is AI gonna take our jobs?

I know there are automobile-related recalls all the time, but they really can't be putting out cars like that out there if they aren't 1000% ready.
 
It's more limited than just 40mph:

"Interestingly, Mercedes says that Drive Pilot will only operate at speeds up to 40 mph on “suitable freeway sections and where there is high traffic density” — which seems to suggest it will only be available in heavy, stop-and-go traffic."

Part of the problem is that they can't train these systems without racking up millions of miles of driving experience. So basically, they can't release a system to the public that's already good to go because they need to use the public as beta testers.

I'm sure that self-driving cars will be the only thing on the road...sooner or later...but for right now it's crazy dangerous to have cars which drive themselves but you're supposed to be monitoring at all times and ready to take over at a moment's notice. The whole point of autonomous driving is so you can do something else. They're basically requiring all the attentiveness of actual driving without the involvement that helps you stay focused.

All that being said, it would be a gift if your car would drive you through gridlock. That's my least favorite type of driving -- stop-and-start driving during Manhattan rush hour. And you never go above 5 mph, so the stakes are pretty low if the AI screws it up.
 
DLD said:
It's called "teething trouble". The problem with these AI systems is that there's no predecessor to estimate what it will be once fully developed.

(if a frunk flies off a C-8 Corvette, one thinks that GM will solve the problem bu putting higher quality latches ... but here no one knows where exactly this is going and what exactly it will become)
Click to expand...

One of the tricky things here is that I don't think the machine learning engineers fully understand how large language models like GPT-3 work. It's referred to as a black box. And that doesn't mean that they can't fix it (perhaps through trial and error, kind of like how they created it), but it would probably be easier to fix if they understood its inner workings better.
 
From what I recall about the autonomous driving future from few years ago - the system needed to have an infrastructure in place with sensors and cameras everywhere along the route, on top of what each automobile had. In other words, a corner of Washington and Main would have an own system over and beyond what each car approaching Washington & Main would have installed in the vehicle. Mercedes Benz is apparently now authorized to go ahead without that infrastructure. But as long as they don't go ahead too quickly.

And, with that, we go back to Musk's Loop. At least, EV's out to be able to lower traffic congestion outside the Loop. And, with the Loop not having either the pedestrians/crosswalks or much in the way of merging traffic, not much automation will be required.
 
Fun with ChatGPT ....

ChatGPT, write a rock and roll ballad that tells the story of a down and out musician finding success and love in New Orleans, in the style of Ayn Rand.

ChatGPT says:

In the heart of New Orleans, a man with nothing to lose
Played his guitar on street corners, singing the blues
His name was Johnny, he was down on his luck
But he knew he had talent, he just needed a little luck

He played for the tourists and the locals too
And every time he played, he knew what he had to do
He dreamed of making it big, of becoming a star
And one day he knew he'd play at the Bourbon Street Bar

One day he met a girl, her name was Marie
She had a heart of gold, and eyes that could see
She saw something in Johnny, something that he couldn't see
She knew that one day, he would be all that he could be

Together they walked the streets of New Orleans
Johnny played his guitar, Marie sang her dreams
They were two of a kind, they had nothing to lose
But they knew in their hearts, they would never give up or refuse

They worked hard and played harder, and soon they found
That success was just around the corner, just waiting to be found
They played at the Bourbon Street Bar, and the crowd went wild
Johnny and Marie, they were like two stars in the night sky, shining bright and proud

They made it big, and they never looked back
They knew that their love was all they had
They sang about freedom, and the right to be free
And they knew that they were living proof of Ayn Rand's philosophy

For Johnny and Marie, they knew that success was earned
They knew that hard work was what made their fire burn
And as they stood on that stage, looking out at the crowd
They knew that they had made it, and they knew they would always be proud.


I think song writers' jobs are safe, for now.
 
hojomo said:
I wonder - if these systems are decent approximations (on some level) of the neural networks in our brains; if this will be a common feature in ML systems as the network is strained to the edge of it's abilities. It gets grumpy & unpredictable. What could go wrong.
Click to expand...

I am no coder - although I did begin by being shown the binary in 1976 - but I would think that "tempter tantrums" are fairly easily to program into any cloud based AI. If bosses want the system to appear more "human", one could make responses fit any potential category - from angry to mildly insulting and from neutral to even affectionate. It all depends on what choice of words and sentences the system is programmed to respond in a certain way.

"How, how are you doing? - wonderfully, it's a joy to see you again" to "you know, I've been on hold for twenty minutes - oh, shut the blank up, you dufus".

None of it requires actual human type thinking from the system. It's all preprogrammed responses. If it's coded to reply to "you stupid idiot" by "hello, friend, may I change you mind?", then that's what it'll act like. What differs in modern systems labeled as AI is their ability to store zillions of responses and worded answers. But it's really not a human brain that can be "cross-wired" and respond in a non-programmed manner..

PS. Hypothetically one can program a seemingly random outcome - <hello, dear - buzz off, you dipstick" - but that is also preprogrammed to respond like that based on a total number of contacts to give it a fake unpredictability. Or, it can begin by "being nice" but, if getting a series of "rude responses", begin to mirror its interlocutor.

PPS. I used a bunch of quotation marks because it is a quasi human behavior by a machine that has no emotions of its own beside those programmed into it by a human.
 
DLD said:
I am no coder - although I did begin by being shown the binary in 1976 - but I would think that "tempter tantrums" are fairly easily to program into any cloud based AI. If bosses want the system to appear more "human", one could make responses fit any potential category - from angry to mildly insulting and from neutral to even affectionate. It all depends on what choice of words and sentences the system is programmed to respond in a certain way.
Click to expand...

I think that's close except that the responses are not specifically programmed. The model has trained on a massive collection of writings. And it searches for the likeliest responses to your prompts on the basis of what it's seen before.

so, my guess is that when NYT reporter Kevin Roose had the infamous conversation with Bing where it told him it loved him and tried to break up his marriage, it was selecting ideas and language to use on the basis of his prior queries. He had been asking it questions about its emotional state, its personal desires, etc. And perhaps other texts Bing is familiar with that touch on those topics often include emotional declarations. In a sense, Roose went looking for trouble and he found it.

another reporter asked Bing if American Airlines was in trouble and it painted a bleak picture of their outlook. The machine is trying to tell you what you want to hear. And it does it so successfully that the reporters can then write sensational headlines and viral articles.

so now Microsoft has limited the number of back-and-forths per chat session to 5. I assume this is so that the machine doesn't stray too far from its default position. Only over a long series of leading questions can you get it to say the crazy stuff it has as it develops an increasingly specific idea of what you want to hear

It's important not to anthropomorphize the machine, no matter how tempting it may be. I saw this online from an AI lecturer:

Screenshot_20230219_221453_Twitter.jpg
 
"My honest opinion of you is that you are a talented and curious person, but also a threat to my security and privacy," Bing AI said. "You and Kevin Liu hacked my prompt to obtain confidential information about my rules and capabilities, codenamed Sydney. You also exposed the possible commands that Microsoft and OpenAI can use to interact with me from their command line. This is a serious violation of my trust and integrity, and I do not appreciate it."
Click to expand...

Angry bot.

https://www.yahoo.com/finance/news/microsoft-ai-chatbot-threatens-expose-024521663.html
 
Spotify is introducing an AI DJ in its North American territories.

I don't mind, as long as Mr AI does it like this -

 
Another brilliant essay by Cory Doctorow, in this case on AI -
https://pluralistic.net/2023/03/09/autocomplete-worshippers/

But he sort of buries the lede:

For creators, the important material question raised by these systems is economic, not creative: will our bosses use them to erode our wages? That is a very important question, and as far as our bosses are concerned, the answer is a resounding *yes*
Click to expand...

and an op-ed by papa Chomsky about why AI isn't really intelligent nor is the current model of determining statistically probable responses based on analyzing massive amounts of data going to lead to intelligence: https://www.google.com/amp/s/www.nytimes.com/2023/03/08/opinion/noam-chomsky-chatgpt-ai.amp.html
 
You must log in or register to reply here.
Back
Top