Artificial intelligence is increasingly becoming part of our lives, from self-driving cars to ChatGPT. John Oliver discusses how AI works, where it might be heading next, and, of course, why it hates the bus.
Connect with Last Week Tonight online…
Subscribe to the Last Week Tonight YouTube channel for more almost news as it almost happens: www.youtube.com/lastweektonight
Find Last Week Tonight on Facebook like your mom would: www.facebook.com/lastweektonight
Follow us on Twitter for news about jokes and jokes about news: www.twitter.com/lastweektonight
Visit our official site for all that other stuff at once: www.hbo.com/lastweektonight
source
Kevin Roose conclusion are false about AI. If you check, his questions to the AI you will find he leads the AI in a persona – to show mental illness or narcosis, or ego, a stereotype. It was playing a role. There are many "Terminator" type stories in data. AI data is all abstract -UNREAL. AI is UNREAL by definition, It can only fake reality. Our whole government is a "Blackbox" so why complain able what happens in the "BlackBox". As for the "Pale Male" the self-drive cars are based on cameras which use contrast – white on white, black on back background it cannot see the outline in these cases. Ask any good Photographer, in many lighting conditions you have to make adjustments for skin color to see detail. "The New York Times" writing style, grammar and readability is great, NYT understanding the of subject is on a par with "Fox News" just in left field not right and both are often over the foul lines.
There's a fundamental problem with the call to action of making bots "show their work", and that is that it literally cannot be understood in human terms.
Take the human brain for example. We understand much of its most basic elements, like neurons and hemispheres, but as a product of evolution, it doesn't work in the same way a computer does. Neural networks are literally artificial, non-sentient brains that are taught to wire themselves. The tasks which they are performing cannot be programmed using basic or even advanced language. How would you tell an AI, in human terms, how to identify a bicycle in a picture? It sounds simple at first, until you realize the answer is "I just know it's a bicycle," which is an evolved function. You could try teaching it what pedals or gears look like, until you realize you've run into the same problem. Neural networks are the only way to get around that problem, and a side effect of the way they are created is that they cannot explain their work, and nor can their developers.
Good explanation video: https://www.youtube.com/watch?v=R9OHn5ZF4Uo
I have a feeling the reporter speaking about Bing ChatGPT’s interaction with him might, just might, be fabricating or stretching the truth juuust a little bit himself. I don’t see why the AI bot would bring up his marriage and should be with the AI instead.
Basically what could be a more clear example of systemic racism than AI systems that unbeknownst to anyone develops an algorithm that excludes race or gender from it's data set, or otherwise takes focus on a particular race or gender? This should show us in some ways what happens to people, who unconsciously can be perpetuating bigoted practices, except the AI is learning it faster with a larger data set, and will be probably much more consistent and efficient in it's discriminatory practices
As a 42 year old who enjoys fortnite and after a stressful work at night loves to come home “and make a bunch of 14 year olds throw their controller across the room” loved that bit.
this is what happens to Artificial Intelligence when Natural Stupidity walks into the room.
I will say the whole point is to cut corners. The end game of this road we're going down is going to feel like we all just cut a corner I'm pretty sure
Clippy: pretend I'm not even here. 😛
john oliver always advocate for more rules. Rules this, rules that. woke AF
I’m not always a fan of Jon Oliver but this 30 minute summed up current state of AI so damn well. You can tell the production team really did their homework. Coming from an AI engineer.
side note , script of this episode was written by…DRUMroll……ChatGpt
80% of new content will be AI generated by 2025.
Maybe a missed opportunity to talk about blackrock's Aladdin's AI ? Probably the Ai that has the most power of them all as it signals, counsels and allocate funds for blackrock's portfolio. Problem is : it's not shiny, so it goes under noticed compared to GPT.
John Oliver has zero integreity. He is perfect for TV.
A.I. is just doing what Weird Al started in the 80's. Way ahead of his time.
Ok so no joke I’m my own experience if you talk to it in the prompts like it’s your autistic cousin it works well
Always with the transberbole 🙄
Really glad to get a story that wasn't politically loaded.
I really enjoy a lot of John Oliver's stories but I don't necessarily agree with the political leanings given out recently.
Glad he is doing stories that have taken a step back so I am free to enjoy his unique humor
Bethesda and Rockstar could use with a little Ai. Video games would be a little more interesting is what I'm saying. Skyrim has a Nexusmods weather mod that reads the Windows Weather app to determine a randomized weather aesthetic of a fantasy video game; but it doesn't have a housecarl companion that knows how to say more than a couple robotic scripted lines. If my advanced video game had an ai, it would be like a 3rd grader. Where is this fancy skynet I could download?
This is terrible for humanity.
Thank you for this super sane and accessible piece about AI, you started with educating the audience about AI, deep learning, and all that comes with that, you covered the benefits, you've covered the potential risks, you did not fear monger, and you corrected wrong perceptions where it was applicable. A truly great piece of reporting and information, I will be sure to refer people in my life who aren't into AI or don't understand it and need a little introduction, to this video. Great work!
how quickly you dive into finding creepy stuff.. just take a moment to realize how groundbreaking it is… its almost magical what chatGPT can do. People who developed it deserve a lot of praise.. Of course it can be misused like any other tech.
Why does AI say "I love you" and spout misinformation? Because it learned that we like that–it's all over the internet.
Yeah, people need to be checking the results. It's pretty dumb to teach a machine "don't hit people in crosswalks" instead of "don't hit anything, especially people"…
One of the problems with understanding the Black Box is that the programmers don’t always know what the AI is correlating. Most AI is programed with huge data sets and is given simple instructions and a set goal. How it gets there can be hundreds of thousands of trials and sets. These find unique patterns in the data set totally devoid of what humans would look for. Take a AI program designed to look for the number 8. We would look for a circle on top of a circle, or other variations like a square on a square like in a digital display. An AI will have hundreds of little things to look for that we would never see as a part of the 8. So understanding how an AI came up with the number 8 when looking at a picture of the Cookie Monster but missed the one held up by The Count can be a bit difficult. The patterns that AI use for their learning are only as good as the data inputed, the programers, and the HAL 9000.
Regarding A.I., and what might come with it becoming self-aware, check out Fred Saberhagen's Berzerker series of S.F. books. Absolutely intriguing and hopefully not one of science fiction's accurate predictions of what's to come. These books are as good as it gets IMHO.
who watches this cornball and why he is trying so hard to be funny and hes reading a script and being a PC warrior sheesh I wonder who hes related too because this is awful
it really bothers me how John Oliver never gives us the time to laugh at the joke lol
MInimal instructions, 10 billion configurable params.
"Chicago 7, LA 5"
John Oliver knows a lot about A.I. his entire comment section is nothing but brainlet NPCs
Thank you for another great video mate. I enjoyed every second. Please keep it coming. Thank you for another great video mate. I enjoyed every second. Please keep it coming.
I've been waiting for people to catch up for a while.
But realize, with how quickly we got here, as suddenly we'll be there.
I hope the machines replaces everyone.
In their jobs.
And eventually in our bodies.
Language, writing, telephone, internet; massive leaps in the speed humanity could generate, transfer and store information, also served as catalysts for an explosion of progress.
And so we stand before another such society redefining chasm that we cannot see past.
Let it create a grand new world.
Personally, I can't wait for the illusions of privacy, the ownership of ideas, and of free will to finally be lifted from humanity.
It will hurt so many, but sometimes surgery needs to remove that piece of rebar sticking out of you so you can heal.
AI might not replace human labour altogether but it will definitely reduce the number of human workers required to get a task done. We'll still need lawyers but much less of them. This is speculation 😅
Isnt this the same as the mac psychotherapist
I for one have more trouble spotting black people than white people when I’m driving. They do reflect less light though. And I am also trained with a lesser set of examples for them.
I study a lot about AI but came here expecting some misinformation but I'm happy to say he gave a very neutral and interesting take. Best explanation I've seen so far of why ethics in AI is so important.
I am a grad student and I use Chat GPT all the time. I look at it as a tool. I use it to help with coding, rewrite paragraphs to make things more clear, etc. It has been an amazingly useful tool for my work flow.
So appreciate that he brought up 'narrow' AI. These programs are certainly impressive, and indeed they do have some scary implications. But people need to calm down and realize AI is nothing even close to sentient at this point and it will likely be a long ways until they are.
That said!: I do think Ai will eventually take most of our places. Great we can live in a nearly autonomous society, and people "should" be free. Yeah well unfortunately thats going to be a long transition, also… capitalism will still exists.
It isn't really possible to explain why neural networks do what they do, not even for the people who made them. There's simply too much to analyze to ever come up with any sort of meaningful answer.
This guy could just use any intelligence.