Sunday, 3 November 2019

Week 2 [4-10.11.2019] Man versus machine.

While searching for curiosities about face recognition, I found the following study.
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150036
Face recognition is used to confirm a person's identity. Research shows that artificial intelligence is better at this task than people. The English have found and selected people who do it much better than AI.
Please read the material, especially the tests and their results. And answer the questions ...

1) Do you still think that AI is much more effective than human?
2) Is it reasonable to invest in such people?
3) Maybe, however, invest in machines and further improve them?

22 comments:

  1. 1. Thank you for raising such an interesting topic. I admire these people from the super-recognisers because it's not an easy task to compare faces from poor quality photos. As for the question, I think that in some cases a person needs cooperation with artificial intelligence. Artificial intelligence can always get something wrong and a trained person should watch over it. In my opinion, the big advantage is that artificial intelligence doesn't get tired. He can do his work 24/7 without a holiday break, etc. In addition, well-trained artificial intelligence replaces several people. It also takes a long time to train special units and they are in one place, and we can use AI globally.

    2. I think so, because not every individual can afford to invest in artificial intelligence or they don't have enough work to use AI. In addition, for safety reasons it is always better to have trained persons available.

    3. I think that despite having trained units, AI should be invested and improved, but on a global scale so that anyone who needs has access to such a system. As I said, such machines work 24/7 which is a big plus.

    ReplyDelete
    Replies
    1. Thank you very much for an interesting contribution to the discussion. I'm glad that in general automation you didn't immediately define the involvement and experience of man. Sometimes it seems to me that we ignore what's quite important to us - humanity.
      I also think that the best combination is just AI + human. The effect, I suppose 100% up.
      Thank you.

      Delete
  2. 1. Вы все еще думаете, что ИИ гораздо эффективнее человека?
    Может быть, не сейчас, но я думаю, что в будущем это будет определенно более эффективно. Я думаю, что искусственный интеллект хорошо справляется со своей задачей, но это не всегда правильно, и чтобы его улучшить, нужно перепроверить его. И после обнаружения ошибки улучшите алгоритм. Но я определенно уверен, что в ближайшем будущем искусственный интеллект будет лучше справляться с этой задачей, чем мы.

    2. Разумно ли инвестировать в таких людей?
    На данный момент определенно да, поскольку искусственный интеллект еще не достиг совершенства, и эти люди могли бы помочь определить, где он допустил ошибку и почему.

    3.Может, однако, инвестировать в машины и улучшать их?
    Я думаю, что да, вам нужно инвестировать в искусственный интеллект и улучшать его, но вам также нужно вкладывать деньги в людей, чтобы улучшить алгоритм распознавания (в ответе № 2)

    ReplyDelete
    Replies
    1. Hello;
      Thank you very much for an interesting answer, especially in a language that I haven't used for a long time. It was nice to remember old times. Yes, I also think that AI is not yet developed as much as we would like and human supervision is advisable. Just like Monika wrote.

      привет всем

      Delete
  3. 1) Do you still think that AI is much more effective than human?

    I think that Artificial Inteligence is becoming the essential part of our time. People try to use AI and limit the human work. “Artificial Inteligence” does not need to sleep or eat in contrast to a human being. For example, AI can recognize faces constantly without any breaks so it could be more effective than human. Recently, I have read that article that the AI can translate better from one to another language than people. In my opinion it could be possible that the most of current activities will be done by AI and people will have other professions which are not known now. Nowadays, AI is good addition to human work and combination of the human work and the AI can be more effective than any of them separately.

    2) Is it reasonable to invest in such people?

    As I wrote above, I think that nowadays the best option is to combine the human work and AI. It could be the most effective because AI is still developing technology and there could be errors and the people can oversee the process for example of recognition faces. Additionally, technology based on AI are very expensive and not every company can afford it. So for now, I think that it is reasonable to invest in such people who were described in the article.

    3) Maybe, however, invest in machines and further improve them?

    I think that investing in machines is justified because AI becomes more and more popular and it can be applied almost everywhere. When I answered your question number 1, I wrote that AI can replaced the human work but in the future, not now. Improving the AI gives big opportunities for the next generation who will use AI in common life.

    ReplyDelete
    Replies
    1. I'm glad that we perceive AI as well as human work. I agree that nowadays these two worlds should be connected and developed each time but separately.
      When choosing this article for discussion I thought that the human factor would be strongly criticized in favor of AI. However, perhaps something is human for everyone in us.

      Delete
  4. 1) Do you still think that AI is much more effective than human?
    1. It is obvious that AI recognizes the face better. He can analyze more rhymes of the face. Than a human being does, but it took a lot of data to do it. People themselves have taught artificial intelligence to recognize faces. All these snapshots, putting on the dog's face and other filters, and at the same time allowing the programs to scan the face and teach them to recognize shapes. Normally, the introduction of such a number of "trial consumers" would cost corporations billions, and in the case of applications they still had a profit from advertising, etc., and so on. And now, trained programs, with the appropriate updates and extensions, will be used by governments and the military.
    2) Is it reasonable to invest in such people?
    There used to be people who had a talent for recognizing people. Such people worked for the military. Especially with photographic memory. It only took a few hours to recognize such a person and now, thanks to the use of coevolutionary neural networks, it takes milliseconds. So are fingerprints.

    3) Maybe, however, invest in machines and further improve them?
    Neural networks for image recognition are not a black box. With a sufficiently large data set, specialised cells are formed on the lower layers to detect certain image characteristics. Most of the currently used algorithms use pre-trained networks, so that their creators do not have to spend a lot of money on a complete training that someone else has done. The final application of such a network can be completely independent of the application of the network from which the pre-trained layers come, as long as they operate on similar input data sets. Neural networks learn all the time, e.g. the algorithm for the imposition of dogs detects facial expressions can be helpful in detecting emotions, without major modifications, which is extremely valuable information for the services. At the same time, training data can be collected all the time, together with the context of the conversation, which can be used for other applications.

    ReplyDelete
    Replies
    1. So, however, the machine. Your statement clearly shows that the machine is better. Because cheaper, faster and generally ... better. But nevertheless from previous statements it appears that man + machine.
      So I think, man for now!

      Delete
  5. 1) Do you still think that AI is much more effective than human?

    In terms of the amount of data processed, artificial intelligence is second to none. It's all about the accuracy of this checking, it all depends on how well we teach it. However, looking at how such systems are used in China, I wonder if this is going in the right direction.

    2) Is it reasonable to invest in such people?

    I think it's worth getting interested in such people. There are probably not many of them, but you can use their skills to better train artificial neural networks. I agree with Monika that the strength lies in the cooperation between man and artificial intelligence.

    3) Maybe, however, invest in machines and further improve them?

    Looking at how many fields of artificial intelligence are used, I think that we will continue to develop algorithms and improve the results. The scale of investment is already very large and I think it will only grow.

    ReplyDelete
    Replies
    1. Hmm ... another voice for human and machine cooperation. So we are thinking individuals.
      people advantage + machine help! cool

      Delete
  6. 1) Do you still think that AI is much more effective than human?

    In my opinion, the effectiveness of artificial intelligence depends largely on properly prepared algorithms and testing them. When it comes to face recognition, it is irreplaceable when it comes to working time, because it never gets tired, it can work without a break and does not demand holidays. However, no artificial intelligence can ever replace a person's work, experience or commitment to work.

    2) Is it reasonable to invest in such people?

    I believe that investment in human development is the best investment that can be. Currently, recognition systems are very poor and will not always work as expected. Each system should be personalized, while a person learns very quickly and can adapt to a changing situation.

    3) Maybe, however, invest in machines and further improve them?

    If we can afford it, we should invest in such equipment to improve it. It can operate regardless of geographical location. He can improve his results so that in time he becomes good enough to use the system as a system supporting people's work.

    ReplyDelete
    Replies
    1. Good voice 100% human. The machine completely by the way. I like the direction of this discussion.
      Thank you very much for voting for people!

      Delete
  7. 1. Do you still think that AI is much more effective than human?
    For sure I never told that AI is better than human. In some case maybe yes but not in all. AI is trained on data to find differences, classify objects to right group or forecasting. Some people that are familiar with some task and have necessary abilities could work better than AI. This article is a proof. Also its depends on task that AI nad human have to do. I think that in some task people are better and in some AI is the winner.

    2. Is it reasonable to invest in such people?
    I think yes. People with important abilities are worth to investing in them. AI is not perfect but it can be helpful for people. We can used knowledge of talented people to improve AI.

    3. Maybe, however, invest in machines and further improve them?
    I think we should improve performance of AI and find new solutions. AI is a great tool for many tasks and it's worth to investing in them. Today we have many algorithms that are very helpful in many tasks.

    ReplyDelete
    Replies
    1. In the strict sense, I also think that AI needs to be developed. Only, watching the Terminator a long time ago, I was wondering when the machines would beat us.
      Do I live to see it, I hope not!

      Delete
  8. AI is definitely more effective in terms of speed. If there would be only a few faces to recognize and the accuracy of the results was extremely important then it would make sense to hire "the police super recognizers".
    But if the task is to unlock smartphones of millions of users with a so-so accurate face recognition than AI is the winner.

    ReplyDelete
    Replies
    1. And if your smartphone unlocks by accident by someone else because AI will not work. And then your "sensitive data" will see the whole world - even your friends ... What will you do then? You write to the US, Korea, Japan for compensation, or put your "ears down" and get over it, explaining the algorithm made a mistake ...

      Delete
  9. 1. In my opinion AI will always be more effective then human, because it is especially adjusted to specific assignments related to data. In case of people who are assigned to recognize faces, we must take into consideration humant factor, which means that the effect of their work can be contaminated by for example health problems, lack of focus and energy at some point of the day, some attractive features in the face of another person, that the worker will be focused on and – because of that – this worker may not recognize other important features and therefore – his work will be worse then effects reached by AI.

    2. I think that the only platform on which in my opinion it would be reasonable to invest in people mentioned in the article is working in the field, where you have no access to cameras or Internet, beacuse it is obvious that in such situation AI can’t do no good. So for sure at this point we should use actual people and it is worth investing in them to do their jobs as best as possible.

    3. For sure it is reasonable to invest in machines and further improvement of AI, but it is nothing new in our times. Systems designed to recognize faces are commonly used for example in China – I once read an article about this and there was mentioned that in this country the only way to log on the Internet is by the face recognition, and when someone commit a crime using the Internet, f.e. wrote unpolite comments about authorities, then the program based on face recognition will deny him or her access to some public infrastructure, like transportation. So in my opinion nowadays investments in machines are on the pretty advanced level. Of course there is a case of ethical correctness of such solutions.

    ReplyDelete
  10. 1)Do you still think that AI is much more effective than human?

    I wasn’t aware of existence of super-recognisers, but even though those people outperform the average results of other people face recognition skill it doesn’t mean they are better than AI. I think that artificial intelligence algorithms will be always better than humans, because of some advantages that people cannot overcome. In very short time AI can be trained on new / another training set focused on some new feature. In case of computing AI power can be easily enlarged through switching to more powerful cluster. AI didn’t get tired or distracted. Considering these facts, I stand that AI is more efficient than human.

    2) Is it reasonable to invest in such people?

    In my opinion society will gain more if a skill of such perceptive individuals will be used in other ways, in a place where the AI is not already implemented. Wasting talent of those people for some thing that can be done even better by algorithm is unreasonable.

    3) Maybe, however, invest in machines and further improve them?

    In my opinion it is the investing in people who are creating and improving the algorithms in first place and then in machines – in improving computing power, and yes I think this is the good direction for the future.

    ReplyDelete
  11. About a year ago there was an attack in the Amsterdams' Central station. A 19-year-old Afghan who stabbed two American tourists in Amsterdam was shot just nine seconds after he launched his attack. It was possible not only with face recognition methods but also with "predictive profiling" whereby a suspect can be identified through suspicious patterns of behavior, by a person called "spotter." In my opinion, it is worth to invest in such person, and it would be an excellent ideat to arm them with additional support from the AI side.

    ReplyDelete
  12. 1) Do you still think that AI is much more effective than human?

    I don't think that AI is more effective than human. We could develop the AI solutions, train neural networks to perform tasks with high accuracy, but even if the effects achieved by artificial intelligence are better than the ones achieved by humans, it should be taken into considerartion that AI is specialized usually in limited number (or even only one) tasks, but human ability allows to do multiple tasks. AI, however more and more impressive, is still a set of algorithms designed by humans and (I hope) won't be a real competition for people for a long time.

    2) Is it reasonable to invest in such people?

    In my opinion it is always reasonable to invest in people. Besides that AI requires 'special conditions' to work, it can only operate on digital data. Human could move more freely and does not need to be connected to electricity ;) Moreover the professional knowledge and experience of super-recognisers could be used to improve algorithm's performance.

    3) Maybe, however, invest in machines and further improve them?

    I also believe that investing in machines is not a bad idea. The number of people with special recognition ability is limited. Ordinary person achieves much lower results in face recognition and in this case the support of an algorithm might be valuable.

    ReplyDelete
  13. 1) Do you still think that AI is much more effective than human?
    I'm not surprised there are people who perform much better in face recognition tasks than general population or even best AI solutions of today. But, in the long run, facial recognition is quite simple task, suited well for boring AI algorithms. And we can toss as much computing power to the task, as it is necessary.
    2) Is it reasonable to invest in such people?
    I don't think it makes sense. Those are oddities that doesn't change anything. We could possibly learn something new about how people recognize faces from studying such persons, but nothing more.
    3) Maybe, however, invest in machines and further improve them?
    Yes, definitively. Machines can be replicated at will, each of replicas having the same recognition skills as the original. In case of people super-recognisers, we'd have to sieve out those from population. Hmm... maybe some free-to-play game on mobile platforms? That would allow to scan thru large group of people for almost free.

    ReplyDelete
  14. 1) Do you still think that AI is much more effective than human?

    No, I don't think so, but from other hand the machine is never exhausted! Some people could be much more better but for how long? How long they can focus on it? One hour? Five hours...?
    The point of replacement human by AI is that human are changed. The AI wouldn't have bad day-it's constantly.

    2) Is it reasonable to invest in such people?

    I think that no. Face recognition is easy case in which we could (and should) replace human by AI. I think that human should do much more harder case which are too hard for AI nowadays.

    3) Maybe, however, invest in machines and further improve them?

    Yes! As I said before-in easy (and boring) case we should replace human, because they wouldn't be focused on it the whole time. Moreover machines could checking many cameras at once... If we talk about face recognition the human have much lower chance if he have to watch 20 cameras at once. For AI is peace of cake!

    ReplyDelete