Should Computers Run the World? – with Hannah Fry
Tagged on:                                     

100 thoughts on “Should Computers Run the World? – with Hannah Fry

  • May 8, 2019 at 11:28 am
    Permalink

    I didn't have a problem hearing the difference between the real and fake Bach.

    Reply
  • May 8, 2019 at 11:40 am
    Permalink

    No.
    But I would still recommend watching the video.

    Reply
  • May 8, 2019 at 12:38 pm
    Permalink

    Algorithms are far less biased than people. We see their decisions as unjust because we are biased to the context instead of relying on facts.

    Reply
  • May 8, 2019 at 1:59 pm
    Permalink

    18:30 I heard about this study in the book "Thinking Fast and Slow" by Daniel Kahnemann I believe. There it has been really well described.

    Also how long will it take to let a pigeon diagnose my sicknesses? Would be really cool to tell the workplace "Sorry, my pigeon doctor said I am not well enough to work today."

    Reply
  • May 8, 2019 at 6:30 pm
    Permalink

    Its people running algorithms on a computer… Shovels don't dig holes.

    Reply
  • May 9, 2019 at 3:16 am
    Permalink

    It is sad that most peoples don't understand the dangers of these developments.
    First of all, most of us will become obsolete because advance algorithms will take over many jobs.
    And second, those in control will give them self even more control over many peoples.

    So yes, I love technology and I'd like to use it as an aid. But unfortunately those in power
    will eventually also use it to control us all. I mean it has already started. YouTube, FaceBook, InstaGram
    etc. All are already using algorithms to control the way we experience and receive information.

    All based on algorithms they think is good for us. But I don't want any of that. I don't want
    a self refreshing page that changes every time. Or that google knows where I live and
    work so it suggest the best time and route to take in the morning. And after a while it even knows
    my day routine.

    See, I wouldn't mind using those technologies if I had full control of it. But it's in the cloud, on their servers
    they can use that information in many ways. I don't like that idea. I have no control. For example you can't
    even use their services if you don't agree with the terms and conditions. And many don't know this but
    almost all terms and condition requires you to share all your information with them and allow them to
    use it to "help" you.

    I am forced to go with the flow. If I don't go with the flow I'll be flagged according to the statistics to be a potential
    danger!? While all I want is to have a choice in what I want. So yes it'll be nice, but humans need to change
    first or many of us will lose.

    Our world is not ready yet to trust computers to run the world.
    Humans need to change their mentality first!

    Reply
  • May 9, 2019 at 5:50 am
    Permalink

    Men run the world, women stay in kitchen!

    Reply
  • May 9, 2019 at 7:41 am
    Permalink

    What if you run an machine learning algorithm on a quantum computer…

    Reply
  • May 9, 2019 at 12:04 pm
    Permalink

    I think the computers should provide advice to everyone in the world, not govern the world. If you Afford all equal benefit from knowlege, it becomes better for everyone.

    Reply
  • May 9, 2019 at 2:06 pm
    Permalink

    That stupid pigeon is me

    Reply
  • May 9, 2019 at 4:14 pm
    Permalink

    Should or Do…

    Reply
  • May 9, 2019 at 4:24 pm
    Permalink

    Sorry but I'm not gonna trust a computer to drive my car for me. Driving is freedom

    Reply
  • May 9, 2019 at 8:33 pm
    Permalink

    Thank you!

    Reply
  • May 10, 2019 at 1:34 am
    Permalink

    So what she is saying is that there is already a beginning gender based pay gap with pigeons….

    Reply
  • May 10, 2019 at 7:40 am
    Permalink

    Humans are uniquely capable of training pigeons to detect cancer. She also explains in the lecture how humans are uniquely capable in other ways.

    Reply
  • May 10, 2019 at 11:02 am
    Permalink

    It should be tried – imagine objective law and power – so much money and opportunities is stolen by the mates of the elected// We lack experience with a society where the 'I scratch yours and you mine' is up to scratch = no hidden harm/externalities to the world outside. I would chose AI to decide upon me… even in Germany.

    Reply
  • May 10, 2019 at 5:50 pm
    Permalink

    I knew it was option 2, because of the very specific 5-10 sec start.

    Reply
  • May 11, 2019 at 2:48 am
    Permalink

    I was born to go to the movies with you guys last night I had to leave for work so I don’t know if I have time to come over tomorrow or not I just want to let you all know I can talk to you later bye

    Reply
  • May 11, 2019 at 4:11 am
    Permalink

    Balance and symbiosis, thanks!

    Reply
  • May 11, 2019 at 10:17 am
    Permalink

    Algorithm is only unbiased if its training data (and possibly the evaluation function) are unbiased. And that is unfortunately never the case.

    Reply
  • May 11, 2019 at 12:54 pm
    Permalink

    Just some police forces controlling a much larger crowd of people is something like 'not desireable' in democratic countries. It is always worth thinking twice before letting algorithms decide what to do over – or against – people's will. It has something to do with the phrase 'the people' as is, not just here in Germany.

    Reply
  • May 11, 2019 at 2:37 pm
    Permalink

    Absolutely brilliant and fascinating… her knowledge and ability to communicate it in a simple and concise manner is fantastic.

    Reply
  • May 11, 2019 at 7:44 pm
    Permalink

    idk that stupid bird is destined for greatness….he will be president of the world someday u just wait

    Reply
  • May 12, 2019 at 11:27 am
    Permalink

    Hannah Fry is my favourite

    Reply
  • May 12, 2019 at 4:33 pm
    Permalink

    I strongly disagree with Hannah's conclusion on 21:41 – "An algorithm is never going to be perfect, because they just don't understand the World in the same way that we do." She just spoke about the horrible biases human judges are affected by and still manages to assume, that the human way of modelling the world is not just better, but perfect? An ideal for machine learning to strive for?
    ".. They don't understand context and they don't understand nuance." She's making this generalization about all AI based on an assumption that future AI will be of similar precision as the algorithms "that we've invited in our lives so far". Instead the current speed of progress in accuracy of predictive models is sound evidence, that current accuracy is not representative of future accuracy and is in fact growing fast.

    Regarding not understanding context and nuance – predictive models take into account all the nuance and context that we give them proportionally to how predictive this context and nuance is of the results. If we disagree with the results, then either we are wrong or we didin't provide data that would include this nuance we consider self-evident, which is probably the case with considering younger sex offenders more dangerous.
    Was that a human-designed algorithm or a predictive ML model? In case of the latter, did its training data contain the victim's age? I bet not.
    This is exactly the case with image classification models that imagine herd animals – their training data probably doesn't contain panoramic images, as the humans selecting the images passed on their biased assumption, that no one will want to classify panoramic images. And cattle is the best fit out of the training data they've learned.
    This is a common mistake people make when learning new words – assigning the best fitting known category to an unknown category instad of saying "I don't know". It would be interesting to know what probability did the image recognition model assign to their best guess – the "herding animal" class. Maybe it was really low and the model wasn't programmed to say "I don't know" for low probabilities.

    Reply
  • May 12, 2019 at 6:13 pm
    Permalink

    You'd think an algorithm doesn't have bias, but not so long ago scientists training an AI noticed that they accidentally had taught it to be biased despite conscious attempts to only give objective input. The AI still picked up on their personal preferences. Completely forgot how though, sadly.
    OK, an AI is not the same thing as a mathematical algorithm, but the more complex a task, the more likely we're going to be using AI for it.

    Still doubt it's going to be as strong as biases from people though.

    Reply
  • May 12, 2019 at 6:31 pm
    Permalink

    So the answer is augmented humans

    Reply
  • May 13, 2019 at 3:05 am
    Permalink

    I think the whole "missing the gorilla in the scan" thing is kind of silly . . . that just tells me they were very focused on looking for tumors (which is a good thing) and not gorillas (which is also a good thing) . . .

    Reply
  • May 13, 2019 at 3:15 am
    Permalink

    I don't think even humans should be able to make laws unless they explain why. "Why" is a very important question.

    Reply
  • May 13, 2019 at 11:13 am
    Permalink

    Seriously though?! …. Birds did that well?!

    Reply
  • May 13, 2019 at 5:45 pm
    Permalink

    Hannah Fry is cute af

    Reply
  • May 13, 2019 at 6:10 pm
    Permalink

    How about, "Should computers/algorithms fly planes, where they're allowed to prevent human correction?" Mention of the Boeing 737 Max 8 is so suspiciously missing, that I thought this video was aged by years, rather than weeks old. As RI's Matt Parkers says, "You've included data that supports your hypothesis, what about the data that doesn't support your hypothesis?"
    (Further, pigeons are excellent pilots and navigators – better than computers; why this insinuation that they're laughably incompetent? With what regard do we hold human pilots, that we think pilots are near idiots?)

    Reply
  • May 13, 2019 at 6:56 pm
    Permalink

    Very nice talk, it underlines completely why people need to have general understanding of how things work. In the case of the benefits spreadsheet and image recognition, we need legislation that forces a basic level of competence in the task it is being applied to on data it has never seen before. This kind of thing is standard among anyone designing this kind of technology, but the problem is when the algorithms are turned over to non-experts or with companies that want to over-sell the capability of their products to non-experts. With the image recognition, they are super accurate in face recognition or verification, but in the general task of identifying images (and scene description even moreso) the results are very varied. In most real-life applications, however, we don't want to solve the general task, but a much more specific one.

    Reply
  • May 13, 2019 at 11:29 pm
    Permalink

    Hannah Fry make me more confident about being a woman.

    Reply
  • May 13, 2019 at 11:33 pm
    Permalink

    I'd trust the world under the magi supercomputers in nge hehe

    Reply
  • May 13, 2019 at 11:41 pm
    Permalink

    All the comments below FOCUSED on the Speaker…. She is telling you that A.I. is taking over every body`s job and the audience is hypnotized by her eloquence… Her mission is to fool people. Several of those references of studies that she used to cut down humans are misrepresented. A.I. has to be a limited tool for humans. Once it becomes independent we will be f""ked.

    Reply
  • May 14, 2019 at 5:11 am
    Permalink

    I for one welcome our new robot overlords.

    Reply
  • May 14, 2019 at 6:03 am
    Permalink

    I wouldn't trust a programmer or group of programmers to not write their own bias into their algorithm or AI.

    Reply
  • May 14, 2019 at 6:44 am
    Permalink

    In case you are wondering how to tell the difference between a real composer and a computer-generated composition, the difference is in the consideration of the room. If you listen closely to sample 2, the room was very much clearly defined in the music. Stereoed, echoed, everything well-considered and deliberate. Sample 1 had absolutely no consideration of the room.

    Reply
  • May 14, 2019 at 9:56 am
    Permalink

    Apparently, IBM built a system for the State of Indiana’s human services department that sent rejection letters to toddlers because they failed to with the state’s requirement that they respond to a questionnaire in writing in a timely fashion.

    Reply
  • May 14, 2019 at 3:06 pm
    Permalink

    computers allrdy run the world for a long time now. nothing works anymore without computers allrdy.

    Reply
  • May 14, 2019 at 6:13 pm
    Permalink

    Great talk/presentation! Unfortunately, machines, just like people, have unconscious bias depending on how they were built (e.g., face detection models being trained mostly with western European faces will have a harder time detecting non-western European faces).

    Reply
  • May 14, 2019 at 9:11 pm
    Permalink

    Scientists should be celebrities, not the childish thespians or jocks!

    Reply
  • May 15, 2019 at 1:41 pm
    Permalink

    Fully 'fledged' pathologists. Ha ah.

    Reply
  • May 15, 2019 at 4:31 pm
    Permalink

    I DEMAND HOSPITALS BE OVERRUN BY BIRD DOCTORS

    Reply
  • May 15, 2019 at 6:50 pm
    Permalink

    Depends on who's running the computers.

    Reply
  • May 15, 2019 at 11:05 pm
    Permalink

    Very nice examples and data. Too many questionable arguments.

    Reply
  • May 16, 2019 at 1:52 pm
    Permalink

    Somewhere, in a parallel universe, humanity did not invent advanced computer technology, but instead relied on vast farms of small containers with problem solving pidgeons.

    Reply
  • May 16, 2019 at 9:51 pm
    Permalink

    9:27 15% mildly aroused

    Reply
  • May 17, 2019 at 11:17 am
    Permalink

    Very interesting content, VERY well delivered.

    Reply
  • May 17, 2019 at 9:54 pm
    Permalink

    Great talk

    Reply
  • May 18, 2019 at 12:10 am
    Permalink

    It is a bit of an outrage to me that psychological tests are being called algorithms. This is because of the consequences to understanding AI. I wish more emphasis was put on construct validity. Algorithms are in fact tweaked by humans. It is us who feed the data. And it is us who set the goal for the algorithm. In essence, algorithms is often completely analogous to empiricism and statistical testing.

    Ps: regardless of my criticism, some great points have been made in this talk and I loved the talk c:

    Reply
  • May 19, 2019 at 2:51 pm
    Permalink

    AI will run the world, and might be our only hope against extinction from runaway capitalism.

    Reply
  • May 19, 2019 at 6:58 pm
    Permalink

    So the best condition for being in court, is when the judge recently celebrated his favorite sports team winning, by having sex and eating a sandwich.

    Reply
  • May 19, 2019 at 10:16 pm
    Permalink

    I believe we need to have computers literally be our government. We program seriously what we want all in all . The world needs to change with these numbers and does not which is another matter? And also as she is saying we believe we are the only species capable of change where is the change in religious ignorance of just change as an example? That the world moves and and computers and MRIs are revealing that if we dish out enviornments of abuse to our citizens we become the vicious animals you see in the zoo. And if you look at these vicious animals they usually tend to be not as bright as a pigeon in truth.

    Reply
  • May 19, 2019 at 10:20 pm
    Permalink

    Potential of artificial inteligence is truly enormous. Imagine what would be possible if you combine AI with quantum computing.

    Reply
  • May 19, 2019 at 10:21 pm
    Permalink

    There's a problem with us we arbitraily believe we can always do either or in life. The problem is we can make an isolated self in jail believing we fight someone else with either or. When you are extreme you truly Are fighting yourself and putting yourself in isolation in jail Always . We do this to our brains which are from nature how can a human mind understand something that is not natural ?
    Because we fear we have our enemy it's us.

    Reply
  • May 21, 2019 at 8:58 am
    Permalink

    What bothers me most about the human judge vs. robo judge analogy is that most people forget that judges are notoriously bad at removing bias from their decisions. A human judge decides whether you should be released subconsciously, and justifies this to you and to themselves by using their words. It would be the same as having the judge merely read the sentence after a computer had made it, because that's what's happening inside the judge's head anyway.

    The only sensible reason to prefer a human judge is because you can at least talk to them and try to persuade them. This may not work, but then you'd at least have felt a little less helpless. Using algorithms to predict recidivism is fine, by the way, but using them to actually decide whether someone should be released or not is ridiculous. People who have high risk of recidivism don't necessarily deserve to be in prison longer, they simply require more assistance after they're released.

    Reply
  • May 22, 2019 at 11:16 am
    Permalink

    It won't take too long before this human/machine partnership is able to clearly define what makes people so amazing at specificity, at which point humans will be factored out of the equation entirely in favor of optimization.
    So personally I'd say that there would likely be a long transition period where we eventually ended up with computers which should indeed run the world.

    Reply
  • May 23, 2019 at 7:53 pm
    Permalink

    Computers already run the world

    Reply
  • May 25, 2019 at 2:31 am
    Permalink

    Who should write the computer program, ahem, AI, that would govern you?

    Reply
  • May 27, 2019 at 5:35 am
    Permalink

    Computers could rule the world within a degree but not totally. Computers cannot feel pain, love, anxiety and other critical elements for humankind.

    Reply
  • May 28, 2019 at 2:20 pm
    Permalink

    Always enjoy listening to Hannah Fry explaining things. She will continue to do well for sure.

    Reply
  • June 1, 2019 at 12:01 pm
    Permalink

    The music experiment does not demonstrate that machines are as good as humans in composing or replicating music, it just demonstrates that the audience had a very poor musical taste… the first option had not the dissonances and usual melodic steps of Bach music; the two options were not so similar.

    Reply
  • June 1, 2019 at 6:18 pm
    Permalink

    The point she's making is
    Pigeons are less bird brained than those who waste money on ridiculous studies. 🙂

    Reply
  • June 4, 2019 at 11:23 am
    Permalink

    I am a pathologist, I think the problem here is that the diagnosis of breast cancer by histopathology slides is really an easy task, any medical student who trained in a pathology lab, can diagnose breast cancer in 1-2 weeks of training. I am really interested to see the results in really difficult cases, in conditions where someone have to judge using different resources together (history and physical examination , Imaging data, Lab data and histopathology slides) the results will be different.

    Reply
  • June 5, 2019 at 1:22 pm
    Permalink

    No. Because the programmers control the computers. And the programmers are controlled by the (most likely MIC) corporations contracted to build the system. And for once, the corps will bid as high as possible. And now a corporation runs the world without resistance. 🙁

    Reply
  • June 6, 2019 at 11:37 am
    Permalink

    anyone else had to think of American gods s2 with that Bach algorithm?

    Reply
  • June 13, 2019 at 1:05 pm
    Permalink

    30:11 Now, we can build a powerful nun-o-machine. MWAHAHAHAAAA!!

    Reply
  • June 14, 2019 at 7:14 am
    Permalink

    So a Technocracy, no thanks. Tools yes, decisions should be left to humans.
    I'm not distracted by pretty faces, this is just a con trick.

    Reply
  • June 14, 2019 at 9:48 am
    Permalink

    Pigeons can save the NHS!

    Reply
  • June 17, 2019 at 1:35 am
    Permalink

    22:40 those houses have the black and white texture of cattle

    Reply
  • July 4, 2019 at 5:49 pm
    Permalink

    Should Computers Run the World? How well did that work out for the A727 Airbus?

    Reply
  • July 5, 2019 at 4:36 pm
    Permalink

    It’s sad this world thinks more about lifeless technology than it does life.

    Reply
  • July 5, 2019 at 5:46 pm
    Permalink

    Short Answer: Yes, but only when they run my software.

    Reply
  • July 6, 2019 at 1:41 pm
    Permalink

    If you ever learned anything about bach then it isn't too hard to guess. The pauses in the first one are off and the harmonics boring and unimaginative.

    (That is coming from someone who does not regard computer creativity as impossible! It needs a better NN architecture though).

    Reply
  • July 8, 2019 at 5:32 pm
    Permalink

    To fool an AI judge, just change your legal name to ;DROP TABLE CONVICTS

    Reply
  • July 10, 2019 at 7:26 pm
    Permalink

    Great!

    Reply
  • July 13, 2019 at 11:00 am
    Permalink

    An excellent talk. This is the first time I've come across Hannah Fry. She's a really impressive presenter.

    Reply
  • July 15, 2019 at 3:14 pm
    Permalink

    Just wait 4 IBM 20/500 QUBIT Quantum device.

    Reply
  • July 16, 2019 at 2:40 pm
    Permalink

    damn it shouldnt be called just comps. anymore but computing systems.. as they are systems not just circutry, but anyhow great lecture..

    Reply
  • July 16, 2019 at 2:42 pm
    Permalink

    well actually algos are becoming soled out nearly by now, cause they are becoming part of a system we cant maintain relly..

    Reply
  • July 16, 2019 at 5:25 pm
    Permalink

    Hmmmph – these young pigeons today don't know how easy they've got it. In World War II a man named Skinner proposed, and was funded to do development work on pigeon guided bombs. The pigeons (in the nose cone of the bomb) would peck at an image of the target projected in front of them. If the target drifted off center, the pigeon pecking at the target image caused the screen to send a corrective signal to the guidance system. This was a one-way trip for the pigeons.

    Pigeons have also been tested, fairly successfully, in pharmaceutical manufacturing, picking bad pills and tablets off of a conveyor.

    Reply
  • July 30, 2019 at 5:34 pm
    Permalink

    Bit disturbed by the footage of the nuns immediately followed by their brains in tubs.

    Reply
  • July 31, 2019 at 3:33 pm
    Permalink

    If you want to make an ai, use a pigeon

    Reply
  • August 6, 2019 at 12:27 pm
    Permalink

    Love her

    Reply
  • August 7, 2019 at 9:01 pm
    Permalink

    Pidgeons taking American jobs..

    Reply
  • August 17, 2019 at 10:20 am
    Permalink

    Quit analysing everyone and everything . Our lives are on hold due to the never ending tests and data input . We are recording a life that we stopped living years ago , due to the constant logging of it all . Did you know that your average doctor spends more time on the paperwork than the patients now . Thats where testing everything has lead to . Most people spend hours and hours writing reports filling in forms logging movements etc . Stop logging life and live it instead .

    Reply
  • September 7, 2019 at 1:18 am
    Permalink

    Who will run the computers ? people who desperately want to. See China.

    Reply
  • September 7, 2019 at 1:18 am
    Permalink

    Women can't handle facts.

    Reply
  • September 15, 2019 at 5:18 pm
    Permalink

    Hanah you are absolutely mesmerizing !!! ❤️❤️❤️

    Reply
  • September 15, 2019 at 5:53 pm
    Permalink

    You are an absolute credit to academia and society ❤️

    Reply
  • October 5, 2019 at 11:43 pm
    Permalink

    I THINK THEY NEED TO REPLACE THE AMERICAN POLITICAL SYSTEM.THE CORRUPTION & GREED INHERENT IN WASHINGTON D.C WILL NOT STOP UNTIL A PARADIGM SHIFT OCCURS.

    Reply
  • October 12, 2019 at 5:16 pm
    Permalink

    Hannah can run my world any day.

    Reply
  • October 12, 2019 at 8:10 pm
    Permalink

    This is hilarious, a friend of mine is a radiologist who co-founded a company for training algorithms and using them in combination with human radiologists to basically increase detection rates to levels never reached before.
    Basically what the algorithm does is to look at scans from mammograms and marks malignant cell structures for the radiologist to check up on (it marks the whole scan as potentially malignant and encircles the area of the potential tumor). The ones that are normal are marked as such, so the radiologist doesn't have to check up on it.

    When he told me about it he basically said everything Hannah Fry said in her talk about humans:
    False positives nearly never occur, but the sheer amount of data tires people and since the vast majority of scans are non-malignant, the radiologists sometimes miss things that are right in front of their eyes. E.g. they look at typical positions of tumors instead of completely analyzing each and every scan the same way. As the day progresses they tend to focus more and more on common positions of tumors because they get more and more tired. The algorithm basically just prevents that.

    Though they trained their algorithm to identify normal samples as well as malignant. The thought process was that the most important thing to train an algorithm with is data and there is (luckily) vastly more data of benign scans than malignant ones, so they used that to train their algorithm.

    Reply
  • October 14, 2019 at 4:20 pm
    Permalink

    2:55 – "We like to imagine ourselves (…) as being uniquely capable of a whole range of different things (…). And I think that this example really cleraly demonstrates that's often not the case. And if you can train birds to diagnose cancer (…)"
    Well, actually you can't. Sorry to burst your bubble, but ALL you can train these birds to do is to find subtle variations and differences in colours and patterns on a presented picture – and then to communicate their "decision" it in a rather crude, "zero-one" way.
    Birds are evolutionary predisposed to be a perfect "spotters" – a vulture can spot a single dead animal while cruising at altitude of 4000-5000 meters / 13000-16000 ft – but that doesn't make them perfect pathologists or criminal investigators, does it?

    Those poor (and rather silly) pigeons after all that "training and coaching" they've received still don't know a sh!… erm, "a thing" about cancer, and they can't "diagnose" anything:
    • diagnose
    verb [ with obj. ] – identify the nature of (an illness or other problem) by examination of the symptoms: two doctors failed to diagnose a punctured lung; identify the nature of the medical condition of: he was finally diagnosed as having epilepsy | 20,000 men are diagnosed with skin cancer every year.

    Likewise, a trained dog can smell drugs or explosives, but that doesn't mean dogs know anything about drugs or explosives per se, least about "drugs smuggling and use" or "IEDs" or "terrorism" – all they are capable of is to "spot" certain smell and communicate their finds.

    Unlike humans, who are indeed capable of a "whole range of different things", even if we might not be as perfect as dogs or pigeons at CERTAIN and rather simple tasks (conceptually simple, that is).
    So, Hannah – to sum it up: one should pay attention to words or phrases one uses, and be critical in his or hers choices – "critical" as in greek "kritikos" – i.e. "involving the objective analysis and evaluation of an issue in order to form a judgement".
    "If you don't think about what you're saying, you will never really say what you're thinking", or something like that : )

    Reply
  • October 14, 2019 at 11:09 pm
    Permalink

    The really good pathologists are concealing a bunch of pigeons in their craniums.

    Reply
  • October 14, 2019 at 11:32 pm
    Permalink

    Hannah Fry is concealing magical powers.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *