[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/biz/ - Business & Finance

View post   
File: 2.75 MB, 2470x1564, file.png [View same] [iqdb] [saucenao] [google]
58632655 No.58632655 [Reply] [Original]

Linus has made a bold prediction that Teslas will never have a functional autonomous driving capability. He says this is because narrow artificial intelligence is not capable of handling an edge case it has never seen before and that only general artificial intelligence would be capable of that.

>> No.58632676
File: 21 KB, 400x400, 6beJNSck_400x400.jpg [View same] [iqdb] [saucenao] [google]
58632676

>>58632655
>>58632655
he is right. AI is the biggest hype train we've seen since VR or something. Elon is basically Elizabeth Holmes from Theranus x 10. He has lied about every product Tesla has ever put out and under delivered at every step of the way. If it were not for an ideological government that supports EVs it wouldn't be a thing at this point in time.

Betting on Elon and Elon related business is the shitcoinery of the legacy markets. We're witnessing the greatest con man in history in terms of wealth. The price target for TSLA is $0. Bankruptcy.

>> No.58632715

>>58632676

human-error accidents that occur on the road wouldn't happen if an AI was driving. And the AI edge cases that do cause fatalities would be far less than the number of those human error accidents.

>> No.58632728

>>58632715
>>58632655
have AI drive trucks and see if insurance rates drop for those trucks

>> No.58632729

>>58632655
He seems to have elon derangement syndrome (EDS). Listen to his voice tone at 7:50.

>> No.58632738

>>58632655
He's retarded. It only needs to be demonstrably safer than human drivers in aggregate to be widely adopted. And it already is. Even if it's not "real" AI, the machine can't drunk drive, get distracted by a phone or conversation, or get tired. The only thing standing in the way is the slow pace of government and changing regulations. While it could take a few years that's much different than "never."

>> No.58632751

>>58632738
it's not safer than humans stop consuming garbage pseudoscience marketing

>>58632729
Elon sucks are now adopting trump cuck language to describe their leader. very sad.

>> No.58632755

>>58632751
disliking a person creates a bias that makes it difficult to arrive at the objective truth. That's probably why Linus is so wrong on this topic.

>> No.58632777

>>58632655
I was literally hating on elon in a different thread but he's not elizabeth holmes tier. She was just a charlatan with NOTHING. He overpromises and underdelivers massively, but he always delivers *something*.
>Claims fully autonomous driving, offers the same assisted driving everybody else already has
Always doing JUST enough to not be a complete fraud. Which is probably the reason why he's so much more succesfull than Holmes or SBF ever was. Why exactly that justifies a market cap of over 500 billion dollars I have no idea

>> No.58632789

>>58632777
wdym? Tesla's assisted driving is light years ahead of the competition

>> No.58632805

I believe fully autonomous driving is quite possible, but only at speeds of about 5km/h.

>> No.58632830

>>58632777
>He overpromises and underdelivers massively, but he always delivers *something*.

This is not the case with SpaceX. Falcon9 is/was actually revolutionary. I don't think people understand just how much SpaceX is kicking everyone's ass in that market.

>> No.58632851

>>58632789
>>58632830
Like I said, he throws out "something" so they can't just put him in prison for obvious fraud. Everybody except for mentally ill musk dickriders know he is a conman, but a succesful one who's perfected his craft of grifting to not be so brazen that he's acutely in danger of landing behind bars.

>> No.58632854

>>58632851
>he throws out "something"
Rocket re-usability is not just "something"

>> No.58632863

How do I become a grifter? I once grifted it and I still feel guilty about it. I'm so tired of these morals.

>> No.58632892

>>58632655
>bold prediction that Teslas will never have a functional autonomous driving capability.
very midwit claim to make considering were at the beginning of the AI spring
>hurr durr ai by definition can't react to everything
doesn't need to
there will be some safety feature that slows down/stops the car for whatever 1 in 100,000km event arises which will be recorded by the car and sent to some server for the ai to learn from

>> No.58632928

>>58632892
i mean, look at this shit 6:25
https://youtu.be/43Lrrhn0CMk?t=382
it realised the incoming car wouldn't fit in the street so it moved slightly out of the way
5 more years and it will be perfected

>> No.58632933
File: 159 KB, 1080x1375, 1693775739343113.jpg [View same] [iqdb] [saucenao] [google]
58632933

>>58632715
>you should trust your life to a mountain of spaghetti code hammered out by pajeets instead of your own skills and abilities
hard pass, my guy

>> No.58632941

>>58632933
AI models aren't coded kek.

>> No.58632964

>>58632655
linus got that whole video wrong. he claims that modern AI does not have the capacity to reason, and can only repeat things it has already seen, but that's completely false. Even chatgpt has the ability to reason

>> No.58632966

>>58632941
please explain the process by which an AI (aka a computer program) gets created in your mind. I'm assuming it involves wizards, magic spells, and flying ponies.

>> No.58632970

>>58632966
new connections are made between data automatically, you don't have to program in an explicit relationship between data.

>> No.58632978

>>58632970
so if you put two documents on a disk, they just start automatically making connections with each other? that is fascinating

>> No.58632989

>>58632978
it's like how God put some very basic amino acids in the primordial soup and then evolution occurred and many complex organisms developed.

Programmers put some very basic pieces of code into a program that parses data. That program then has the ability to identify very complex relationships between the data on its own.

>> No.58633016

>>58632989
Jesus fucking christ are you retarded.
No wonder a low iq subhuman like you loves musk so much because you are the textbook definition of dunning kruger.
Yea I get it that DK is overused but ur "understanding" of AI makes me think you didn't even graduate high school.

>> No.58633063

>>58633016
it's true tho. the 'simple program' is literally just a framework for assembling virtual neurons. those neurons are just single cells that perform a very basic task.

>> No.58633081

>>58632751
You don't seem to understand how fucking terrible people are at driving a car.

>> No.58633107

>>58632964
>Even chatgpt has the ability to reason
no it doesnt

>> No.58633109

>>58633063
This. AI has not evolved since the 60's and we're still using those same algorithms. It's the computing capability that has multiplied.

>> No.58633151

>>58633107
Do you think something like 12312321 + 5324323 and its result was in the GPT4o training data set? I don't think so. But it can perfectly compute any simple math operation without the need of a calculator. I don't think that a monkey on a typewriter can do that, without knowing how summation works. Same with multiplication

>> No.58633160
File: 1.65 MB, 4096x4096, file.png [View same] [iqdb] [saucenao] [google]
58633160

>>58632655
>He says this is because narrow artificial intelligence is not capable of handling an edge case it has never seen before

it's a short sighted argument because he's not taking into account that this is an active area of research so they will be training the cars even better in the future. plus now they have a lot more training data in the last year than they ever had before as the amount of miles their cars are self driving has skyrocketed. 10 years ago we had pretty much no self driving cars on the road and they were only self driving at research facilities. now we have millions of cars with the capacity to self drive and they have recently crossed the threshhold of driving over a billion miles. what happens when these cars have trillions of more miles driven, or hundreds of trillion. and who's to say that future self driving cars won't be able to see things better and more accurately and understand the world better. who's to say that in 10 years we wont have self driving cars that are orders of magnitude smarter than they are now.

also a lot of times when people say "Never" in the technology field they are proven wrong eventually. these cars are not an impossible technology, they already exist and are already quite good at a lot of things. they will for sure get much better.

>> No.58633372

>>58632655
>>58632715
For this reason, I envision a future where the state bans cars that arent run by AI or which even have the capacity for human input.

>> No.58633389

>>58632715
The problem is with human error someone is clearly liable, there is a clear path for legal subrogation. With an AI jews wont be able to sue anyone and that wont fly.

>> No.58633497
File: 38 KB, 562x1000, 81LOK+RazfL._AC_UF1000,1000_QL80_.jpg [View same] [iqdb] [saucenao] [google]
58633497

>>58633160
Your argument is missing the context that AI has been an "active area of research" since the 1980's and before. It has produced more disappointments than successes. The entire death of Lisp and Lisp Machines was due to DARPA/DOD pulling funding on AI: https://en.wikipedia.org/wiki/AI_winter

The only major new revelation was using drastically more sophisticated computer power to shift all focus to a machine learning/LLM paradigm, because modern computers can crunch tons of data. All modern AI relies on essentially stochastic processing of input against a large volume of data, whether it's from language/NLP (ChatGPT) or images/video (Tesla driver aides for example).

These systems "reason" by running a statistical match against things they've already seen. They do not possess fundamental concepts inherit to human beings or animals that make logical decisions, nor do they perform self-analysis. For instance, I've never seen an overturned truck on a highway, but if I do I'm sure as fuck not going to drive my car into it. Telsa "autopilot" drove right into it due to the machine learning basically have no concept for it: https://www.youtube.com/watch?v=ZQA1cY6n7YM

If I see someone wearing a shirt with a stop sign on it, I'm not going to slam on my brakes. No luck for a robotaxi: https://www.motortrend.com/news/can-a-t-shirt-stop-a-waymo-driverless-taxi-vehicle/

This is also why you can turn ChatGPT into a psychotic parrot. It does not have the ability to look at what it's saying and go, "hey that makes no fucking sense". It's just processing your input and running a match.

The modern AI hype cycle is mostly about securing venture capital funding for a novel technology by making it sound like the Next Big Thing so investors don't want to "miss out". They're relying on the dishonest idea that AI in real life is like AI in movies (a human-like intelligence), which, to paraphrase a book, is like inventing an airplane by trying to make a giant mechanical bird.

>> No.58633533

>>58632655
>>58632676
I find yours and Linus' opinions bemusing.
Linus may be good at building PC's but he frequently makes it obvious he has never formally studied engineering, or IT, or compsci, or anything really. He probably hasn't written a Hello World in his life.
So, you only just figured out that AI doesn't mean sentient technological beings like wat they dun showed in the movies, and suddenly that means recent AI developments are hype, are a lie, are nothing that special?
I am 45 years old. I did my undergraduate bachelor of IT in 1996. Even back then, the distinction was made between neural networks that just performed machine learning, or specialised in narrow fields, and general intelligence. It's not a scam that things changed - when AI was first thought about I guess people didn't realise it would end up being sliced like that.
It's amazing to me when normies have their bubble burst because they realise ChatGPT isn't some sentient being. No shit, you fucking retards.
The recent developments in AI ARE revolutionary and the shockwaves will be felt around the world for years to come. You just haven't imagined how the tech might affect you yet.
To pull just a few examples - mass surveillance is now orders of magnitude more capable, with all our conversations not just stored in some giant data warehouse, searchable by keywords, but now parsed and understood by machines to levels of granularity never dreamed of 5 years ago. AI's can parse petabytes of legal documents and outperform lawyers in applying law and finding cases. The drone assassins we see over on /gif/ doing the devils work in Ukraine are now able to do it without the input of signals (that might get jammed), with AI pilots given a human target and their approximate location at the start of the mission. The Tesla autpilot is amazing and the best in the industry, even with edge cases that humans will be better at.

>> No.58633604

>>58633497
Your whole argument seems to be that because AI isn't AGI, and that this is contrary to what many gullible normies and boomers out there think, that therefore the AI we see in action is a cheap trick and hyped up for venture capital.
You fail to see that in their current, flawed state, aided by modern computation and storage capacity as much as they are, the AI's we have are doing terrifying things, they are extraordinarily useful and capable tools that are going to change our social, technical, political, creative landscapes.
Yep, a Tesla drove into an overturned truck and balked at a Stop sign on somebody's t-shirt. That is only a mindfuck for people who expect it to be an AGI. All we should really care about is the number of accidents, near misses etc per mile driven. What Tesla Autopilot lacks in general intelligence and its ability to deal with edge cases, it makes up for (statistically speaking) in doing what it's good at reliably, for millions of miles, with fewer incidents than humans. I still wouldn't trust it in congested cities and laneways, but it looks like it's doing really well there now (it used to only be great on the highway), and the rate at which it is improving is phenomenal. Other auto makers won't be too far behind.

>> No.58633650

>>58632655
The problem with AI and all technology in general really is that as soon as we invent something it immedietly becomes boring and unimpressive. 50 years ago the idea that you could carry a wireless phone around in your pocket was seen as super futuristic, now we have something way way way more advanced than that and we all think its gay. The tesla thing is already really impressive, but its not impressive 'enough' so its gay. When it does actually work people will still think its gay, and so on forever until we're living in gay space communist future world with robots acting out tiktok videos for us on the moon

>> No.58633667

>>58632655
>I know more than the one of the most prominent AI developers in the world.


Thats just stupid, people said that Space X wouldn't achieve a reusable rocket and here we are with those things launching and landing effortlessly.

>> No.58633680

>>58633604
>>58633533
cont. I guess another way to put it is, I think a lot of you young critics have never tried to write a piece of software that does some of these tasks that AI can do now, WITHOUT using AI.
I have. Plenty of nerds back in the 90's tried to make chatbots by writing thousands of lines of C or Basic, trying to bruteforce responses to millions of possible inputs, or trying to get sophisticated and parse the human inputs for some basic grammar, using dictionaries, thesaurus's to find synonyms. Absolutely hopeless and the best rated and raved about chatbots were utter garbage.
I worked with a guy who did his masters in the 90's trying to perform line detection on images. He even tried using a neural network. It was fucking trash and he not only gave up, but declared it was an impossible task that nobody would ever solve.
Can you imagine trying to write a car autoilot in C, with millions of lines of code trying to output the appropriate steering, throttle, brake inputs given the inputs coming through a camera and a radar? The stage we are at now is fucking science fiction compared to back then.

>> No.58633804

>>58633533
>>58633604
>>58633680
I'm not gonna do a line by line of what you wrote. I know you have some IT knowledge, but I think you're overstepping the bounds of both your knowledge and the argument I'm actually making. Nobody is contesting that machine learning will probably be around forever and will continue to evolve - I am simply not optimistic that the childish idea that actually self-driving cars (not, "pattern recognition until an accident is happening then it switches back to human control for liability purposes") and human-like robots are anywhere close. All new technology has to be tempered with reason about what it actually is and what it does - with so much experience you should know this.

I do have to laugh at the part about having to write everything in C and Basic back in the day. You should really read up on Lisp machines, their performance and what was done with them. You're the Dunning Kruger effect on display, not because you're stupid, but because you're not informed and like to spout off very arrogantly anyway.

>> No.58634524

>>58632655
That's not a bold statement, that's just how computers work.

Also he's shitting on self driving while self driving is in a low year.

He's bobo bottom signaling.

>> No.58634544

>>58633680
Yea actually because I met a guy who did.

Kid was a zoomzoom and the company he worked for wouldn't let him run his code on an on the road vehicle because it was.

>"an infinite liability glitch"
Maybe the kid ran language models in his off time but from what I understand it was pretty clearly a baby script function that just kept trying new solutions until something worked.

It never hit anything important in the private lots it ran at, and when it did. It was because sensors fed it bad info.

>> No.58634587

>>58633497

stop confusing us with the truth

>> No.58634593

>>58632933
I would return the vehicle and ask them to pay the tow truck fee,

If they couldn't, We would go to court and they might squish me. But not before I called a journalist so i could become a lolcow.

>> No.58634596

>>58633533

ok boomer

>> No.58634601
File: 112 KB, 760x739, 2a0-4272627488.jpg [View same] [iqdb] [saucenao] [google]
58634601

>> No.58634665

>>58632933
I would trust a Tesla (same for elon's spacex ships) if it was
>open source
>audited by autists
>computer simulations made public
instead of it's dude trust me oops you can't stay in your car during update season because it might cook you. LMAO
>>58632941 kill yourself

>> No.58634710

>>58632655

The real question is: Will the AI car be able to dodge the IED I planted by the roadside? Because fuck AI. Sabotage skynet.

>> No.58634730

>>58632715
I think we should finally agree that equality is a lie and some people are objectively better than others. I should be able to drive myself. Your 80 year old grandma? Your retarded inbred meth addicted cousin? 400 lbs 85 IQ Shaniqua? No they should be driven by AI cars (straight off a cliff). I should also be able to drive at a much higher ABV because, again, I am better than you.

>> No.58634733

>>58633533
>He probably hasn't written a Hello World in his life.

>import You

>if You.are_a_faggot:
> print('Hello, world!')

>> No.58634740

>>58633063
If youre point is computer "AI" isn't actually AI its just a very basic attempt at what they call a "neural network" then yes. But your point was that no coding is involved in machine learning which is asinine.

>> No.58634760

>>58633151
gpt does inline results from programmatic operations. it will even tell you sometimes
ie if you ask what word is at x position in a sentence it will run a python script to count words in a string

this isnt reasoning

>> No.58635275

>>58632676
>>58632830
>>58632854
Did it never occur to you lot that Spacex is fake af? You deride all his other efforts but you totally buy the one that you really have no proof of.
Space really IS fake. It has always been a psyop to soften you up for the great alien 'invasion'

>> No.58635288

What we call "AI" is just a massive table of probabilities. Anything that can't be solved by a massive table of probabilities can't be solved by "AI." What we call "AI" is the next generation of computational technology, but it has hard limits to its abilities.

>> No.58635301

>>58632751
you type like you've got massive hairy titties

>> No.58635601

>>58633804
Your argument about self driving cars might hold water if humans did not cause any accidents on the roads.

>> No.58635628

>>58635275
Elon does not exist. He is completely made up human. You have no proof that he exists.

>> No.58635641

>>58632655

Who gives a fuck about this soulless, beedy eyed nigger!?

>> No.58635790

>>58632966
Holy fucking reddit moment.

>> No.58635849

>>58632655
we are in here >>58629037

>> No.58636223

I think he’s right but eventually the entire world will be designed in a way that AI can handle. Humans will be the actual robots facilitating it all

>> No.58636327

>>58632655
I want all of the retards on the road to be forced to submit to AI chauffeur mode because they can't seem to help but drive 10MPH under the limit and ride their brakes.

>> No.58636363

>>58632676
this

>> No.58636382

>>58632655
DUH, SOMETHING YOU COULD HAVE KNOWN OVER A FUCKING DECADE AGO FUCKING DUH DUH NO SHIT NO FUCKING SHIT DUMB FUCKING COCK SUCKERS

>> No.58636393
File: 655 KB, 2386x1718, 1705545445423906.jpg [View same] [iqdb] [saucenao] [google]
58636393

>Linus
lmao
kek even

>> No.58636441

>>58636393
daaaaamn i wasn't aware he was quite this cringe.

there are literally already self-driving taxis he is an actual retard for that take.

>> No.58636462

>>58632715
>conclusions arrived to by absolutely nothing at all.
Bro I just feeeeeeeel like MuH Ai.. Blah blah. Ai is fake and gay, fuck self driving cars, what you need more time to goon so badly that you can't be bothered to drive anymore? Fuck off.

>> No.58636474

>>58636441
>gets in self driving taxi
>dies

>> No.58636494

>>58636474
>gets in human driving taxi
>dies

>> No.58636547

>>58636494
self driving vehicles are death machines

>> No.58636565

>>58632655
There are millions of teslas on the road gathering data for the self driving algorithm. The number of teslas on the road grows every day. More cars means more data, more data means less unknown cases for the AI. If minorities can drive on the road then a computer program designed by white and Asian engineers can drive on the road even betterily. Stfu doomer bitch

>> No.58636696

>>58636547
you need to look up some statistics. if everyone were in self driving vehicles there would be so few vehicular fatalities it would be mind-boggling. human error is a factor in every single accident. if you remove the human error it almost eliminates car accidents.

>> No.58636756

>>58633804
>I do have to laugh at the part about having to write everything in C and Basic back in the day. You should really read up on Lisp machines, their performance and what was done with them. You're the Dunning Kruger effect on display, not because you're stupid, but because you're not informed and like to spout off very arrogantly anyway.
Bro, I may have to throw a little of the Dunning Kreuger accusation back at you (with the same respect you showed me though). Lisp is just a language, it doesn't have some intrinsic neural network or ML ability, it just so happened to be the language used in early AI experiments. As much due to the language as due to the hardware that got a foothold in research labs, that were built on Lisp. The programming language itself had some attempts at being "human readable", which many languages attempted. E.g. Basic, Cobol. Some sections of code could look more like pseudo code than indecipherable machine language, but that is not AI!
I never programmed in Lisp however the reason I know, and knew at the time, that it wasn't some miracle AI language, is that I can tell you throughout 90's, tens of thousands of nerds of all persuasions were trying to pass the Turing test (create a chat bot that can make a person think it's another person). They used all the technologies available to them, including Lisp, including older neural networks. And they were all fucking shit. "I'm sorry, I didn't get that, can you say it in a different way?" x 1000.
I talked to these bots, many from individuals, some from companies, to test them, to think about writing my own.
There were huge incentives to pass the Turing test, including a $100k reward in the Loebner comp. https://en.wikipedia.org/wiki/Loebner_Prize
Trust me, nobody passed, and it wasn't even close. Until the current generation of AI came along. The competition ended in 201.

>> No.58636785

>>58632655
At first i didn't look at the entire image and i thought you were talking about Linus Torvalds. I took the opinion seriously enough to consider that it may be correct, even though Torvalds is kind of a retarded faggot these days. But when i noticed it was THIS linus, i discarded the opinion immediately. I know very little about ai. Funny how that works.

>> No.58636910

Get aload of this kike...

What a heeb.... If you find your 'tech tips' via Linus....

You are the fucking problem.... Kys kikeniggerfaggots