04:41.960 --> 00:04:46.639
less a good time worse when generative

00:04:45.080 --> 00:04:51.280
models are faced with a concept that they don't understand or they simply run

00:04:49.000 --> 00:04:56.160
out of tokens they can begin to hallucinate that is to say they just

00:04:53.919 --> 00:05:00.400
make things up as they go which is why sometimes you get eldrich Abominations

00:04:57.960 --> 00:05:04.479
like these with that said these limitations don't mean that machine

00:05:02.160 --> 00:05:09.680
learning AI is a dead end it's been deployed very effectively for diagnosing

00:05:07.240 --> 00:05:14.400
diseases and in other highly complex scenarios where the data is dense and

00:05:11.880 --> 00:05:19.720
the conclusions require interpretation these specialized models are extremely

00:05:16.680 --> 00:05:22.120
useful they're just also extremely not

00:05:19.720 --> 00:05:26.720
new simple neural networks have been in use for decades for things ranging from

00:05:24.680 --> 00:05:32.759
handwriting recognition to web traffic analysis and yes even video game

00:05:30.280 --> 00:05:36.880
these are not scripted sequences the AI is determining when your allies choose

00:05:35.080 --> 00:05:42.280
to advance and how to best help you out in combat and chatbots the main

00:05:39.520 --> 00:05:47.080
difference is that they run much faster on Modern Hardware if I had to distill

00:05:45.039 --> 00:05:51.520
down what artificial narrow intelligence really means then I would say it's like

00:05:49.319 --> 00:05:55.720
having a thousand monkeys at a thousand typewriters with a thousand pieces of

00:05:53.880 --> 00:05:59.639
reference material for what the outputs are supposed to look like with enough

00:05:58.000 --> 00:06:04.000
trial and error then they do arrive arrive at a point where they're likely

00:06:01.720 --> 00:06:09.560
to spit out a correct or at least correct enough solution then we take all

00:06:07.400 --> 00:06:14.800
those monkeys and we take a snapshot of the model State and we start feeding it

00:06:11.599 --> 00:06:17.599
inputs for both Fun and Profit what ani

00:06:14.800 --> 00:06:22.120
is to a brain then is kind of what a single app is to a computer it's a

00:06:20.080 --> 00:06:28.000
building block it's something your brain is capable of but it's just one of its

00:06:25.000 --> 00:06:30.080
many many functions shifting GE a bit

00:06:28.000 --> 00:06:34.360
then what would artificial general intelligence look like well it

00:06:33.080 --> 00:06:38.360
would need to be able to handle everything we've talked about so far

00:06:36.400 --> 00:06:43.240
just like your brain can take some past experiences and turn them into a new

00:06:40.080 --> 00:06:45.560
creation but again like your own brain

00:06:43.240 --> 00:06:50.400
it would need to be able to run many of these models concurrently and

00:06:47.599 --> 00:06:55.599
continuously train and iterate on them rather than relying on fixed snapshots

00:06:53.199 --> 00:07:00.000
only then would an AGI have the ability to truly learn and adapt to new things

00:06:58.400 --> 00:07:04.919
bringing it closer to that that classical definition of AI and really

00:07:02.520 --> 00:07:10.479
blur the lines between machine learning and machine Consciousness the problem is

00:07:08.560 --> 00:07:15.000
even if we had software that sophisticated we are nowhere close to

00:07:12.840 --> 00:07:22.599
being able to run an AGI even on a modern supercomput let alone on your AI

00:07:19.400 --> 00:07:24.199
smartphone but all right lonus you still

00:07:22.599 --> 00:07:29.360
haven't explained why any of this is even a problem I mean freerange meat is

00:07:26.840 --> 00:07:34.879
just marketing bollocks too so who cares well well truthfully in most cases I

00:07:31.960 --> 00:07:39.639
don't I mean cooler Master's AI thermal paste snafu I was never bothered by it

00:07:37.479 --> 00:07:45.280
because I never expected my paste to be sentient anyway but there are situations

00:07:43.199 --> 00:07:51.360
where this kind of marketing can have an impact on user safety and therefore does

00:07:47.800 --> 00:07:53.879
matter let's talk about Tesla Mr musk

00:07:51.360 --> 00:07:59.159
has said among other things that any vehicle from 2019 onward will be able to

00:07:56.280 --> 00:08:03.319
reach full autonomy and he's certainly put out some impressive demos both

00:08:01.280 --> 00:08:08.520
canned and even in the form of public beta software that you really can use

00:08:05.720 --> 00:08:13.360
and that's really cool but unfortunately it isn't much more than that you see to

00:08:11.000 --> 00:08:18.440
operate a vehicle safely it's not enough to be trained with images of painted

00:08:15.960 --> 00:08:23.400
lines and traffic cones stop signs pedestrians vehicle Telemetry data it's

00:08:21.159 --> 00:08:29.039
not even enough to be trained to predict the likely maneuvers of nearby vehicles

00:08:25.400 --> 00:08:32.399
and life forms on the road anything can

00:08:29.039 --> 00:08:35.839
happen and by definition by its very

00:08:32.399 --> 00:08:37.599
definition Ani is not capable of

00:08:35.839 --> 00:08:42.479
handling an edge case that it has never seen before even if it was by the way I

00:08:40.680 --> 00:08:49.040
have some really bad news for you Tesla owners out there Hardware 3.0 has about

00:08:46.120 --> 00:08:54.160
144 tops or trillion operations per second worth of processing power for

00:08:51.480 --> 00:08:58.680
context Windows 11 recall a feature that does little more than take screenshots

00:08:56.040 --> 00:09:04.600
and analyze your PC usage for search asks for 40 tops now to be clear tops is

00:09:02.880 --> 00:09:08.640
not a be all andall measure of performance and there is no way that

00:09:06.959 --> 00:09:13.000
Microsoft has optimized the code for recall nearly as much as Tesla has for

00:09:10.720 --> 00:09:17.320
full self-driving but this should still illustrate the point that Tesla either

00:09:14.880 --> 00:09:23.200
did or should have known that a vehicle with the AI capabilities of a family of

00:09:20.040 --> 00:09:25.519
iPhone 15 Pro users would never achieve

00:09:23.200 --> 00:09:29.680
that kind of realtime contextual awareness that's required for complex

00:09:27.839 --> 00:09:33.680
situations like operating a motor vehicle and they misrepresented its

00:09:31.800 --> 00:09:38.880
capabilities in order to sell more software that was never going to leave

00:09:35.760 --> 00:09:41.320
beta that is going to be a doozy of a

00:09:38.880 --> 00:09:46.360
class action and it's a common story that has led to this current mess where

00:09:43.959 --> 00:09:51.360
fuzzy definitions and impossible promises have turned AI into this

00:09:49.120 --> 00:09:56.800
meaningless buzzword like all the rest of them all of them refer to legitimate

00:09:54.680 --> 00:10:00.600
useful Technologies some of which have really come to fruition but their

00:09:58.800 --> 00:10:05.680
meanings have become diluted with overuse and it means that when computer

00:10:03.480 --> 00:10:08.959
cognition finally happens we're going to have to call it something completely

00:10:07.200 --> 00:10:15.279
different in order to differentiate it from all of the marketing wank on the

00:10:11.680 --> 00:10:17.959
subject of marketing calling ai ai

00:10:15.279 --> 00:10:23.640
wasn't an accident the people behind that marketing know what you think AI

00:10:21.079 --> 00:10:27.720
means they know that the promise and the Mystique behind the term is tantalizing

00:10:26.079 --> 00:10:32.640
and they know you'll click on an article that is interviewing an AI expert who

00:10:29.720 --> 00:10:37.079
discusses how dangerous or already alive it is they want you to buy into their

00:10:34.760 --> 00:10:43.000
hype they want you to buy into their stocks what we have now though really

00:10:39.800 --> 00:10:44.720
are decent summarization engines and

00:10:43.000 --> 00:10:49.360
lukewarm guessing machines that are tuned for working with different types

00:10:46.240 --> 00:10:52.000
of media they can't reason and anyone

00:10:49.360 --> 00:10:57.200
who's seen chat GPT get something ridiculously confidently wrong can

00:10:54.680 --> 00:11:01.240
attest to that and they also can't understand why they get these things

00:10:59.320 --> 00:11:06.320
wrong which mean they can't learn or improve on their own even if you

00:11:03.600 --> 00:11:10.600
explicitly tell them they also have a finite number of things that they can

00:11:08.000 --> 00:11:14.760
remember called tokens which limit their ability to maintain a train of thought

00:11:12.760 --> 00:11:19.160
for very long anyone who's tried to get chat GPT to write a novel or even a

00:11:17.160 --> 00:11:24.040
moderately complex python script can probably attest to that of course as

00:11:22.000 --> 00:11:28.320
time goes on some of these limitations will begin to lift and these a Ani

00:11:25.959 --> 00:11:32.920
models will start to look more and more like a GI to the lay person I mean just

00:11:31.200 --> 00:11:37.800
ask the guy who had Bing fall in love with him but at the end of the day it's

00:11:35.279 --> 00:11:42.120
still the same thing as it's always been to paraphrase a paper co-written by

00:11:39.560 --> 00:11:46.399
Emily bender and Alexander caller it's a hyper intelligent octopus that is

00:11:44.560 --> 00:11:50.600
observing and learning the patterns that are expected of communication and

00:11:48.560 --> 00:11:55.079
repeating them back more accurately over time so when faced with a novel

00:11:52.639 --> 00:11:59.480
situation like a bear attack that octopus has no hope of being able to

00:11:57.800 --> 00:12:04.800
guide you on how to defend yourself with the materials you have on hand it has no

00:12:01.680 --> 00:12:06.480
concept of bear or stick only what words

00:12:04.800 --> 00:12:10.959
most closely match the pattern that it has previously observed to further

00:12:08.880 --> 00:12:15.040
illustrate this if you haven't tried it before try to craft a prompt that coaxes

00:12:13.320 --> 00:12:20.040
stable diffusion to spit out something very specific it turns out it's pretty

00:12:18.079 --> 00:12:23.480
challenging and when it finally does it it's generally composed of a mosaic of

00:12:21.959 --> 00:12:28.839
shapes and patterns from its training model that best match the keywords that

00:12:26.360 --> 00:12:33.600
sounds overly simplified but that's because it is in essence what it's doing

00:12:31.880 --> 00:12:37.800
it takes the keywords from your prompt and starts compositing filling in the

00:12:35.519 --> 00:12:43.279
image until it hits for example a certain percentage of computer and desk

00:12:40.800 --> 00:12:47.639
and it does this without any context so if you asked it to show you a computer

00:12:45.160 --> 00:12:53.000
would it show you a desktop PC a laptop a server rack and for that matter from

00:12:49.720 --> 00:12:55.680
what era the answer is often simply yes

00:12:53.000 --> 00:13:00.360
because it lacks the context behind what computer means to you and it can't

00:12:58.440 --> 00:13:05.120
always even for a coherent depiction even in the same image so then Linus

00:13:03.279 --> 00:13:09.560
just don't worry about it and wait till AGI shows up and you don't have to

00:13:06.600 --> 00:13:14.279
bother with online dating anymore well not quite as generative AI models learn

00:13:12.600 --> 00:13:17.959
to create works that are closer and closer to our expectations and to

00:13:16.440 --> 00:13:21.880
finally understand how many fingers and teeth humans normally have we're going

00:13:20.240 --> 00:13:27.639
to start to find it more and more difficult to distinguish their output from real photographs or artwork or

00:13:25.680 --> 00:13:32.000
other people ushering in an era of unprecedented distrust and the bad news

00:13:30.639 --> 00:13:35.920
is that the folks in charge of helping us deal with the consequences of all of

00:13:33.760 --> 00:13:41.240
this have a lot less funding than the ones who are trying to sell it to us so

00:13:38.600 --> 00:13:45.320
be safe out there and safely check out our sponsor manscaped if you're looking

00:13:43.639 --> 00:13:50.720
for a way to quickly take care of your more delicate Parts manscaped is on the

00:13:48.279 --> 00:13:55.720
case with a performance package 5.0 Ultra it includes the lawnmower 5.0

00:13:53.399 --> 00:14:00.279
Ultra to trim the grass as well as the weed whacker 2.0 for your ears and nose

00:13:58.800 --> 00:14:03.639
so it doesn't feel like you're always tickling yourself you'll even get their

00:14:01.920 --> 00:14:07.839
crop Soother made with shea butter and Coco seed butter designed to pamper your

00:14:05.759 --> 00:14:12.160
delicate areas with essential moisture and soothing relief and for an added

00:14:09.920 --> 00:14:16.160
layer of freshness give your package a Royal Treatment with their crop

00:14:13.639 --> 00:14:21.160
preserver to keep the funk at Bay check out the performance package 5.0 Ultra at

00:14:18.560 --> 00:14:25.800
manscape.com techtips and use code techtips for 20% off and free shipping

00:14:24.600 --> 00:14:32.480
if you guys enjoyed this video and you're looking for some more debunking spelunking maybe check out recent video

00:14:29.920 --> 00:14:36.720
on PC speed up tools to see if they really do what they say they do