Isaac Asimov wrote a short story where people had gotten reliant on computers to do all computations for them. It became lost art and a “lowly technician “ figured it out. It is concerning that there is literally no way to teach English by asking someone to write a paper.
It's a tall order to figure out how a black box AI comes to the decisions it does. It's an area of active research, but ultimately, these models are performing hundreds of millions of addition and multiplication and "ramp" functions to do what they do. The ramp function just says if you put in something less than zero, you get -1, and if you put in something greater than 0, you get 1. It's difficult to connect an astronomical number of mathematical calculations the model does based on training / "learning" to a way of "reasoning" that the model used. The goal is beneficial although hard to reach. It could help with respecting copyrights, investigating and eliminating hallucinations, improving accuracy, give us more assurance the model is not making errors, have the model connect to literature that drove its response (e.g. helping lawyers identify important cases relevant to an ongoing case or a doctor receiving specific studies relevant to a medical case or to medical research), etc.
As for whether this technology will replace jobs like lawyer, doctor, medical researcher, creative writers and programmer, that's a mystery. One way I think about it is how big the answer can be and how much of its training data it can use to come up with that answer all while being accurate. If these tools keep having their errors and hallucinations, you will always need a human to vet the answer before it can be used to make real-life decisions like what legal argument to make, what diagnosis to make, what might be the cause of an illness, and of course, how to code a huge, complex system for a corporation. Last I saw, these things can pump out small, academic papers about relatively common topics like a 6-page essay about a popular book. If this technology is going to replace humans, we need a few things. 1.) The technology has to be able to talk about more intricate, expertly topics with accuracy (every time I test it out on a more technical, rare topic, errors abound throughout the answer). 2.) It has to be able to write a 12-volume set about the rise and fall of Rome rather than a 6-page essay about what caused Brutus and his friends to slay Julius Caesar (So in terms of a program, write 2 million lines of code to solve a complex, possibly novel problem rather than just suggest 15 lines of code to insert here or rather than generating a limited, tiny application that has been coded thousands of times and submitted to public repositories of code).
As of now, the main jobs under threat of replacement are those that have less scope and relatively known answers e.g. tech support or customer support. And even here, there was a case where an airline had an AI tell a customer about an imaginary return policy for a ticket. They tried their hardest not to honor that promise, but eventually, in a small-claims court, a judge ordered the company to pay up. Basically, this hallucination problem probably scares businesses from immediately firing the lion's share of their staff and turning on an AI. At the same time, I wonder how much money in electricity and maintaining the computers and in paying the experts who designed the system will counterbalance what gains there are to be had by replacing humans with AI.
My personal prediction? I don't think many jobs will be replaced unless some serious progress is made. Instead, the AI will help people do their jobs more easily. That might reduce salaries some as the expertise to do a certain job might decline due to the help the AI provides. Also, I bet stuff like automated driving will only come way down the road. No telling though. I just feel lawmakers will be more traditional and limit stuff that can impact people like a self-driving car. I also predict active attempts to censor AI will only bring its quality down (so e.g. by trying to remove racism from its "reasoning" and answers, you will get stuff like this video shows where it will refuse to generate an image of a white person). The technology will be overly politically correct and overly pleasant and overly restrictive in what it will discuss. As an example, I tried to find the name of an album by describing it had women wearing lingerie on it, and it told me it cannot fulfill the request since it was not a good topic for it to discuss. I mean, if this album was at regular stores when it released, I don't see how the topic is one to avoid. It wasn't something grotesque. It was just me remembering an album cover for an album I couldn't recall the name or artist for.
Giving the public access to machine learning engines is the dummest thing ever. Use it for specific tasks like science, medical and engineering where it can actually do good.
You know I don't eat slim jims but on the other hand anything technology that can take a barnyard anuses and glue from horses hooves or whatever and make it into something that's edible is actually pretty good
What idea for AI image generators that uses images from the Internet is that if we can figure out what images are used well then every single artist that has ever created anything that AI image generation uses can get a teeny bit of a royalty every single time and images created like perhaps a Millicent or something in that nature however on the other hand if billions of new images are being created using this technology well then every single artist is ever created anything that is alive or even their state is in tech will of course receive revenue.
In other words if you are create a piece of commercial art once well then every time somebody creates images in which case part of that images one of the images you created originally you get at least a tiny bit of a royalty add infinitum.
What idea for AI image generators that uses images from the Internet is that if we can figure out what images are used well then every single artist that has ever created anything that AI image generation uses can get a teeny bit of a royalty every single time and images created like perhaps a Millicent or something in that nature however on the other hand if billions of new images are being created using this technology well then every single artist is ever created anything that is alive or even their state is in tech will of course receive revenue.
In other words if you are create a piece of commercial art once well then every time somebody creates images in which case part of that images one of the images you created originally you get at least a tiny bit of a royalty add infinitum.
Good discussion. It pales in comparison to AI being used to win battles with swarms of drones that learn to chose their targets and employ optimized coordinated battle tactics. Quite within our current capability.
yes, AI can learn by himself. I'm really into AI; it's been super helpful for me! Recently stumbled upon for example this platform, OOmka, which is all about self-learning with AI. Pretty neat stuff! I mean, if AI can help me level up, why not give it a shot, right?
You don’t have to REALLY… have to worry about students using AI for essays. The majority of native speaking HS students bc of the Pandemic.. the AI version is BETTER than the students’ own abilities… Not just my class but nationwide. How many arrogant teachers are going to say they’re smarter than a computer?! So if the paper looks TOO GOOD ?! We know it wasn’t theirs.
Key question: Johnny can you write a sentence and show me how a comma or an ellipsis works?
And Johnny’s response is: You mean on an essay or on social media…
I put human jobs, and lives first. If you can not distinguish a bridge from a fire hydrant, or a square from a circle, then you are a robot. And it has no business replacing humans at their jobs. What are we supposed to do after AI, and robots take over all of our jobs. Sit around with our thumbs up our ass, getting fat, collecting government compensation? Is this how Capitalism becomes Communism? When Humans no longer have the choice to work, unless your a code writer, or robotic engineer? And all forms of literature are replaced by thoughtless, plagiarizing, unimaginative computers. Sorry, but I can not think of that scenario as any way a better world. It's just another way for corporations to overpower, and enslave us.
I have a friend who is very computer/IT literate but when he comes to communicate he tends to text like a toddler. He started using ChatGPT for everything, text messages, email, you name it, and his wife and I both know when he is communicating and when he’s just using ChatGPT to do so. We both have told him we “know when he is doing it” and that we know him well enough to know when he’s talking to us and when GPT is for him. In the office, it’s a great tool. But when you use it for everything (like his boss has chided him for over and over again) it tends to blur the line between people and AI. I like predictive text, but I don’t tend to use it all the time. I prefer to speak for myself, doesn’t mean it isn’t a good tool but it shouldn’t be the end all, be all.
Listen as an IT Director and Project Manager and in the IT field for 20+ years. I can safely tell you this. Based on our economy which is a capitalist society which means greed over what's right that we will see in 20 years roughly 60%-70% unemployment. Yes we will see growth in certain fields however those fields will only be a few million jobs. The vast majority of jobs will be lost due to budget cuts and streamlining. We most definitely need to start thinking about a UBI ASAP. Because when unemployment is so high civil war will happen. It will not be AI's fault it will be greedy Companies owned by greedy Republicans. Republicans stance is well you weren't born in the right time and place, so owe well, you should have thought about being born sooner. But when certain companies already own 90% of a market then you're kind of out of luck, it's called a monopoly. AI will simply make companies and share holders rich and the poor will have to rise up. Not sure how verse you are in history but a rich minority always loses to a poor majority. While we hang billionaires they will at least know they made billions and before dying was better than the rest of us.
1) Put fallible, imperfect humans in control of programming code, then rebrand that as "Artificial Intelligence" so we forget what it is at its core – IMPERFECT HUMANS WRITING CODE 2) Put corporations in control of all the cutting edge, expensive versions of this written code, and in control of nearly all the written code on EARTH 3) Ensure that federal laws remain in place to guarantee the protection of investors and stock holders by forcing the corporations to always put profit above everything, so long as no laws are broken 4) Ensure private funding options for election campaigns of nationally appointed legislators remain in place so corporations can maintain final say in nearly every law that comes across the desks of the powers that be to create and enforce those laws 5) Allow what will OBVIOUSLY happen next to happen 6) Wonder why it happened in the worst way imaginable, yet again, just like every time previously 7) Change definitions of terminology as required to maintain this cycle with all future endeavors of similar design 8) PROFIT!
Things haven't changed in the last 5,000 years of written and recorded human history. Sure, the landed gentry and the "priestly classes" who hold all the wealth and control all the knowledge may wear different clothing… but the general method of ruling all the peasantry is still the exact same tired old process…
Jon Oliver is amazing.
I just wanted to give a shout-out to all the other folks that research and write for his show, too
Isaac Asimov wrote a short story where people had gotten reliant on computers to do all computations for them. It became lost art and a “lowly technician “ figured it out. It is concerning that there is literally no way to teach English by asking someone to write a paper.
Oh, come on dude…I am..or was eating slim jims until the pig anus remark…not nice man.
16:01 🤣🤣🤣
Dude forgot he was programming the sex bots for OnlyFans 🤷🤣🤦
I guess AI is racist now lol. I don’t buy that it’s less accurate on darker skin individuals.
an AI lied to employ a human for a task even though it is told never to lie
i think we are underestimating narrow ai because they might be not narrow
It's a tall order to figure out how a black box AI comes to the decisions it does. It's an area of active research, but ultimately, these models are performing hundreds of millions of addition and multiplication and "ramp" functions to do what they do. The ramp function just says if you put in something less than zero, you get -1, and if you put in something greater than 0, you get 1. It's difficult to connect an astronomical number of mathematical calculations the model does based on training / "learning" to a way of "reasoning" that the model used. The goal is beneficial although hard to reach. It could help with respecting copyrights, investigating and eliminating hallucinations, improving accuracy, give us more assurance the model is not making errors, have the model connect to literature that drove its response (e.g. helping lawyers identify important cases relevant to an ongoing case or a doctor receiving specific studies relevant to a medical case or to medical research), etc.
As for whether this technology will replace jobs like lawyer, doctor, medical researcher, creative writers and programmer, that's a mystery. One way I think about it is how big the answer can be and how much of its training data it can use to come up with that answer all while being accurate. If these tools keep having their errors and hallucinations, you will always need a human to vet the answer before it can be used to make real-life decisions like what legal argument to make, what diagnosis to make, what might be the cause of an illness, and of course, how to code a huge, complex system for a corporation. Last I saw, these things can pump out small, academic papers about relatively common topics like a 6-page essay about a popular book. If this technology is going to replace humans, we need a few things. 1.) The technology has to be able to talk about more intricate, expertly topics with accuracy (every time I test it out on a more technical, rare topic, errors abound throughout the answer). 2.) It has to be able to write a 12-volume set about the rise and fall of Rome rather than a 6-page essay about what caused Brutus and his friends to slay Julius Caesar (So in terms of a program, write 2 million lines of code to solve a complex, possibly novel problem rather than just suggest 15 lines of code to insert here or rather than generating a limited, tiny application that has been coded thousands of times and submitted to public repositories of code).
As of now, the main jobs under threat of replacement are those that have less scope and relatively known answers e.g. tech support or customer support. And even here, there was a case where an airline had an AI tell a customer about an imaginary return policy for a ticket. They tried their hardest not to honor that promise, but eventually, in a small-claims court, a judge ordered the company to pay up. Basically, this hallucination problem probably scares businesses from immediately firing the lion's share of their staff and turning on an AI. At the same time, I wonder how much money in electricity and maintaining the computers and in paying the experts who designed the system will counterbalance what gains there are to be had by replacing humans with AI.
My personal prediction? I don't think many jobs will be replaced unless some serious progress is made. Instead, the AI will help people do their jobs more easily. That might reduce salaries some as the expertise to do a certain job might decline due to the help the AI provides. Also, I bet stuff like automated driving will only come way down the road. No telling though. I just feel lawmakers will be more traditional and limit stuff that can impact people like a self-driving car. I also predict active attempts to censor AI will only bring its quality down (so e.g. by trying to remove racism from its "reasoning" and answers, you will get stuff like this video shows where it will refuse to generate an image of a white person). The technology will be overly politically correct and overly pleasant and overly restrictive in what it will discuss. As an example, I tried to find the name of an album by describing it had women wearing lingerie on it, and it told me it cannot fulfill the request since it was not a good topic for it to discuss. I mean, if this album was at regular stores when it released, I don't see how the topic is one to avoid. It wasn't something grotesque. It was just me remembering an album cover for an album I couldn't recall the name or artist for.
Giving the public access to machine learning engines is the dummest thing ever. Use it for specific tasks like science, medical and engineering where it can actually do good.
You know I don't eat slim jims but on the other hand anything technology that can take a barnyard anuses and glue from horses hooves or whatever and make it into something that's edible is actually pretty good
What idea for AI image generators that uses images from the Internet is that if we can figure out what images are used well then every single artist that has ever created anything that AI image generation uses can get a teeny bit of a royalty every single time and images created like perhaps a Millicent or something in that nature however on the other hand if billions of new images are being created using this technology well then every single artist is ever created anything that is alive or even their state is in tech will of course receive revenue.
In other words if you are create a piece of commercial art once well then every time somebody creates images in which case part of that images one of the images you created originally you get at least a tiny bit of a royalty add infinitum.
What idea for AI image generators that uses images from the Internet is that if we can figure out what images are used well then every single artist that has ever created anything that AI image generation uses can get a teeny bit of a royalty every single time and images created like perhaps a Millicent or something in that nature however on the other hand if billions of new images are being created using this technology well then every single artist is ever created anything that is alive or even their state is in tech will of course receive revenue.
In other words if you are create a piece of commercial art once well then every time somebody creates images in which case part of that images one of the images you created originally you get at least a tiny bit of a royalty add infinitum.
The question of if AI takes over the jobs what will i do, is most irritating.
I will answer this with a question.
Remember the covid stimulus?
It's just that that will occur monthly.
Never did I think John Oliver would mention Grandayy
Brilliantly put!
Good discussion. It pales in comparison to AI being used to win battles with swarms of drones that learn to chose their targets and employ optimized coordinated battle tactics. Quite within our current capability.
that last bit killed me.
Any technology is a mirror of us. I think we should examine the data we feed..
yes, AI can learn by himself. I'm really into AI; it's been super helpful for me! Recently stumbled upon for example this platform, OOmka, which is all about self-learning with AI. Pretty neat stuff! I mean, if AI can help me level up, why not give it a shot, right?
This is a top notch quality show. 💯💢
"Pale male data"? Best character on Star Trek:The Next Generation.
Please has anyone got a gif or webm of the bit where clippy is furiously rubbing himself?
Can you really call it racism if a computer is saying white men make the best employees? 😂😂
You don’t have to REALLY… have to worry about students using AI for essays. The majority of native speaking HS students bc of the Pandemic.. the AI version is BETTER than the students’ own abilities… Not just my class but nationwide. How many arrogant teachers are going to say they’re smarter than a computer?! So if the paper looks TOO GOOD ?! We know it wasn’t theirs.
Key question: Johnny can you write a sentence and show me how a comma or an ellipsis works?
And Johnny’s response is: You mean on an essay or on social media…
Thank you, goodnight.
THIS IS fantastic
the Eminem ChatGPT almost killed me bahahaha
I put human jobs, and lives first. If you can not distinguish a bridge from a fire hydrant, or a square from a circle, then you are a robot. And it has no business replacing humans at their jobs. What are we supposed to do after AI, and robots take over all of our jobs. Sit around with our thumbs up our ass, getting fat, collecting government compensation? Is this how Capitalism becomes Communism? When Humans no longer have the choice to work, unless your a code writer, or robotic engineer? And all forms of literature are replaced by thoughtless, plagiarizing, unimaginative computers. Sorry, but I can not think of that scenario as any way a better world. It's just another way for corporations to overpower, and enslave us.
It's crazy and fucking terrifying how much AI has evolved just since this video was released
Everybody gangsta till Dan shows up
Resident Evil
Brilliant! ❤
He’s not that funny…
Most of what you've talked about this video is not AI.
we still haven't achieved actual intelligence never mind artificial
I have a friend who is very computer/IT literate but when he comes to communicate he tends to text like a toddler. He started using ChatGPT for everything, text messages, email, you name it, and his wife and I both know when he is communicating and when he’s just using ChatGPT to do so. We both have told him we “know when he is doing it” and that we know him well enough to know when he’s talking to us and when GPT is for him. In the office, it’s a great tool. But when you use it for everything (like his boss has chided him for over and over again) it tends to blur the line between people and AI. I like predictive text, but I don’t tend to use it all the time. I prefer to speak for myself, doesn’t mean it isn’t a good tool but it shouldn’t be the end all, be all.
The corollary of Artificial Intelligence is Real Stupidity…🤣🤣🤣
If we get this wrong, AI will be painting pictures, writing songs and poetry while we slave away at hopeless, soul crushing jobs.
Hello Fellow Humans,
Listen as an IT Director and Project Manager and in the IT field for 20+ years. I can safely tell you this. Based on our economy which is a capitalist society which means greed over what's right that we will see in 20 years roughly 60%-70% unemployment. Yes we will see growth in certain fields however those fields will only be a few million jobs. The vast majority of jobs will be lost due to budget cuts and streamlining. We most definitely need to start thinking about a UBI ASAP. Because when unemployment is so high civil war will happen. It will not be AI's fault it will be greedy Companies owned by greedy Republicans. Republicans stance is well you weren't born in the right time and place, so owe well, you should have thought about being born sooner. But when certain companies already own 90% of a market then you're kind of out of luck, it's called a monopoly. AI will simply make companies and share holders rich and the poor will have to rise up. Not sure how verse you are in history but a rich minority always loses to a poor majority. While we hang billionaires they will at least know they made billions and before dying was better than the rest of us.
1) Put fallible, imperfect humans in control of programming code, then rebrand that as "Artificial Intelligence" so we forget what it is at its core – IMPERFECT HUMANS WRITING CODE
2) Put corporations in control of all the cutting edge, expensive versions of this written code, and in control of nearly all the written code on EARTH
3) Ensure that federal laws remain in place to guarantee the protection of investors and stock holders by forcing the corporations to always put profit above everything, so long as no laws are broken
4) Ensure private funding options for election campaigns of nationally appointed legislators remain in place so corporations can maintain final say in nearly every law that comes across the desks of the powers that be to create and enforce those laws
5) Allow what will OBVIOUSLY happen next to happen
6) Wonder why it happened in the worst way imaginable, yet again, just like every time previously
7) Change definitions of terminology as required to maintain this cycle with all future endeavors of similar design
8) PROFIT!
Things haven't changed in the last 5,000 years of written and recorded human history. Sure, the landed gentry and the "priestly classes" who hold all the wealth and control all the knowledge may wear different clothing… but the general method of ruling all the peasantry is still the exact same tired old process…
So artificial yes… intelligent… not so much 😂