The difference between brains and computers

I believe that within my lifetime processors will get powerful enough and software good enough that we will see computers that emulate human brains and pass the Turing Test (for more detail see my earlier posts Kurzweil predicts we will reverse engineer the human brain using software, Kurzweil predicts computers with the power of the human brain by 2025, Scientists create artificial brain with 2.3m simulated neurons).

Today I read a good Economist article which explains the challenges of building ‘human computers’. Brains currently have three key characteristics that computers do not:

These are: low power consumption (human brains use about 20 watts, whereas the supercomputers currently used to try to simulate them need megawatts); fault tolerance (losing just one transistor can wreck a microprocessor, but brains lose neurons all the time); and a lack of need to be programmed (brains learn and change spontaneously as they interact with the world, instead of following the fixed paths and branches of a predetermined algorithm).

Having identified these characteristics scientists are now working to design around them, as detailed in the rest of the Economist article. In summary progress is being made but it is very early days. The third characteristic of not needing to be programmed runs contrary to our current notions of development, and is perhaps the most challenging.

The other issue at play here is consciousness. ‘Would human computers be conscious or not?’ and ‘What is consciousness anyway?’ are unresolved questions with inherently unknowable answers. My view is that consciousness arises from the mind and human computers would be to all intents and purposes be conscious. Any other answer creates more questions than it answers and hence falls foul of Occam’s razor. This question was debated earlier today here and on Hacker News.


  • Richard Kelly

    A project similar to this is Honda’s project/friend – ASIMO. The concept behind his abilities is tuition. He is capable of learning objects, both static and in motion, and then store this data – building up a memory bank, just like Humans do.

    However to be fully-humanized the ‘Human Computer’ will need to learn to lie correctly. It’s simple to build a program to lie because you tell it what to lie about. And this precisely is the problem – it needs to decide what to lie about and what to use as a lie on its own accord. We all lie at some point in our lives and the reason is probably to benefit yourself or others’ feelings. In order for a Human Computer to lie it first needs to learn passionate emotion. All of our thoughts are processed via this kind of emotion: When we feel angry we process a more suited emotional response, such as lots of swearing and expression upon key words. When we’re happy we are more polite and positive with our vocabulary.

    To build a fully-humanized Human Computer you need to show it objects/media and learn it reasons to hate those objects/media. We all have a song we don’t like and there’ll be a reason behind that. No one taught us that reason because we decided for ourselves. We need to use that same process for Human Computers because emotion is the drive behind our thought process. We can store our lives and thoughts on a hard drive but the hard drive can not think like us. It needs a process and that process is emotion.

    As for the consciousness debate: Nothing is conscious unless its heart is healthily beating.

  • Randomised opinions about songs that generate simulated feelings and emotions could be programmed.

  • Richard Kelly

    If you can find a copy of July 2013 issue of Wired UK magazine there’s a feature about Neuroscientist Henry Markram, who has 1 Billion Euros to build a supercomputer replica of a Human brain. I’ve found a link to the web version of the article if you fancy a read:

  • Richard Kelly

    That may be possible but that’s merely telling the computer how to feel about a song. The emotion is programmed not spontaneously thought of. In order for a computer to give an opinion it needs to be told its choices and it may use a simple algorithm to choose one. Very similar to your iPod’s Shuffle feature. We could even expand on this ‘Shuffle’ process and use data to determine which opinion the computer chooses. But that’s not how we think as Humans. I love rock music such as AC/DC but I also like some classic piano such as Ludovico Einaudi but I only favor some of his work. I used to like his song ‘Primavera’ but I’ve grown tired of it. The computer needs to have this oddness to become tired of a song. Back in the 1990’s rave music dominated most radios but as we’ve grown, rave music sounds absolutely awful!! If a computer were to learn that rave music was brilliant back in the 1990’s it would still think rave music is brilliant today because it doesn’t know anything of tiredness or even changing taste. And I think that’s because our minds are molded by our surrounding environment. It’s why we have different accents, prefer certain types of scenery and dress a certain way. Even your health can define you to a certain environment (Southern UK residents live longer than Northern UK residents).

    It’s like teaching a child in school: We don’t tell them WHAT to think but HOW to think. The Human Computer needs to adapt the correct hardware/software so it is capable of learning how we learn. It also needs to have a Human environment: If you put a baby in the wild and left it to be brought up in the wilderness it would grow up to be like the type of animal it was raised by because it only knows from what it has learnt from its surroundings. It would know nothing of Human speech and it would most likely walk and move like an animal.

    ASIMO is the closest Human Computer to being Human and that’s because he has Human features – correct layout of the Human body (minus the battery backpack design feature). All it/he needs is access to the right learning tools and the right environment to start thinking like a Human. We merely mimic each other because we are influenced by one another. We’re a learning species.

    So I agree Nic, I do also believe within your lifetime (not assuming your old as grass) you will see a Human Computer pass the Turing test with flying colours. With innovations happening in this specific field, this very day, that hypothesis is more certain than ever 🙂