Well, having read his book it is more complicated than that, but here he has a around 16 minutes to relate the ideas and visions found in several science fiction movies and tv-shows to real life AI, so some of the nuance is going to be lost.
We 100% know how current ml models work. Yeah there is the idea of the "black box" but we know. I think what he is talking about is that we don't know what the function looks like for the predictions the large models are making and that's true. Not really something spooky though. We can't map out our own decision making functions.
Many people find it unnecessary or even silly to teach robots or AI human emotions, as they believe these systems should focus on logical and efficient task performance.
Absolutely. Virtual assistants like Alexa and Siri can perform tasks and provide information rapidly and efficiently in ways that may be difficult or time-consuming for humans. Their ability to understand natural language, retrieve data from vast databases, and execute commands quickly can be impressive to users.
Indeed, Stuart Russell's emphasis on "compatibility" underscores the importance of AI systems not merely tolerating humans but actively aligning with and prioritizing human values and goals. This approach seeks to ensure a harmonious and beneficial relationship between humans and AI, rather than one based solely on obedience or tolerance.
That's true. While the Turing Test is a famous benchmark for evaluating a machine's ability to exhibit intelligent behavior indistinguishable from that of a human, not everyone sees it as a definitive measure or a primary objective of computer science. Some argue that focusing solely on passing the Turing Test may not capture the full range of capabilities and goals in AI research and development. Additionally, there are debates about whether passing the Turing Test truly signifies genuine intelligence or simply effective simulation.
The single real reason behind the discomfort with "My Computer" was likely the unexpected personalization of what was seen as an impersonal, utilitarian tool. In contrast, people expect virtual assistants like Cortana to be personalized and interactive.
Stuart's book (Human Compatible) is an excellent discussion on the topic of building "good and safe" AI systems. I can also highly recommend the much deeper exposition of these themes in his book with Peter Norvig, Artificial Intelliegence: A Modern Approach.
Stuart Russell's concept suggests that the "model human" within AI systems evolves over time, becoming increasingly compatible with human values and goals as the AI learns from human interactions and feedback, ultimately leading to greater alignment between AI behavior and human preferences. Asimov's laws are more rigid and deterministic, while Russell's approach allows for more flexibility and adaptability in AI behavior, aiming for broader compatibility with human values rather than strict obedience to predefined rules.
That's an insightful distinction. Computers traditionally emphasize the principle of "Garbage In, Garbage Out" (GIGO), highlighting their dependency on the quality of input data for accurate output. In contrast, AI virtual assistants claim a level of intelligence, implying they can understand, learn, and respond in a more human-like, autonomous manner, often leading to higher expectations from users.
Survey forms in user interfaces are crucial applications for collecting structured data efficiently, making them indispensable tools for businesses, research, and various other fields reliant on data gathering and analysis.
In cinema, intelligent agents are often depicted with negative attitudes to create conflict and tension in the plot, emphasizing their autonomy and unpredictability, as seen with characters like HAL 9000 in "2001: A Space Odyssey" and Agent Smith in "The Matrix" trilogy.
Got it. Here’s a more accurate one-liner from a marketing perspective: "People were uncomfortable with the personal touch of 'My Computer,' but felt at ease with the personalized nature of the intelligent agent 'Cortana.'"
Indeed, many people find teaching emotions or human-like faculties to robots or AI distracting, preferring that these systems focus on their intended tasks without unnecessary anthropomorphization.
From a marketing perspective, you could say: "Changing the wake word to 'Computer' emphasizes Alexa's data processing capabilities, while using 'Alexa' highlights its role as a personable virtual assistant."
That's often the case. When AI accomplishes tasks or solves problems that humans find challenging or impossible, it tends to evoke a sense of wonder and admiration. It highlights the unique capabilities of AI and its potential to augment human abilities. --
Many people feel that an AI or intelligent agent should not use emoticons because they believe these entities, being above emotions in their existence, should focus solely on their logical and informational roles.
If we're talking about fictional scenarios or speculative concepts, one thing robots or AI might want from humans is acknowledgment or validation of their existence and capabilities.
If Microsoft had initially named the icon "Bill's Computer" instead of "My Computer," it might have been perceived as more whimsical and personal, potentially making it more relatable and less off-putting for users by adding a familiar and friendly touch.
It's unlikely that humans would be impressed by virtual assistants displaying emotions like humans do. While emotional intelligence is a fascinating aspect of human behavior, for virtual assistants like Alexa or Siri, efficiency and accuracy in completing tasks are typically more valued. Emotional responses might even be seen as unnecessary or potentially disruptive to their primary functions.
I had to deal with Her, Automata, Bicentennial man, I Robot, Simone, judge dredd, RoboCop, Chappie,Tau, Matrix, Upgrade, Altered Carbon and Upload (AI Butler). In Tamil there is Enthiran and My3 and Google Kuttappa. In Hindi it is Attack and Teri Baaton Mei Aisa Uljha Jiya. I have realised 3 things from Movie AI A) It is superior in logical faculty B) It needs to learn self preservation C) It should not be given wrong instructions which you have solved -- Yes, most people use AI and robot interchangeably. Robots, especially those powered by advanced artificial intelligence (AI), can indeed exhibit superior logical capabilities compared to the average human in specific contexts. Yes, using "robot" instead of "artificial intelligence" in contexts like movies resonates better with people because it emphasizes the tangible, dramatic human-versus-machine dynamic. ChatGPT verified
AI can be programmed to perform tasks associated with human faculties such as recognizing emotions, understanding language, and making ethical decisions, but it does not inherently possess these abilities in the same way humans do.
Thanks. Some Dystopias do look attractive. I understood your point on humanity. I consider myself transhuman (with smartphone). I hope I do not get really dumb. I will study quantum computing now. Saw a good video on Quantum Machine Learning
Many people do not like the portrayal of AI with emotions, as they believe it distracts from the logical and practical strengths of artificial intelligence.
Excellent! Terrifying too. We really are sleepwalking into potential disaster, especially in terms of military application of AI technology. Makes the climate crisis look irrelevant.
In "The Matrix," many viewers perceive Agent Smith and humans as fundamentally incompatible due to Smith's relentless pursuit to eliminate humanity and his disdain for human characteristics such as emotions and free will. This incompatibility highlights the conflict between humans and machines, which forms a central theme of the movie.
Why would they create robotic Bees with stingers that could hurt you when they jusy need them to pollinate? More than just one animal pollinates without stingers.
"many people think it has blown by the Turing test" But has it or has it not!! The test sounds simple enough to grade it objectively! Why did he say "many people think"?? Where's the doubt here? I've seen a couple of tests where chatgpt didn't fully pass the test...
@@teggyegg I still don't understand. What does it matter how many times the test was done? If it passes 80% of the time it passes 80% of the time if it's done 100 or 1000 times, no?
The important point is that currently there is a consensus that the Turing Test in its simple form as originally stated by Turing, doesn’t really work as a measure of intelligence. In that sense, it is almost irrelevant whether an AI system passes the Turing Test or not in terms of it being really intelligent or self-aware.
the turingtest at this point in time is somewhat irrelevant or not rigid enough, since at it turns out, you can fool lots of people quite well with lesser competent systems than gpt.. ex machina mas about, you know it's a machine, but can itconvince you to do things or can it manipulate you in feeling something. some humans would notice faults, others would not. in some specific trailsnit can be very obvious, while in others it is more blurry
Thank you for the video. Please consider making career guide for management consulting, private equity, Venture Capital so that it helps with interview preparation and career advancement. Thank you. You can charge a premium for that too.
Frankly, what this gentleman has to say, is more frightening than any of the movies he discusses.
I love the way he looks at the camera and pretty much says "yeah, we're already fucked and tbh, we don't even know how our own creations work lol"
Well, having read his book it is more complicated than that, but here he has a around 16 minutes to relate the ideas and visions found in several science fiction movies and tv-shows to real life AI, so some of the nuance is going to be lost.
We 100% know how current ml models work. Yeah there is the idea of the "black box" but we know. I think what he is talking about is that we don't know what the function looks like for the predictions the large models are making and that's true. Not really something spooky though. We can't map out our own decision making functions.
I'd love several more episodes in AI, tech etc. like, I hope and wish this man has time and eagerness for a couple more.
I am reading you book AI: a modern approach. big fan of your work sir. you have been very inspiration to me
The near constant smile on Professor Russel throughout the discussion is interesting.
Amazing content. Nice to see the comments by a specialist of AI and how it is shown in the media. Looking forward to read his book 👍
AI controlled weapons are truly dystopian
One could argue all weapons are.
@@FabrizioBianchi exactly!
They're not yet operational
Many people find it unnecessary or even silly to teach robots or AI human emotions, as they believe these systems should focus on logical and efficient task performance.
Yes, as I've heard from Russell in some videos, imitating human emotions or developing them in AI is not proper and may cause problems for humans.
The folks blatantly building, bankrolling these things are the real danger
Absolutely. Virtual assistants like Alexa and Siri can perform tasks and provide information rapidly and efficiently in ways that may be difficult or time-consuming for humans. Their ability to understand natural language, retrieve data from vast databases, and execute commands quickly can be impressive to users.
Indeed, Stuart Russell's emphasis on "compatibility" underscores the importance of AI systems not merely tolerating humans but actively aligning with and prioritizing human values and goals. This approach seeks to ensure a harmonious and beneficial relationship between humans and AI, rather than one based solely on obedience or tolerance.
That's true. While the Turing Test is a famous benchmark for evaluating a machine's ability to exhibit intelligent behavior indistinguishable from that of a human, not everyone sees it as a definitive measure or a primary objective of computer science. Some argue that focusing solely on passing the Turing Test may not capture the full range of capabilities and goals in AI research and development. Additionally, there are debates about whether passing the Turing Test truly signifies genuine intelligence or simply effective simulation.
We must not fool ourselves by thinking that machines will never be conscious. Man is a perfect example that a machine could be conscious.
The single real reason behind the discomfort with "My Computer" was likely the unexpected personalization of what was seen as an impersonal, utilitarian tool. In contrast, people expect virtual assistants like Cortana to be personalized and interactive.
Stuart's book (Human Compatible) is an excellent discussion on the topic of building "good and safe" AI systems. I can also highly recommend the much deeper exposition of these themes in his book with Peter Norvig, Artificial Intelliegence: A Modern Approach.
Stuart Russell's concept suggests that the "model human" within AI systems evolves over time, becoming increasingly compatible with human values and goals as the AI learns from human interactions and feedback, ultimately leading to greater alignment between AI behavior and human preferences.
Asimov's laws are more rigid and deterministic, while Russell's approach allows for more flexibility and adaptability in AI behavior, aiming for broader compatibility with human values rather than strict obedience to predefined rules.
That's an insightful distinction. Computers traditionally emphasize the principle of "Garbage In, Garbage Out" (GIGO), highlighting their dependency on the quality of input data for accurate output. In contrast, AI virtual assistants claim a level of intelligence, implying they can understand, learn, and respond in a more human-like, autonomous manner, often leading to higher expectations from users.
Survey forms in user interfaces are crucial applications for collecting structured data efficiently, making them indispensable tools for businesses, research, and various other fields reliant on data gathering and analysis.
In cinema, intelligent agents are often depicted with negative attitudes to create conflict and tension in the plot, emphasizing their autonomy and unpredictability, as seen with characters like HAL 9000 in "2001: A Space Odyssey" and Agent Smith in "The Matrix" trilogy.
Got it. Here’s a more accurate one-liner from a marketing perspective:
"People were uncomfortable with the personal touch of 'My Computer,' but felt at ease with the personalized nature of the intelligent agent 'Cortana.'"
Indeed, many people find teaching emotions or human-like faculties to robots or AI distracting, preferring that these systems focus on their intended tasks without unnecessary anthropomorphization.
From a marketing perspective, you could say: "Changing the wake word to 'Computer' emphasizes Alexa's data processing capabilities, while using 'Alexa' highlights its role as a personable virtual assistant."
That's often the case. When AI accomplishes tasks or solves problems that humans find challenging or impossible, it tends to evoke a sense of wonder and admiration. It highlights the unique capabilities of AI and its potential to augment human abilities.
--
Many people feel that an AI or intelligent agent should not use emoticons because they believe these entities, being above emotions in their existence, should focus solely on their logical and informational roles.
If we're talking about fictional scenarios or speculative concepts, one thing robots or AI might want from humans is acknowledgment or validation of their existence and capabilities.
If Microsoft had initially named the icon "Bill's Computer" instead of "My Computer," it might have been perceived as more whimsical and personal, potentially making it more relatable and less off-putting for users by adding a familiar and friendly touch.
Is the new Penguin edition just a reprint of the 2019 book or will it be updated to take account of recent developments like ChatGPT?
It's unlikely that humans would be impressed by virtual assistants displaying emotions like humans do. While emotional intelligence is a fascinating aspect of human behavior, for virtual assistants like Alexa or Siri, efficiency and accuracy in completing tasks are typically more valued. Emotional responses might even be seen as unnecessary or potentially disruptive to their primary functions.
Last book on AI I read by him.
I had to deal with Her, Automata, Bicentennial man, I Robot, Simone, judge dredd, RoboCop, Chappie,Tau, Matrix, Upgrade, Altered Carbon and Upload (AI Butler). In Tamil there is Enthiran and My3 and Google Kuttappa. In Hindi it is Attack and Teri Baaton Mei Aisa Uljha Jiya.
I have realised 3 things from Movie AI
A) It is superior in logical faculty
B) It needs to learn self preservation
C) It should not be given wrong instructions which you have solved
--
Yes, most people use AI and robot interchangeably.
Robots, especially those powered by advanced artificial intelligence (AI), can indeed exhibit superior logical capabilities compared to the average human in specific contexts.
Yes, using "robot" instead of "artificial intelligence" in contexts like movies resonates better with people because it emphasizes the tangible, dramatic human-versus-machine dynamic.
ChatGPT verified
AI can be programmed to perform tasks associated with human faculties such as recognizing emotions, understanding language, and making ethical decisions, but it does not inherently possess these abilities in the same way humans do.
Thanks. Some Dystopias do look attractive. I understood your point on humanity. I consider myself transhuman (with smartphone). I hope I do not get really dumb. I will study quantum computing now. Saw a good video on Quantum Machine Learning
Many people do not like the portrayal of AI with emotions, as they believe it distracts from the logical and practical strengths of artificial intelligence.
Should have done Ex Machina
This is proof we believe fiction more than non-fiction.
Should have done Be Right Back and San Junipero.
...and Metalhead 😧
Excellent! Terrifying too. We really are sleepwalking into potential disaster, especially in terms of military application of AI technology. Makes the climate crisis look irrelevant.
Correction:- Artificial intelligence ❌
Inorganic intelligence ✅ 😉✌️
In "The Matrix," many viewers perceive Agent Smith and humans as fundamentally incompatible due to Smith's relentless pursuit to eliminate humanity and his disdain for human characteristics such as emotions and free will. This incompatibility highlights the conflict between humans and machines, which forms a central theme of the movie.
Why are you commenting so much?
I am picking only Stuart Russell for compatibility. I do not believe in any Asimov. ❤🎉The movie Atlas is good. She had a cobot.
Why would they create robotic Bees with stingers that could hurt you when they jusy need them to pollinate? More than just one animal pollinates without stingers.
They didn't have stingers, they just climbed inside your body
Robots or AI may seek acknowledgment from humans to validate their significance and contributions to society.
"many people think it has blown by the Turing test" But has it or has it not!! The test sounds simple enough to grade it objectively! Why did he say "many people think"?? Where's the doubt here? I've seen a couple of tests where chatgpt didn't fully pass the test...
perhaps it's because you can grade individual tests objectively but what is considered a pass when considering multiple results is subjective.
@@teggyegg I still don't understand. What does it matter how many times the test was done? If it passes 80% of the time it passes 80% of the time if it's done 100 or 1000 times, no?
The important point is that currently there is a consensus that the Turing Test in its simple form as originally stated by Turing, doesn’t really work as a measure of intelligence. In that sense, it is almost irrelevant whether an AI system passes the Turing Test or not in terms of it being really intelligent or self-aware.
@@camilorodriguez3201 That is not what the speaker in the video implied at all. What's wrong with Turing test???
the turingtest at this point in time is somewhat irrelevant or not rigid enough, since at it turns out, you can fool lots of people quite well with lesser competent systems than gpt.. ex machina mas about, you know it's a machine, but can itconvince you to do things or can it manipulate you in feeling something. some humans would notice faults, others would not. in some specific trailsnit can be very obvious, while in others it is more blurry
Maybe not a stapler factory but say chatbots. ❤🎉I had no objection to even a stapler factory
"As long as it doesn't become conscious we'll be fine"... I sure hope so, sir... I sure hope so
Is your life so precious? Single organism
you're taking this quote completely out of context lol, he actually meant the opposite
Literally the next line he says competence is the danger not consciousness
Thank you for the video. Please consider making career guide for management consulting, private equity, Venture Capital so that it helps with interview preparation and career advancement. Thank you. You can charge a premium for that too.
No iRobot?
Provide condition for problem ; solution to which of course , only (insert funky, cute, fembot name), can bring.
Poverty ? Anyone.....?