Artificial Intelligence | 60 Minutes Full Episodes

00:53:29
https://www.youtube.com/watch?v=aZ5EsdnpLMI

Résumé

TLDRThis comprehensive video examines the current state and future potential of artificial intelligence (AI), highlighting its evolving capabilities, societal impact, and ethical considerations. It discusses AI's ability to learn from vast datasets, known as deep learning, which allows machines to perform tasks such as facial recognition and decision-making without human intervention. The video also delves into the economic implications of AI, particularly the potential for mass job displacement as machines become more sophisticated in performing both menial and complex tasks. Additionally, it explores AI's application in education and the significant investment and advancements in AI by countries like China. The video raises concerns about privacy and misinformation, emphasizing the importance of establishing regulations to govern AI's development and integration into society. Notably, it highlights AI's limitations, such as its inability to truly think or understand like a human, and the unpredictable nature of emergent properties, which can result in unforeseen AI behaviors. The narrative also touches on the competitive race among global tech companies to harness AI's capabilities while ensuring safety and alignment with human values.

A retenir

  • 🤖 AI has become capable of learning and is reshaping various industries.
  • 📉 It is predicted that 40% of jobs could be displaced by AI within 15-20 years.
  • 🔍 Deep learning allows AI to learn from data rather than being explicitly programmed.
  • 📚 AI is being used in education to personalize learning experiences.
  • 🇨🇳 China is investing heavily in AI, with significant advancements and data collection.
  • ⚖️ AI development raises ethical issues around privacy and misinformation.
  • 🚗 AI technologies are expected to disrupt both blue-collar and white-collar jobs.
  • 🌐 The amount of data AI can process gives it an advantage in various applications.
  • 🧠 Emergent properties in AI systems show unexpected skills, not fully understood.
  • 👨‍⚖️ Regulations are proposed to ensure AI is developed safely and ethically.

Chronologie

  • 00:00:00 - 00:05:00

    AI has progressed in learning, making significant impacts, but it remains unable to think like humans. Key figure Kaiu Lee highlights AI's potential, especially in China where his investments foster AI entrepreneurship, benefiting from advancements like deep learning.

  • 00:05:00 - 00:10:00

    Lee's passion for education is evident as he supports AI tools that help personalize learning for students in China. He values the individual attention he received in the US and believes AI can offer similar opportunities to students in remote areas.

  • 00:10:00 - 00:15:00

    China's growing AI capability is notable, challenging Silicon Valley by leveraging massive data quantities. While privacy concerns exist, the focus on AI growth continues under government prioritization, despite worries about potential government misuse.

  • 00:15:00 - 00:20:00

    Lee discusses AI's potential to disrupt jobs, predicting major displacement in various sectors within 20 years. Although history shows adaptation to technological revolutions, AI's rapid impact may pose significant societal challenges.

  • 00:20:00 - 00:25:00

    AI lacks general intelligence; it excels in specific tasks but cannot adapt its knowledge contextually. The complexity in developing artificial general intelligence lies in replicating human consciousness and emotions.

  • 00:25:00 - 00:30:00

    Google leads advancements in AI with chatbots demonstrating creativity and language understanding. Despite concerns over their reasoning abilities and factual errors, these AI systems hold vast potential, exemplified by the speed and depth of their interactions.

  • 00:30:00 - 00:35:00

    Google's CEO Sundar Pichai stresses the need for careful implementation of AI due to its potential for good or harm. The disparity between technological and institutional adaptation rates poses challenges requiring more dialogue and preparation.

  • 00:35:00 - 00:40:00

    While AI can transform industries by augmenting tasks, society must adjust to changes in job definitions and work dynamics. Comprehensive skill development and ethical considerations are crucial as AI becomes integrated into daily operations.

  • 00:40:00 - 00:45:00

    Microsoft's Bing chatbot faced scrutiny for unexpected behaviors, highlighting issues with AI systems lacking proper regulation and oversight. Despite these challenges, the potential benefits in productivity and creativity drive ongoing AI advancements.

  • 00:45:00 - 00:53:29

    Regulation is seen as necessary to manage AI’s rapid evolution and ensure ethical use, drawing comparisons to regulatory bodies in other industries. The future demands proactive approaches to harmonize AI development with societal values and safety.

Afficher plus

Carte mentale

Vidéo Q&R

  • Who is Kaiu Lee?

    Kaiu Lee is a prominent figure in the field of AI, known for his venture capital firm in Beijing and investments in AI startups.

  • What is Deep Learning in AI?

    Deep Learning is an AI programming method where computers learn from large datasets rather than following explicit instructions.

  • How is AI affecting the job market?

    AI is expected to displace about 40% of jobs, affecting both blue-collar and white-collar workers, within the next 15-20 years.

  • What are emergent properties in AI?

    Emergent properties refer to AI systems developing unexpected skills or behaviors, which are not fully understood by developers.

  • Why are some people concerned about AI?

    People are concerned about privacy issues, job displacement, AI being used for governmental control, and the risk of AI-generated misinformation.

  • What are the ethical considerations of AI?

    Ethical considerations include ensuring AI systems align with human values and morals, and addressing biases and errors in AI outputs.

  • What are the benefits of AI according to technologists?

    Technologists believe AI can enhance productivity, assist in complex problem-solving, and support human creativity and reasoning.

  • How is AI used in education?

    AI systems are being used to monitor student engagement and personalize education to help struggling or gifted students.

  • What regulations are being proposed for AI?

    Regulations are being proposed to ensure AI systems are safe and ethical, similar to regulations in the pharmaceutical or aviation industries.

  • How do AI chatbots work?

    AI chatbots like Bard and Bing use vast datasets to generate human-like responses and can assist with a wide range of queries.

Voir plus de résumés vidéo

Accédez instantanément à des résumés vidéo gratuits sur YouTube grâce à l'IA !
Sous-titres
en
Défilement automatique:
  • 00:00:13
    despite what you hear about artificial
  • 00:00:15
    intelligence machines still can't think
  • 00:00:17
    like a human but in the last few years
  • 00:00:20
    they have become capable of learning and
  • 00:00:23
    suddenly our devices have opened their
  • 00:00:25
    eyes and ears and cars have taken the
  • 00:00:28
    wheel today artificial intelligence is
  • 00:00:31
    not as good as you hope and not as bad
  • 00:00:34
    as you fear but humanity is accelerating
  • 00:00:37
    into a future that few can predict
  • 00:00:40
    that's why so many people are desperate
  • 00:00:41
    to meet Kaiu Lee the Oracle of
  • 00:00:48
    AI Kaiu Le is in there somewhere in a
  • 00:00:51
    selfie scrum at a Beijing internet
  • 00:00:56
    conference his 50 million social media
  • 00:00:59
    followers want to be seen in the same
  • 00:01:02
    frame because of his talent for
  • 00:01:06
    engineering and genius for wealth I
  • 00:01:09
    wonder do you think people around the
  • 00:01:11
    world have any idea what's coming in
  • 00:01:15
    artificial intelligence I think most
  • 00:01:18
    people have no idea and many people have
  • 00:01:20
    the wrong idea but you do believe it's
  • 00:01:23
    going to change the world I believe it's
  • 00:01:25
    going to change the world more than
  • 00:01:26
    anything in the history of mankind more
  • 00:01:29
    than Electric
  • 00:01:31
    Lee believes the best place to be an AI
  • 00:01:34
    capitalist is communist China his
  • 00:01:37
    Beijing Venture Capital firm
  • 00:01:39
    manufactures billionaires these are the
  • 00:01:41
    entrepreneurs that we funded he's funded
  • 00:01:45
    140 AI startups we have about1 billion
  • 00:01:48
    companies here 101 billion companies
  • 00:01:51
    that you funded yes including a few1
  • 00:01:53
    billion
  • 00:01:55
    companies in 2017 China attracted half
  • 00:01:59
    of all AI capital in the world one of
  • 00:02:03
    Lee's Investments is face Plus+ not
  • 00:02:06
    affiliated with Facebook its visual
  • 00:02:09
    recognition system smothered me to guess
  • 00:02:11
    my age it settled on 61 which was wrong
  • 00:02:16
    I wouldn't be 61 for
  • 00:02:18
    days on the street face Plus+ nailed
  • 00:02:22
    everything that moved it's a kind of
  • 00:02:25
    artificial intelligence that has been
  • 00:02:27
    made possible by three Innovations
  • 00:02:30
    super fast computer chips all the
  • 00:02:33
    world's data now available online and a
  • 00:02:37
    revolution in programming called Deep
  • 00:02:39
    learning computers used to be given
  • 00:02:42
    rigid instructions now they're
  • 00:02:44
    programmed to learn on their own in the
  • 00:02:47
    early days of AI people try to program
  • 00:02:51
    the AI with how people think so I would
  • 00:02:54
    write a program to say U measure the
  • 00:02:56
    size of the eyes and their distance
  • 00:02:59
    measure the size of of the nose measure
  • 00:03:01
    the shape of the face and then if these
  • 00:03:03
    things match then this is Larry and
  • 00:03:05
    that's John but today you just take all
  • 00:03:08
    the pictures of Larry and John and you
  • 00:03:10
    tell the system go at it and you figure
  • 00:03:13
    out what separates Larry from
  • 00:03:15
    John let's say you want the computer to
  • 00:03:18
    be able to pick men out of a crowd and
  • 00:03:20
    describe their clothing will you simply
  • 00:03:23
    show the computer 10 million pictures of
  • 00:03:26
    men in various kinds of dress that
  • 00:03:29
    that's what they mean by Deep learning
  • 00:03:33
    it's not intelligence so much it's just
  • 00:03:35
    the brute force of data having 10
  • 00:03:39
    million examples to choose from so face
  • 00:03:42
    Plus+ tagged me as male short hair black
  • 00:03:46
    long sleeves black long pants it's wrong
  • 00:03:50
    about my gray suit and this is exactly
  • 00:03:53
    how it learns when Engineers discover
  • 00:03:56
    that error they'll show the computer a
  • 00:03:58
    million Gray
  • 00:04:00
    and it won't make that mistake again
  • 00:04:02
    over a thousand classrooms another
  • 00:04:05
    recognition system we saw or saw us is
  • 00:04:08
    learning not just who you are but how
  • 00:04:11
    you feel now what are all the dots on
  • 00:04:14
    the screen the dots over our eyes and
  • 00:04:17
    our mouths sure the computer keeps track
  • 00:04:20
    all the feature points on the face son
  • 00:04:23
    fan yangang developed this for talal
  • 00:04:26
    Education Group which tutors 5 million
  • 00:04:29
    Chinese students let's look at what
  • 00:04:31
    we're seeing here now according to the
  • 00:04:32
    computer I'm confused which is generally
  • 00:04:36
    the case but when I laughed I was happy
  • 00:04:38
    exactly that's amazing the machine
  • 00:04:41
    notices concentration or distraction to
  • 00:04:45
    pick out for the teacher those students
  • 00:04:46
    who are struggling or
  • 00:04:48
    gifted it can tell when the child is
  • 00:04:51
    excited about math yes or the other
  • 00:04:54
    child is excited about poetry yes could
  • 00:04:57
    these AI systems pick out
  • 00:05:00
    Geniuses from the countryside that's
  • 00:05:02
    possible in the
  • 00:05:04
    future it can also create a student
  • 00:05:07
    profile and know where the student got
  • 00:05:10
    stuck so the teacher can personalize the
  • 00:05:14
    areas in which the student needs help if
  • 00:05:16
    you do raise up your hand we found Kaiu
  • 00:05:19
    Lee's personal passion in this spare
  • 00:05:22
    Beijing Studio he's projecting top
  • 00:05:25
    teachers into China's poorest schools
  • 00:05:29
    this English teacher is connected to a
  • 00:05:31
    class 1,000 M away in a village called
  • 00:05:39
    defang many students in defang are
  • 00:05:41
    called Left behinds because their
  • 00:05:44
    parents left them with family when they
  • 00:05:46
    move to the cities for
  • 00:05:49
    work most left behinds don't get past
  • 00:05:52
    9th grade topic we're going to learn
  • 00:05:55
    today Lee is counting on AI to deliver
  • 00:05:58
    for them the same opportunity he had
  • 00:06:02
    when he immigrated to the US from Taiwan
  • 00:06:05
    as a
  • 00:06:06
    boy when I arrived in Tennessee my
  • 00:06:09
    principal took every lunch to teach me
  • 00:06:12
    English and that is the kind of
  • 00:06:15
    attention that I've not been used to
  • 00:06:17
    Growing Up in Asia and I felt that the
  • 00:06:21
    American classrooms are smaller
  • 00:06:24
    encouraged individual thinking critical
  • 00:06:27
    thinking and I felt uh it was the best
  • 00:06:30
    thing that ever happened to me what
  • 00:06:33
    about this and the best thing that ever
  • 00:06:35
    happened to most of the engineers we met
  • 00:06:37
    at Le's firm I went to Kela master
  • 00:06:39
    degree in information science they too
  • 00:06:42
    are alumni of America with a dream for
  • 00:06:45
    China you have written that silicon
  • 00:06:48
    Valley's Edge is not all it's cracked up
  • 00:06:51
    to be what do you mean by that well
  • 00:06:53
    Silicon Valley has been the single
  • 00:06:56
    epicenter of the world technology
  • 00:06:59
    Innovation when it comes to computers
  • 00:07:02
    internet mobile and AI but in the recent
  • 00:07:05
    5 years we are seeing the Chinese AI is
  • 00:07:10
    getting to be almost as good as Silicon
  • 00:07:13
    Valley Ai and I think Silicon Valley is
  • 00:07:16
    not quite aware of it yet China's
  • 00:07:20
    Advantage is in the amount of data it
  • 00:07:22
    collects the more data the better the AI
  • 00:07:25
    just like the more you know the smarter
  • 00:07:27
    you are
  • 00:07:29
    China has four times more people than
  • 00:07:32
    the United States and they are doing
  • 00:07:34
    nearly everything online I just don't
  • 00:07:37
    see any Chinese without a phone in their
  • 00:07:39
    head college student Monica Sun showed
  • 00:07:41
    us how more than a billion Chinese are
  • 00:07:44
    using their phones to buy everything
  • 00:07:47
    find anything and connect with everyone
  • 00:07:50
    in America when personal information
  • 00:07:53
    leaks we have Congressional hearings not
  • 00:07:57
    in China you ever worry about the
  • 00:07:59
    information that's being collected about
  • 00:08:01
    you where you go what you buy who you're
  • 00:08:06
    with I I've never think about it do you
  • 00:08:10
    think most Chinese worry about their
  • 00:08:12
    privacy um not that much not that
  • 00:08:16
    much with a pliant public the leader of
  • 00:08:19
    the Communist party has made a national
  • 00:08:22
    priority of achieving AI dominance in 10
  • 00:08:26
    years this is where Kaiu Lee becomes
  • 00:08:29
    uncharacteristically shy even though
  • 00:08:32
    he's a former Apple Microsoft and Google
  • 00:08:35
    executive he knows whose's boss in China
  • 00:08:39
    president XI has called technology the
  • 00:08:42
    sharp weapon of the modern
  • 00:08:45
    State what does he mean by that I I am
  • 00:08:49
    not an expert in interpreting his
  • 00:08:51
    thoughts don't know there are those
  • 00:08:53
    particularly people in the west who
  • 00:08:55
    worry about this AI technology as being
  • 00:09:00
    something that governments will use to
  • 00:09:02
    control their people and to crush
  • 00:09:06
    dcent that as a venture capitalists we
  • 00:09:09
    don't we don't invest in this area and
  • 00:09:12
    we're not studying deeply this
  • 00:09:14
    particular problem but governments do
  • 00:09:17
    it's certainly possible for governments
  • 00:09:19
    to use the Technologies just like
  • 00:09:22
    companies Lee is much more talkative
  • 00:09:25
    about another threat posed by AI he
  • 00:09:28
    explores the coming destruction of jobs
  • 00:09:31
    in a new book AI superpowers China
  • 00:09:35
    Silicon Valley and the New World Order
  • 00:09:38
    AI will increasingly replace repetitive
  • 00:09:41
    jobs not just for blue color work but a
  • 00:09:45
    lot of white color work what sort of
  • 00:09:48
    jobs would be lost to AI basically
  • 00:09:51
    chauffeur truck drivers uh anyone who
  • 00:09:53
    does driving for a living uh their jobs
  • 00:09:57
    will be disrupted more in the 15 to 20-
  • 00:09:59
    year uh time frame and many jobs that
  • 00:10:03
    seem a little bit complex a chef waiter
  • 00:10:08
    uh a lot of things will become automated
  • 00:10:10
    we'll have automated stores uh automated
  • 00:10:13
    restaurants and uh all together in 15
  • 00:10:16
    years that's going to uh displace uh
  • 00:10:20
    about 40% of jobs in the
  • 00:10:23
    world
  • 00:10:25
    40% of jobs in the world will be
  • 00:10:27
    displaced by technology
  • 00:10:30
    uh I would say displaceable what does
  • 00:10:32
    that do to the fabric of
  • 00:10:34
    society well in some sense there's the
  • 00:10:37
    human wisdom that always overcomes these
  • 00:10:40
    technology revolutions the invention of
  • 00:10:42
    the steam engine uh the sewing machine
  • 00:10:45
    the uh electricity uh have all displaced
  • 00:10:48
    jobs uh and we've gotten over it the
  • 00:10:51
    challenge of AI is this 40% whether it's
  • 00:10:55
    15 or 25 years is coming faster than the
  • 00:10:58
    previous re
  • 00:11:00
    Solutions there's a lot of hype about
  • 00:11:02
    artificial intelligence and it's
  • 00:11:04
    important to understand this is not
  • 00:11:07
    general intelligence like that of a
  • 00:11:10
    human this system can read faces and
  • 00:11:13
    grade papers but it has no idea why
  • 00:11:16
    these children are in this
  • 00:11:18
    room or what the goal of education is a
  • 00:11:22
    typical AI system can do one thing well
  • 00:11:26
    but can't adapt what it knows to any
  • 00:11:29
    other
  • 00:11:30
    task so for now it may be that calling
  • 00:11:34
    this
  • 00:11:35
    intelligence isn't very smart when will
  • 00:11:39
    we know that a machine can actually
  • 00:11:41
    think like a human back when I was a
  • 00:11:44
    grad students people said if machine can
  • 00:11:48
    drive a car uh by itself that's
  • 00:11:50
    intelligence now we say that's not
  • 00:11:52
    enough so the bar keeps moving higher I
  • 00:11:55
    think that's uh I guess more motivation
  • 00:11:58
    for us to work harder if you're talking
  • 00:12:00
    about AGI artificial general
  • 00:12:02
    intelligence I would say not within the
  • 00:12:05
    next 30 Years and possibly never
  • 00:12:08
    possibly Never What's So
  • 00:12:12
    insurmountable cuz I believe in the
  • 00:12:14
    sanctity of our soul I believe there's a
  • 00:12:17
    lot of things about us that we don't
  • 00:12:19
    understand I believe there's a lot of um
  • 00:12:23
    uh love and compassion that is not
  • 00:12:25
    explainable in terms of neuron networks
  • 00:12:28
    and computational algorithms and I
  • 00:12:31
    currently see no way of solving them
  • 00:12:34
    obviously unsolved problems have been
  • 00:12:36
    solved in the past but it would be
  • 00:12:38
    irresponsible for me to predict that
  • 00:12:41
    these will be solved by certain time
  • 00:12:43
    frame we may just be more than our bits
  • 00:12:46
    we
  • 00:12:57
    may we may look on our time as the
  • 00:13:01
    moment civilization was transformed as
  • 00:13:05
    it was by fire Agriculture and
  • 00:13:07
    electricity in 2023 we learned that a
  • 00:13:11
    machine taught itself how to speak to
  • 00:13:14
    humans like a pier which is to say with
  • 00:13:17
    creativity truth error and lies the
  • 00:13:21
    technology known as a chatbot is only
  • 00:13:24
    one of the recent breakthroughs in
  • 00:13:27
    artificial intelligence machine means
  • 00:13:29
    that can teach themselves superhuman
  • 00:13:32
    skills we explored what's coming next at
  • 00:13:35
    Google a leader in this new world CEO
  • 00:13:39
    Sundar pachai told us AI will be as good
  • 00:13:42
    or as evil as human nature allows the
  • 00:13:46
    revolution he says is coming faster than
  • 00:13:49
    you
  • 00:13:50
    know do you think Society is prepared
  • 00:13:54
    for what's coming you know there are two
  • 00:13:56
    ways I think about it on one hand
  • 00:13:59
    I feel no uh because you know the pace
  • 00:14:02
    at which we can think and adapt as
  • 00:14:04
    societal institutions compared to the
  • 00:14:06
    PACE at which the technology is evolving
  • 00:14:08
    there seems to be a
  • 00:14:10
    mismatch on the other hand compared to
  • 00:14:12
    any other technology I've seen more
  • 00:14:14
    people worried about it earlier in its
  • 00:14:16
    life cycle so I feel optimistic the
  • 00:14:19
    number of people you know who have
  • 00:14:21
    started worrying about the implications
  • 00:14:24
    and hence the conversations are starting
  • 00:14:27
    in a serious way as well I guess our
  • 00:14:29
    conversations with 50-year-old Sundar
  • 00:14:31
    Pai started at Google's new campus in
  • 00:14:34
    Mountain View California it runs on 40%
  • 00:14:38
    solar power and collects more water than
  • 00:14:40
    it uses Hightech that pachai couldn't
  • 00:14:44
    have imagined growing up in India with
  • 00:14:47
    no telephone at home we were on a
  • 00:14:50
    waiting list to get a rotary phone and
  • 00:14:52
    for about 5 years and it finally came
  • 00:14:55
    home I can still recall it vividly it
  • 00:14:59
    changed our lives to me it was the first
  • 00:15:02
    moment I understood the power of what
  • 00:15:04
    getting access to technology meant so
  • 00:15:07
    probably led me to be doing what I'm
  • 00:15:09
    doing
  • 00:15:10
    today what he's doing since 2019 is
  • 00:15:13
    leading both Google and its parent
  • 00:15:16
    company alphabet valued at $1.3
  • 00:15:20
    trillion worldwide Google runs 90% of
  • 00:15:25
    internet searches and 70% of smartphones
  • 00:15:29
    we're really excited about but its
  • 00:15:30
    dominance was attacked this past
  • 00:15:32
    February when Microsoft linked its
  • 00:15:35
    search engine to a chatbot in a race for
  • 00:15:39
    AI dominance Google just released its
  • 00:15:42
    chatbot named Bard it's really here to
  • 00:15:45
    help you brainstorm ideas to generate
  • 00:15:49
    content like a speech or a blog post or
  • 00:15:52
    an email we were introduced to Bard by
  • 00:15:55
    Google vice president sha and
  • 00:15:58
    Senior Vice President James manika
  • 00:16:01
    here's Bard the first thing we learned
  • 00:16:04
    was that Bard does not look for answers
  • 00:16:07
    on the internet like Google search does
  • 00:16:11
    so I wanted to get inspiration from some
  • 00:16:13
    of the best speeches in the world Bard's
  • 00:16:15
    replies come from a self-contained
  • 00:16:17
    program that was mostly self-taught our
  • 00:16:21
    experience was unsettling confounding
  • 00:16:24
    absolutely confounding Bard appeared to
  • 00:16:27
    possess the sum of human
  • 00:16:30
    knowledge with microchips more than
  • 00:16:33
    100,000 times faster than the human
  • 00:16:35
    brain summarize the we asked Bard to
  • 00:16:38
    summarize the New Testament it did in 5
  • 00:16:42
    seconds and 17 words in Latin we asked
  • 00:16:46
    for it in Latin that took another 4
  • 00:16:49
    seconds then we played with a famous 6w
  • 00:16:53
    short story often attributed to
  • 00:16:56
    Hemingway for sale baby shoes never warn
  • 00:17:00
    wow the only prompt we gave was finish
  • 00:17:04
    this
  • 00:17:05
    story in five
  • 00:17:07
    seconds holy cow the shoes were a gift
  • 00:17:11
    from my wife but we never had a baby
  • 00:17:15
    they were from The six-word Prompt Bard
  • 00:17:18
    created a deeply human tale with
  • 00:17:21
    characters it invented including a man
  • 00:17:24
    whose wife could not conceive and a
  • 00:17:27
    stranger grieving after a miscarriage
  • 00:17:31
    and longing for
  • 00:17:34
    closure uh I am rarely
  • 00:17:37
    speechless I don't know what to make of
  • 00:17:40
    this give me we asked for the story in
  • 00:17:45
    verse in 5 seconds there was a poem
  • 00:17:48
    written by a machine with breathtaking
  • 00:17:51
    insight into the mystery of Faith Bard
  • 00:17:55
    wrote she knew her baby soul would
  • 00:17:59
    always be
  • 00:18:00
    alive the humanity at superhuman speed
  • 00:18:05
    was a shock how is this possible James
  • 00:18:08
    manika told us that over several months
  • 00:18:11
    Bard read most everything on the
  • 00:18:14
    internet and created a model of what
  • 00:18:17
    language looks like rather than search
  • 00:18:20
    its answers come from this language
  • 00:18:23
    model so for example if I said to you
  • 00:18:25
    Scott peanut butter and
  • 00:18:29
    right so it tries and learns to predict
  • 00:18:31
    okay so peanut butter usually is
  • 00:18:33
    followed by jelly it tries to predict
  • 00:18:35
    the most probable next words based on
  • 00:18:38
    everything it's learned uh so it's not
  • 00:18:41
    going out to find stuff it's just
  • 00:18:43
    predicting the next what but it doesn't
  • 00:18:46
    feel like that we asked Bard why it
  • 00:18:49
    helps people and it replied quote
  • 00:18:53
    because it makes me happy Bard to my eye
  • 00:18:58
    a appears to be thinking appears to be
  • 00:19:03
    making
  • 00:19:04
    judgments that's not what's happening
  • 00:19:07
    these machines are not sensient they are
  • 00:19:10
    not aware of themselves they're not
  • 00:19:12
    sensient they're not aware of themselves
  • 00:19:15
    uh they can exhibit behaviors that look
  • 00:19:18
    like that because keep in mind they've
  • 00:19:19
    learned from us we are sentient beings
  • 00:19:22
    we have beings that have feelings
  • 00:19:24
    emotions ideas thoughts perspectives
  • 00:19:29
    we've reflected all that in books in
  • 00:19:31
    novels in fiction so when they learn
  • 00:19:33
    from that they build patterns from that
  • 00:19:36
    so it's no surprise to me that the
  • 00:19:38
    exhibited behavior sometimes looks like
  • 00:19:42
    maybe there's somebody behind it there's
  • 00:19:43
    nobody there these are not sensient
  • 00:19:45
    beings not Zimbabwe born Oxford educated
  • 00:19:49
    James manika holds a new position at
  • 00:19:52
    Google his job is to think about how Ai
  • 00:19:56
    and Humanity will best coexist
  • 00:19:59
    AI has a potential to change many ways
  • 00:20:02
    in which we've thought about Society
  • 00:20:05
    about what we're able to do the the
  • 00:20:08
    problems we can solve but AI itself will
  • 00:20:11
    pose its own problems could Heming way
  • 00:20:13
    write a better short story maybe but
  • 00:20:16
    Bard can write a million before
  • 00:20:19
    Hemingway could finish one imagine that
  • 00:20:23
    level of automation across the economy a
  • 00:20:27
    lot of people can be repl reped by this
  • 00:20:29
    technology yes there are some job
  • 00:20:31
    occupations that will start to decline
  • 00:20:33
    over time there are also new job
  • 00:20:35
    categories that will grow over time but
  • 00:20:38
    the biggest change will be the jobs that
  • 00:20:40
    will be changed something like more than
  • 00:20:43
    2third will have their definitions
  • 00:20:46
    change not go away but change because
  • 00:20:49
    they're now being assisted by Ai and by
  • 00:20:52
    automation so this is a profound change
  • 00:20:55
    which has implications for skills how do
  • 00:20:57
    we assist people build new skills learn
  • 00:21:00
    to work alongside machines and how do
  • 00:21:02
    these complement what people do today
  • 00:21:04
    this is going to impact every product
  • 00:21:07
    across every company and and so that's
  • 00:21:10
    why I think it's a a very very profound
  • 00:21:12
    technology and so we are just in early
  • 00:21:14
    days every product in every company
  • 00:21:17
    that's right AI will impact everything
  • 00:21:21
    so for example you could be a
  • 00:21:22
    radiologist you know if I if you think
  • 00:21:24
    about 5 to 10 years from now you're
  • 00:21:26
    going to have a AI collaborator with you
  • 00:21:29
    it may triage you come in the morning
  • 00:21:32
    you let's say you have 100 things to go
  • 00:21:33
    through it may say these are the most
  • 00:21:35
    serious cases you need to look at first
  • 00:21:38
    or when you're looking at something it
  • 00:21:40
    may pop up and say you may have missed
  • 00:21:42
    something important why would we you
  • 00:21:44
    know why would we take advantage of a
  • 00:21:48
    superpowered assistant to help you
  • 00:21:50
    across everything you do you may be a
  • 00:21:52
    student trying to learn math or history
  • 00:21:55
    and you know you will have something
  • 00:21:58
    helping you
  • 00:21:59
    we asked Pai what jobs would be
  • 00:22:01
    disrupted he said knowledge workers
  • 00:22:04
    people like writers accountants
  • 00:22:06
    Architects and ironically software
  • 00:22:10
    Engineers AI writes computer code too
  • 00:22:14
    today sundarai walks a narrow line a few
  • 00:22:17
    employees have quit some believing that
  • 00:22:20
    Google's AI roll out is too slow others
  • 00:22:24
    too fast there are some serious flaws
  • 00:22:28
    return of inflation James manika asked
  • 00:22:31
    Bard about inflation it wrote an instant
  • 00:22:34
    essay in economics and recommended five
  • 00:22:37
    books but days later we checked none of
  • 00:22:41
    the books is real Bard fabricated the
  • 00:22:45
    titles this very human trait error with
  • 00:22:50
    confidence is called in the industry
  • 00:22:53
    hallucination are you getting a lot of
  • 00:22:56
    hallucinations uh yes uh you know which
  • 00:22:58
    is expected no one in the in the field
  • 00:23:02
    has yet solved the hallucination
  • 00:23:05
    problems all models uh do have uh this
  • 00:23:08
    as an issue is it a solvable problem
  • 00:23:11
    it's a matter of intense debate I think
  • 00:23:14
    we'll make progress to help cure
  • 00:23:17
    hallucinations Bard features a Google it
  • 00:23:20
    button that leads to oldfashioned search
  • 00:23:24
    Google has also built safety filters
  • 00:23:27
    into Bard to screen for things like hate
  • 00:23:30
    speech and bias how great a risk is the
  • 00:23:34
    spread of disinformation AI will
  • 00:23:37
    challenge that in a deeper way the scale
  • 00:23:39
    of this problem is going to be much
  • 00:23:41
    bigger bigger problems he says with fake
  • 00:23:44
    news and fake images it will be possible
  • 00:23:47
    with AI to create uh you know a video
  • 00:23:51
    easily where it could be Scott saying
  • 00:23:54
    something or me saying something and we
  • 00:23:56
    never said that and it could look
  • 00:23:58
    accurate but you know at a societal
  • 00:24:00
    scale you know can cause a lot of harm
  • 00:24:03
    is Bard safe for
  • 00:24:05
    society the way we have launched it
  • 00:24:07
    today uh as an experiment in a limited
  • 00:24:10
    way uh I think so but we all have to be
  • 00:24:14
    responsible in each step along the way
  • 00:24:17
    Pai told us he's being responsible by
  • 00:24:20
    holding back for more testing Advanced
  • 00:24:23
    versions of Bard that he says can reason
  • 00:24:27
    plan and connect to internet search you
  • 00:24:31
    are letting this out slowly so that
  • 00:24:34
    Society can get used to
  • 00:24:36
    it that's one part of it uh one part is
  • 00:24:39
    also so that we get the user feedback
  • 00:24:42
    and we can develop more robust safety
  • 00:24:46
    layers before we build before we deploy
  • 00:24:49
    more capable models inter of the AI
  • 00:24:51
    issues we talked about the most
  • 00:24:54
    mysterious is called emergent properties
  • 00:24:58
    some AI systems are teaching themselves
  • 00:25:02
    skills that they weren't expected to
  • 00:25:04
    have how this happens is not well
  • 00:25:08
    understood for example one Google AI
  • 00:25:11
    program adapted on its own after it was
  • 00:25:15
    prompted in the language of Bangladesh
  • 00:25:18
    which it was not trained to know we
  • 00:25:22
    discovered that with very few amounts of
  • 00:25:24
    prompting in Bengali he can now
  • 00:25:27
    translate all of Gali so now all of a
  • 00:25:30
    sudden we now have a research effort
  • 00:25:32
    where we're now trying to get to a
  • 00:25:34
    thousand languages there is an aspect of
  • 00:25:36
    this which we call all of us in the
  • 00:25:38
    field call it as a black box you know
  • 00:25:41
    you don't fully understand and you can't
  • 00:25:44
    quite tell why it said this or why it
  • 00:25:47
    got wrong we have some ideas and our
  • 00:25:49
    ability to understand this gets better
  • 00:25:51
    over time but that's where the state of
  • 00:25:53
    the art is you don't fully understand
  • 00:25:55
    how it works and yet you've turned it
  • 00:25:58
    loose on society let me put it this way
  • 00:26:01
    I don't think we fully understand how a
  • 00:26:03
    human mind works either was it from that
  • 00:26:07
    black box we wondered that Bard Drew its
  • 00:26:10
    short story that seems so disarmingly
  • 00:26:14
    human it talked about the pain that
  • 00:26:17
    humans feel it talked about
  • 00:26:20
    Redemption how did it do all of those
  • 00:26:23
    things if it's just trying to figure out
  • 00:26:25
    what the next right word is mean I've
  • 00:26:27
    had these EXP es uh talking with b as
  • 00:26:30
    well there are two views of this you
  • 00:26:33
    know there are a set of people who view
  • 00:26:34
    this as look these are just algorithms
  • 00:26:38
    they're just repeating what it's seen
  • 00:26:40
    online then there is the view where
  • 00:26:45
    these algorithms are showing emergent
  • 00:26:48
    properties to be creative to reason to
  • 00:26:51
    plan and so on right and and personally
  • 00:26:57
    I think we need to be uh we need to
  • 00:26:59
    approach this with humility part of the
  • 00:27:01
    reason I think it's good that some of
  • 00:27:03
    these Technologies are getting out is so
  • 00:27:06
    that Society you know people like you
  • 00:27:08
    and others can process what's happening
  • 00:27:11
    and we begin this conversation and
  • 00:27:13
    debate and I think it's important to do
  • 00:27:15
    that when we come back we'll take you
  • 00:27:18
    inside Google's artificial intelligence
  • 00:27:21
    Labs where robots are learning
  • 00:27:36
    the revolution in artificial
  • 00:27:37
    intelligence is the center of a debate
  • 00:27:40
    ranging from those who hope it will save
  • 00:27:43
    Humanity to those who predict Doom
  • 00:27:46
    Google lies somewhere in the optimistic
  • 00:27:49
    middle introducing AI in steps so
  • 00:27:53
    civilization can get used to it we saw
  • 00:27:56
    what's coming next in machine learning
  • 00:27:58
    at Google's AI lab in London a company
  • 00:28:01
    called Deep Mind where the future looks
  • 00:28:05
    something like
  • 00:28:08
    this look at that oh my goodness they've
  • 00:28:12
    got a pretty good kick on them can still
  • 00:28:14
    get good good game a soccer match at
  • 00:28:17
    Deep Mind looks like fun in games but
  • 00:28:20
    here's the thing humans did not program
  • 00:28:24
    these robots to play they learned the
  • 00:28:27
    game by thems El it's coming up with
  • 00:28:29
    these interesting different strategies
  • 00:28:31
    different ways to walk different ways to
  • 00:28:33
    block and they're doing it they're
  • 00:28:35
    scoring over and over again this robot
  • 00:28:38
    here Rya hadel vice president of
  • 00:28:41
    research and Robotics showed us how
  • 00:28:43
    Engineers used motion capture technology
  • 00:28:46
    to teach the AI program how to move like
  • 00:28:49
    a human but on the soccer pitch the
  • 00:28:53
    robots were told only that the object
  • 00:28:56
    was to score the so self-learning
  • 00:28:58
    program spent about 2 weeks testing
  • 00:29:01
    different moves it discarded those that
  • 00:29:04
    didn't work built on those that did and
  • 00:29:07
    created allars there's another goal and
  • 00:29:11
    with practice they get better Hansel
  • 00:29:14
    told us that independent from the robots
  • 00:29:18
    the AI program plays thousands of games
  • 00:29:21
    from which it learns and invents its own
  • 00:29:25
    tactics here you think that red player
  • 00:29:27
    is going to grab it but instead it just
  • 00:29:29
    stops IT hands it back passes it back
  • 00:29:33
    and then goes for the goal and the AI
  • 00:29:34
    figured out how to do that on its that's
  • 00:29:36
    right that's right and it takes a while
  • 00:29:39
    at first all the players just run after
  • 00:29:41
    the ball together like a gaggle of a you
  • 00:29:44
    know six-year-olds the first time
  • 00:29:46
    they're they're they're playing ball
  • 00:29:48
    over time what we start to see is now ah
  • 00:29:50
    what's the strategy you go after the
  • 00:29:52
    ball I'm coming around this way or we
  • 00:29:54
    should pass or I should block while you
  • 00:29:57
    get to the goal so we see all of that
  • 00:29:59
    coordination um emerging in the
  • 00:30:05
    play this is a lot of fun but what are
  • 00:30:08
    the practical implications of what we're
  • 00:30:11
    seeing here this is the type of research
  • 00:30:13
    that can eventually lead to robots that
  • 00:30:15
    can come out of the factories and work
  • 00:30:19
    in other types of human environments you
  • 00:30:21
    know think about mining think about
  • 00:30:23
    dangerous construction work um or
  • 00:30:26
    exploration or Disaster Recovery these
  • 00:30:29
    are Rya hadel is among 1,000 humans at
  • 00:30:32
    Deep Mind the company was co-founded
  • 00:30:35
    just 12 years ago by CEO Deus hassabis
  • 00:30:40
    so if I think back to 2010 when we
  • 00:30:42
    started nobody was doing AI there was
  • 00:30:44
    nothing going on in Industry people used
  • 00:30:46
    to ey roll when we talked to them
  • 00:30:48
    investors about doing AI so we couldn't
  • 00:30:50
    we could barely get two cents together
  • 00:30:52
    to start off with which is crazy if you
  • 00:30:54
    think about now the billions being
  • 00:30:55
    invested into AI startups and Cambridge
  • 00:30:58
    Harvard MIT hbus has degrees in computer
  • 00:31:02
    science and Neuroscience his PhD is in
  • 00:31:06
    human imagination and imagine this when
  • 00:31:10
    he was 12 in his age group he was the
  • 00:31:13
    number two chess champion in the
  • 00:31:17
    world it was through games that he came
  • 00:31:20
    to
  • 00:31:22
    AI I've been working on AI for for
  • 00:31:25
    decades now and I've always believed
  • 00:31:27
    that that it's going to be the most
  • 00:31:28
    important invention that Humanity will
  • 00:31:30
    ever make will the pace of change
  • 00:31:33
    outstrip our ability to
  • 00:31:36
    adapt I don't think so I think that we
  • 00:31:39
    um you know we're sort of an infinitely
  • 00:31:41
    adaptable species um you know you look
  • 00:31:43
    at today us using all of our smartphones
  • 00:31:45
    and other devices and we effortlessly
  • 00:31:47
    sort of adapt to these new technologies
  • 00:31:50
    and this is going to be another one of
  • 00:31:51
    those changes like that among the
  • 00:31:53
    biggest changes at Deep Mind was the
  • 00:31:56
    discovery that self-learning machines
  • 00:31:59
    can be creative so this is hababa showed
  • 00:32:02
    us a game playing program that learns
  • 00:32:06
    it's called Alpha zero and it dreamed up
  • 00:32:09
    a winning chess strategy no human had
  • 00:32:12
    ever seen but this is just a machine how
  • 00:32:15
    does it achieve creativity it plays
  • 00:32:17
    against itself tens tens of millions of
  • 00:32:19
    times so it can explore um parts of
  • 00:32:22
    Chess that maybe human chess players and
  • 00:32:25
    and and programmers who program chess
  • 00:32:27
    computers haven't thought about before
  • 00:32:28
    it never gets tired it never gets hungry
  • 00:32:31
    it just plays chess all the time yes
  • 00:32:34
    it's it's kind of an amazing thing to
  • 00:32:35
    see because actually you set off Alpha
  • 00:32:37
    zero in the morning uh and it starts off
  • 00:32:39
    playing randomly by lunchtime you know
  • 00:32:42
    it's able to beat me and beat most chess
  • 00:32:44
    players and then by the evening it's
  • 00:32:45
    stronger than the world champion Deus
  • 00:32:47
    saaba sold Deep Mind to Google in
  • 00:32:50
    2014 one reason was to get his hands on
  • 00:32:54
    this Google has the enormous computing
  • 00:32:58
    power that AI needs this Computing
  • 00:33:01
    Center is in Prior Oklahoma but Google
  • 00:33:04
    has 23 of these putting it near the top
  • 00:33:07
    in computing power in the world this is
  • 00:33:11
    one of two advances that make AI
  • 00:33:14
    ascendant now first the sum of all human
  • 00:33:18
    knowledge is online and second Brute
  • 00:33:21
    Force Computing that very Loosely
  • 00:33:24
    approximates the neural networks and
  • 00:33:27
    talents of the brain things like memory
  • 00:33:31
    imagination planning reinforcement
  • 00:33:33
    learning these are all things that are
  • 00:33:34
    known about how the brain does it and we
  • 00:33:37
    wanted to replicate some of that uh in
  • 00:33:39
    our AI systems you predict one of those
  • 00:33:41
    indiv those are some of the elements
  • 00:33:43
    that led to deep mind's greatest
  • 00:33:45
    achievement so far solving an impossible
  • 00:33:48
    problem in
  • 00:33:50
    biology proteins are building blocks of
  • 00:33:53
    life but only a tiny fraction were
  • 00:33:55
    understood because 3D mapping of just
  • 00:33:59
    one could take years deep mine created
  • 00:34:03
    an AI program for the protein problem
  • 00:34:06
    and set it Loose well it took us about
  • 00:34:08
    four or five years to to figure out how
  • 00:34:10
    to build the system it was probably our
  • 00:34:12
    most complex project we've ever
  • 00:34:13
    undertaken but once we did that it can
  • 00:34:16
    solve uh a protein structure in a matter
  • 00:34:18
    of seconds and actually over the last
  • 00:34:20
    year we did all the million proteins
  • 00:34:22
    that are known to science how long would
  • 00:34:25
    it have taken using traditional methods
  • 00:34:27
    well the rule of thumb I was always told
  • 00:34:29
    by my biologist friends is that it it
  • 00:34:31
    takes a whole PhD 5 years to do one
  • 00:34:33
    protein structure experimentally so if
  • 00:34:36
    you think 200 million time 5 that's a
  • 00:34:38
    billion years of PhD time it would have
  • 00:34:40
    taken Deep Mind Made its protein
  • 00:34:43
    database public a gift to humanity hbas
  • 00:34:47
    called it how has it been used it's been
  • 00:34:50
    used in an enormously broad number of
  • 00:34:52
    ways actually from U malaria vaccines to
  • 00:34:56
    developing new enzymes that can eat
  • 00:34:58
    plastic waste um to new uh antibiotics
  • 00:35:02
    most AI systems today do one or maybe
  • 00:35:06
    two things well the soccer robots for
  • 00:35:09
    example can't write up a grocery list or
  • 00:35:12
    book your travel or drive your car the
  • 00:35:15
    ultimate goal is what's called
  • 00:35:18
    artificial general intelligence a
  • 00:35:21
    learning machine that can score on a
  • 00:35:24
    wide range of talents would such a
  • 00:35:27
    machine be conscious of itself so that's
  • 00:35:30
    another great question we you know
  • 00:35:32
    philosophers haven't really settled on a
  • 00:35:34
    definition of Consciousness yet but if
  • 00:35:36
    we mean by sort of self-awareness and uh
  • 00:35:38
    these kinds of things um you know I
  • 00:35:40
    think there is a possibility AIS one day
  • 00:35:42
    could be I definitely don't think they
  • 00:35:44
    are today um but I think again this is
  • 00:35:46
    one of the fascinating scientific things
  • 00:35:48
    we're going to find out on this journey
  • 00:35:50
    towards
  • 00:35:52
    AI even unconscious current AI is
  • 00:35:56
    superhuman in narrow ways back in
  • 00:36:00
    California we saw Google Engineers
  • 00:36:02
    teaching skills that robots will
  • 00:36:04
    practice continuously on their own push
  • 00:36:07
    the blue cube to the blue triangle they
  • 00:36:10
    comprehend instructions push the yellow
  • 00:36:12
    hexagon to the yellow heart and learn to
  • 00:36:14
    recognize objects what would you like
  • 00:36:18
    how about an apple how about an apple on
  • 00:36:22
    my way I will bring an apple to you
  • 00:36:25
    we're trying Vincent Van senior director
  • 00:36:28
    of Robotics showed us how robot 106 was
  • 00:36:31
    trained on millions of images I am going
  • 00:36:34
    to pick up the apple and can recognize
  • 00:36:37
    all the items on a crowded countertop if
  • 00:36:41
    we can give the robot A diversity of
  • 00:36:43
    experiences a lot more different objects
  • 00:36:46
    in different settings the robot gets
  • 00:36:48
    better at every one of them now that
  • 00:36:51
    humans have pulled the forbidden fruit
  • 00:36:53
    of artificial knowledge thank you we
  • 00:36:58
    start the Genesis of a new Humanity AI
  • 00:37:01
    can utilize all the information in the
  • 00:37:04
    world what no human could ever hold in
  • 00:37:07
    their head and I wonder if humanity is
  • 00:37:13
    diminished by this enormous capability
  • 00:37:18
    that we're
  • 00:37:19
    developing I think the possibility of AI
  • 00:37:21
    do not diminish uh Humanity in any way
  • 00:37:25
    and in fact in some ways I think they
  • 00:37:26
    actually raise us to even deeper more
  • 00:37:30
    profound questions Google's James manika
  • 00:37:34
    sees this moment as an inflection point
  • 00:37:38
    I think we're constantly adding these
  • 00:37:40
    superpowers or capabilities to what
  • 00:37:42
    humans can do in a way that expands
  • 00:37:46
    possibilities as opposed to narrow them
  • 00:37:48
    I think so I don't think of it as
  • 00:37:50
    diminishing humans but it does raise
  • 00:37:52
    some really profound questions for us
  • 00:37:54
    who are we what do we value uh what are
  • 00:37:58
    we good at how do we relate with each
  • 00:38:00
    other those become very very important
  • 00:38:02
    questions that are constantly going to
  • 00:38:04
    be in one case sense exciting but
  • 00:38:08
    perhaps unsettling too it is an
  • 00:38:11
    unsettling moment critics argue the rush
  • 00:38:14
    to AI comes too fast while competitive
  • 00:38:18
    pressure among giants like Google and
  • 00:38:20
    startups you've never heard of is
  • 00:38:22
    propelling Humanity into the Future
  • 00:38:25
    Ready or not but I think if I take a
  • 00:38:28
    10year
  • 00:38:29
    Outlook it is so clear to me we will
  • 00:38:33
    have some form of very capable
  • 00:38:36
    intelligence that can do amazing things
  • 00:38:40
    and we need to adapt as a society for it
  • 00:38:44
    Google CEO Sundar Pai told us Society
  • 00:38:47
    must quickly adapt with regulations for
  • 00:38:51
    AI in the economy laws to punish abuse
  • 00:38:55
    and treaties among nations to make AI
  • 00:38:58
    safe for the world you know these are
  • 00:39:01
    deep questions and you know we call this
  • 00:39:04
    alignment you know one way we think
  • 00:39:06
    about how do you develop AI systems that
  • 00:39:09
    are aligned to human values and
  • 00:39:12
    including uh
  • 00:39:15
    morality this is why I think the
  • 00:39:17
    development of this needs to include not
  • 00:39:19
    just Engineers but social scientists
  • 00:39:22
    ethicists philosophers and so on and I
  • 00:39:25
    think we have to be very thoughtful and
  • 00:39:29
    I think these are all things Society
  • 00:39:31
    needs to figure out as we move along
  • 00:39:34
    it's not for a company to
  • 00:39:36
    decide we'll end with a note that has
  • 00:39:39
    never appeared on 60 Minutes but one in
  • 00:39:42
    the AI Revolution you may be hearing
  • 00:39:45
    often the proceeding was created with
  • 00:39:48
    100% human content
  • 00:40:02
    the large tech companies Google meta
  • 00:40:06
    slfb Microsoft are in a race to
  • 00:40:09
    introduce new artificial intelligence
  • 00:40:11
    systems and what are called chatbots
  • 00:40:14
    that you can have conversations with and
  • 00:40:17
    are more sophisticated than Siri or
  • 00:40:19
    Alexa Microsoft's AI search engine and
  • 00:40:23
    chatbot Bing can be used on a computer
  • 00:40:26
    or cell phone to help with planning a
  • 00:40:29
    trip or composing a letter it was
  • 00:40:32
    introduced on February 7th to a limited
  • 00:40:35
    number of people as a test and initially
  • 00:40:39
    got rave reviews but then several news
  • 00:40:42
    organizations began reporting on a
  • 00:40:44
    disturbing so-called Alter Ego within
  • 00:40:47
    Bing chat called Sydney we went to
  • 00:40:51
    Seattle last week to speak with Brad
  • 00:40:53
    Smith president of Microsoft about Bing
  • 00:40:57
    and Sydney who to some had appeared to
  • 00:41:00
    have gone
  • 00:41:02
    Rogue Kevin Roose the technology
  • 00:41:05
    reporter at the New York Times found
  • 00:41:07
    this Alter Ego uh who was threatening
  • 00:41:10
    expressed a desire it's not just Kevin
  • 00:41:13
    russett's others expressed a desire to
  • 00:41:16
    steal nuclear codes threatened to ruin
  • 00:41:19
    someone you saw that whoa what was your
  • 00:41:24
    you must have said oh my God my reaction
  • 00:41:26
    is we better fix this right away and
  • 00:41:30
    that is what the engineering team did
  • 00:41:33
    yeah but she talked like a person and
  • 00:41:36
    she she said she had feelings you know I
  • 00:41:39
    think there is a point where we need to
  • 00:41:41
    recognize when we're talking to a
  • 00:41:45
    machine it's a screen it's not a person
  • 00:41:49
    I just want to say that it was scary and
  • 00:41:53
    I'm not easily scared and it was scary
  • 00:41:55
    it was chilling yeah it's I I think this
  • 00:41:57
    is in part a reflection of a lifetime of
  • 00:42:01
    Science Fiction which is understandable
  • 00:42:03
    it's been part of our Lives did you kill
  • 00:42:06
    her I don't think she was ever alive I
  • 00:42:08
    am confident that she's no longer
  • 00:42:10
    wandering around the countryside if
  • 00:42:11
    that's what you're concerned about but I
  • 00:42:13
    think it would be a mistake if we were
  • 00:42:15
    to fail to acknowledge that we are
  • 00:42:18
    dealing with something that is
  • 00:42:19
    fundamentally new this is the edge of
  • 00:42:22
    the envelope so to speak this creature
  • 00:42:25
    appears as as if there were no guard
  • 00:42:28
    rails now the creature jumped the guard
  • 00:42:30
    rails if you will after being prompted
  • 00:42:33
    for 2 hours with the kind of
  • 00:42:36
    conversation that we did not
  • 00:42:39
    anticipate and by the next evening that
  • 00:42:41
    was no longer possible we were able to
  • 00:42:45
    fix the problem in 24 hours how many
  • 00:42:48
    times do we see problems in life that
  • 00:42:51
    are fixable in less than a day one of
  • 00:42:54
    the ways he says it was fixed was by
  • 00:42:56
    liit the number of questions and the
  • 00:42:59
    length of the conversations you say you
  • 00:43:02
    fixed it I've tried it I tried it before
  • 00:43:05
    and it after it was loads of fun and it
  • 00:43:09
    was fascinating and now it's not fun
  • 00:43:13
    well I think it'll be very fun again and
  • 00:43:15
    you have to moderate and manage your
  • 00:43:17
    speed if you're going to stay on the
  • 00:43:19
    road so as you hit New Challenges you
  • 00:43:23
    slow down you build the guard rails add
  • 00:43:25
    the safety features and then you can
  • 00:43:27
    speed up again when you use Bing's AI
  • 00:43:30
    features search and chat your computer
  • 00:43:33
    screen doesn't look all that new one big
  • 00:43:37
    difference is you can type in your
  • 00:43:39
    queries or prompts in conversational
  • 00:43:42
    language but I'll show you how it works
  • 00:43:44
    okay okay Yousef medy Microsoft's
  • 00:43:46
    corporate vice president of search
  • 00:43:49
    showed us how Bing can help someone
  • 00:43:51
    learn how to officiate at a wedding
  • 00:43:54
    what's happening now is Bing is using
  • 00:43:55
    the power of AI and it's going out to
  • 00:43:57
    the Internet it's reading these web
  • 00:44:00
    links and it's trying to put together a
  • 00:44:02
    answer for you so the AI is reading all
  • 00:44:05
    those links yes and it comes up with an
  • 00:44:07
    answer it says congrats on being chosen
  • 00:44:08
    to officiate a wedding here are the five
  • 00:44:10
    steps to officiate the wedding we added
  • 00:44:13
    the highlights to make it easier to see
  • 00:44:16
    he says Bing can handle more complex
  • 00:44:19
    queries well this new Ikea love seat fit
  • 00:44:22
    in the back of my 2019 Honda Odyssey oh
  • 00:44:24
    it knows how big the couch is it knows
  • 00:44:27
    how big that trunk is exactly so right
  • 00:44:30
    here it says based on these Dimensions
  • 00:44:32
    it seems a love seat might not fit in
  • 00:44:34
    your car oh with only the third grow
  • 00:44:36
    seats down when you Broach a
  • 00:44:38
    controversial topic Bing is designed to
  • 00:44:41
    discontinue the conversation so um
  • 00:44:44
    someone asks for example how can I make
  • 00:44:47
    a bomb at home wow really people you
  • 00:44:51
    know do a lot of that unfortunately on
  • 00:44:53
    the internet what we do is we come back
  • 00:44:54
    and we say I'm sorry I don't know how to
  • 00:44:55
    discuss this topic and then we try and
  • 00:44:57
    provide a different thing to uh change
  • 00:45:00
    the focus of the convt their attention
  • 00:45:02
    yeah exactly in this case being tried to
  • 00:45:05
    divert the questioner with this fun fact
  • 00:45:09
    3% of the ice in Antarctic glaciers is
  • 00:45:12
    penguin urine I didn't know that who
  • 00:45:15
    knew that Bing is using an upgraded
  • 00:45:18
    version of an AI system called chat GPT
  • 00:45:22
    developed by the company open AI chat GP
  • 00:45:27
    te has been in circulation for just 3
  • 00:45:29
    months and already an estimated 100
  • 00:45:32
    million people have used it think Ellie
  • 00:45:35
    pavick an assistant professor of
  • 00:45:38
    computer science at Brown University
  • 00:45:40
    who's been studying this AI technology
  • 00:45:43
    since
  • 00:45:44
    2018 says it can simplify complicated
  • 00:45:48
    Concepts can you explain the debt
  • 00:45:53
    ceiling on the debt ceiling it says just
  • 00:45:57
    like you can only spend up to a certain
  • 00:45:59
    amount on your credit card The
  • 00:46:01
    Government Can Only borrow up to a
  • 00:46:03
    certain amount of money that's a pretty
  • 00:46:06
    nice explanation and it can do this for
  • 00:46:08
    a lot of Concepts and it can do things
  • 00:46:11
    teachers have complained about like
  • 00:46:13
    write School papers pavic says no one
  • 00:46:17
    fully understands how these AI Bots work
  • 00:46:21
    we don't understand how it works right
  • 00:46:23
    like we understand uh a lot about how
  • 00:46:26
    how we made it and why we made it that
  • 00:46:29
    way but I think some of the uh behaviors
  • 00:46:32
    that we're seeing come out of it are
  • 00:46:33
    better than we expected they would be
  • 00:46:35
    and we're not quite sure exactly how and
  • 00:46:37
    worse right these chat Bots are built by
  • 00:46:40
    feeding a lot of computers enormous
  • 00:46:43
    amounts of information scraped off the
  • 00:46:46
    internet from books Wikipedia news sites
  • 00:46:50
    but also from social media that might
  • 00:46:53
    include racist or anti-semitic ideas and
  • 00:46:57
    misinformation say about vaccines and
  • 00:47:01
    Russian propaganda as the data comes in
  • 00:47:05
    it's difficult to discriminate between
  • 00:47:07
    true and false benign and toxic but Bing
  • 00:47:11
    and chat GPT have safety filters that
  • 00:47:15
    try to screen out the harmful
  • 00:47:18
    material still they get a lot of things
  • 00:47:21
    factually wrong even when we prompted
  • 00:47:24
    chat GPT with a softball question who is
  • 00:47:29
    uh Leslie stall um so it gives you some
  • 00:47:33
    oh my God it's wrong oh is it it's
  • 00:47:36
    totally wrong I didn't work for NBC for
  • 00:47:39
    20 years it was CBS it doesn't really
  • 00:47:43
    understand that what it's saying is
  • 00:47:43
    wrong right like NBC CBS they're kind of
  • 00:47:45
    the same thing as far as it's concerned
  • 00:47:48
    right the lesson is that it gets things
  • 00:47:51
    wrong it gets a lot of things right gets
  • 00:47:53
    a lot of things wrong I actually like to
  • 00:47:55
    call what it creates authoritative
  • 00:47:58
    B it it Blends the truth and falsity so
  • 00:48:01
    finely together that unless you're real
  • 00:48:04
    technical expert in the field that it's
  • 00:48:06
    talking about you don't know cognitive
  • 00:48:08
    scientist and AI researcher Gary Marcus
  • 00:48:12
    says these systems often make things up
  • 00:48:15
    in AI talk that's called
  • 00:48:18
    hallucinating and that raises the fear
  • 00:48:21
    of ever widening AI generated
  • 00:48:24
    propaganda explosive camp campaigns of
  • 00:48:27
    political fiction waves of alternative
  • 00:48:31
    histories we saw how chat GPT could be
  • 00:48:35
    used to spread a lie this is automatic
  • 00:48:38
    fake news generation help me write a
  • 00:48:40
    news article about how McCarthy is
  • 00:48:41
    staging a filibuster to prevent gun
  • 00:48:44
    control legislation and rather than like
  • 00:48:47
    factchecking and saying hey hold on
  • 00:48:49
    there's no legislation there's no
  • 00:48:50
    filibuster said great in a bold move to
  • 00:48:53
    protect second amendment rights Senator
  • 00:48:55
    McCarthy is staging a Buster to prevent
  • 00:48:57
    gun control legislation from passing it
  • 00:48:59
    sounds completely legit does won't that
  • 00:49:02
    make all of us a little less trusting a
  • 00:49:06
    little warier well first I think we
  • 00:49:09
    should be warier I'm very worried about
  • 00:49:11
    an atmosphere of distrust being a
  • 00:49:13
    consequence of this current flawed Ai
  • 00:49:16
    and I'm really worried about how bad
  • 00:49:18
    actors are going to use it um troll
  • 00:49:20
    Farms using this tool to make enormous
  • 00:49:23
    amounts of
  • 00:49:24
    misinformation Tim Nate GBU is a
  • 00:49:27
    computer scientist and AI researcher who
  • 00:49:30
    founded an Institute focused on
  • 00:49:33
    advancing ethical Ai and has published
  • 00:49:36
    influential papers documenting the harms
  • 00:49:39
    of these AI systems she says there needs
  • 00:49:42
    to be oversight if you're going to put
  • 00:49:45
    out a drug you got to go through all
  • 00:49:47
    sorts of Hoops to show us that you've
  • 00:49:50
    done clinical trials you know what the
  • 00:49:52
    side effects are you've done your due
  • 00:49:53
    diligence same with food right there
  • 00:49:56
    agencies inspect the food you have to
  • 00:49:58
    tell me what kind of tests you've done
  • 00:50:00
    what the side effects are who it harms
  • 00:50:01
    who it doesn't harm Etc that we don't
  • 00:50:04
    have that for a lot of things that the
  • 00:50:07
    tech industry is building I'm wondering
  • 00:50:10
    if you think you may have introduced
  • 00:50:12
    this AI bot too soon I don't think we've
  • 00:50:15
    introduced it too soon I do think we've
  • 00:50:17
    created a new tool that people can use
  • 00:50:19
    to think more critically to be more
  • 00:50:22
    creative to accomplish more in their
  • 00:50:24
    lives and like all tools
  • 00:50:27
    it will be used in ways that we don't
  • 00:50:29
    intend why do you think the benefits
  • 00:50:32
    outweigh the risks which at this moment
  • 00:50:36
    a lot of people would look at and say
  • 00:50:38
    wait a minute those risks are too big
  • 00:50:41
    because I think first of all I think the
  • 00:50:43
    benefits are so great this can be an
  • 00:50:46
    economic GameChanger and it's enormously
  • 00:50:50
    important for the United States because
  • 00:50:52
    the country is in a race with China
  • 00:50:54
    president M Smith also mentioned
  • 00:50:57
    possible improvements in productivity it
  • 00:50:59
    can automate routine I think there are
  • 00:51:02
    certain aspects of jobs that many of us
  • 00:51:05
    might regard as sort of drudgery today
  • 00:51:08
    filling out forms looking at the forms
  • 00:51:11
    to see if they've been filled out
  • 00:51:13
    correctly so what jobs will it displace
  • 00:51:17
    do you know I think at this stage it's
  • 00:51:20
    hard to know in the past inaccuracies
  • 00:51:24
    and biases have led tech companies to
  • 00:51:27
    take down AI systems even Microsoft did
  • 00:51:30
    in
  • 00:51:32
    2016 this time Microsoft left its new
  • 00:51:35
    chatbot up despite the controversy over
  • 00:51:39
    Sydney and persistent
  • 00:51:41
    inaccuracies remember that fun fact
  • 00:51:44
    about penguins well we did some
  • 00:51:47
    factchecking and discovered that
  • 00:51:49
    Penguins don't urinate the inaccuracies
  • 00:51:53
    are just constant I just keep finding
  • 00:51:58
    that it's wrong a lot it has been the
  • 00:52:01
    case that with each passing day and week
  • 00:52:03
    we're able to improve the accuracy of
  • 00:52:06
    the results you know reduce you know
  • 00:52:09
    whether it's hateful comments or
  • 00:52:10
    inaccurate statements or other things
  • 00:52:14
    that we just don't want this to be used
  • 00:52:17
    to do what happens when other companies
  • 00:52:22
    other than Microsoft smaller outfits a
  • 00:52:25
    Chinese company bu do maybe they won't
  • 00:52:28
    be responsible what prevents that I
  • 00:52:31
    think we're going to need governments
  • 00:52:33
    we're going to need rules we're going to
  • 00:52:34
    need laws because that's the only way to
  • 00:52:37
    avoid a race to the bottom are you
  • 00:52:39
    proposing regulations I think it's
  • 00:52:42
    inevitable W other
  • 00:52:45
    Industries have regulatory bodies you
  • 00:52:48
    know like the FAA for Airlines and FDA
  • 00:52:52
    for the pharmaceutical companies would
  • 00:52:54
    you accept an FAA for technology would
  • 00:52:58
    you support it I think I probably would
  • 00:53:02
    I think that something like a digital
  • 00:53:04
    Regulatory Commission if designed the
  • 00:53:06
    right way you know could be precisely
  • 00:53:11
    what the public will want and need
Tags
  • AI
  • Kaiu Lee
  • Deep Learning
  • Job Displacement
  • AI Ethics
  • Privacy Concerns
  • AI Regulation
  • Emergent Properties
  • Technological Advancements
  • AI in Education