TED Conversations

Josh Mayourian

Student , Cooper Union for the Advancement of Science and Art

TEDCRED 50+

This conversation is closed. Start a new conversation
or join one »

Will we ever truly be able to model nature?

My Bioelectricity professor Nina Tandon recently gave a TED talk “Caring for engineered tissue” and I was amazed how we are able to copy the environment of artificially grown cells. There are many techniques used to reduce error and create accurate results. Such amazing replications allow us to grow artificial hearts and bones, enhancing research opportunities on these
parts of the body. This made me wonder how successful we are at modeling
other living systems, so I watched the TED talk “Robert Full on engineering and evolution.” Many years ago, engineer's claimed bees shouldn't be able to fly, dolphins shouldn’t be able to swim, and geckos shouldn't be able to climb from their calculations. However, in the past few years we've been able to explain these phenomenons, showing how much we have progressed. Through watching these great talks, I was curious: How close are we to modeling nature and making predictions without ideal assumptions? Will we ever be able to reach this point and truly copy nature?

Related Talks:
Topics: Bioelectricity TEDinClass
+6
Share:

Showing single comment thread. View the full conversation.

  • thumb

    Thomas Brucia

    • +4
    Mar 1 2012: One issue is modeling... another is the degree to which the models are accurate... and a third is how far out they remain accurate to some agreed-upon degree -- and when the Butterfly Effect makes them useless. The problem in modeling is feedback loops which amplify small divergences into enormous ones. Sometimes these are major factors (as in weather forecasting) and sometimes there are no divergences (as in modeling the acceleration of a falling apple). The real issue is knowing the characteristics of the aspect of our natural world we're trying to model...... We might reflect that all models are *approximations* and *simplifications* of reality and therefore there's a difference between the more comprehensible model and the incomprehensibility of the more variegated reality being modeled. The only exact model of the universe would be an exactly identical universe... and what use would that be, even if it were possible.....

    • Michael M 30+

      • 0
      Mar 1 2012: Thomas
      Thank you for this comment and you actually wrote what I was going to say. Lorenz was correct, simple dependence on initial conditions is always how the model turns out something "similar." But is finally similar not exact. I laugh sometimes at the "5-day forecast". Ok sometimes they are right, but sometimes, you just want to say turn off the computer and look out the window. A great answer to this very good question.

      Josh, I think we approximate well a number of things. non-linear dynamics says I think, good job, but you forgot this little tweak.
      • thumb

        Josh Mayourian 50+

        • +1
        Mar 1 2012: Thomas and Michael,

        Using Runge-Kutta methods when programming to approximate, we are able to estimate very well, but only in an ideal setting. Nature is far from an ideal setting. Therefore, I totally agree with you, as nature is so complex that changes would occur in a way that no approximation method used will be able to model perfectly.
    • thumb

      Sophie Rand 50+

      • +3
      Mar 2 2012: I agree that we could get close to accuracy, but what would be the point? Getting near accuracy is good enough in most cases-- in engineering we make assumptions so often-- linearity, frictionless, causal...all these things which help us vastly to come up with solutions to natural problems, without needing to truly imitate those natural phenomena.
      • thumb

        Matthew Wieder 50+

        • 0
        Mar 7 2012: There seem to be two directions that this question is being attacked. One from the point of view of whether or not we can create a model for how nature works, how molecules interact to produce the visible results we see in our daily lives. The second is to look at specific instances of how we can use materials to mimic the properties of nature. I will attempt to answer the latter question. In my tissue engineering class as well as my biomechanics class we have been shown over and over again, whether it be a replacement for a material replacement for a tendon or an artificial knee replacement, these artificial replacements cannot match up to those of the native tissues they are replacing. For example, while we might be able to find a material that is capable of reproducing a single material property of the native tendon for example elasticity, the replacements will always fail with respect to a different material property for example strength. To this end, tissue engineering takes as an assumption that in order to truly model a native tissue from the body or in nature in general, the material must be made up for the most part of the same constituent parts as the native one. The question then becomes not finding a material that matches the properties, but rather a set of environmental conditions which will allow a living tissue to grow and gain the same material properties as the native tissue.

        • Mary M. 50+

          • 0
          Mar 7 2012: I replied to your comment on being human (it did not have a reply button and your email isn't accessible).....I will come back to erase this one in a few minutes.

          Mary

    • Veronica Shalotenko 50+

      • +3
      Mar 6 2012: Hello Thomas,

      I agree that every model is inherently a simplification of reality. Certainly, this is acceptable for most applications. Going back to the weather example, it is usually not completely devastating if the meteorologist tells us that it’s going to rain in the morning but it actually rains in the evening. But, what about tissue engineering applications, to which Josh referred in his original question? Is an approximate model of human tissue sufficient in the in vitro cultivation of tissues and organs for implantation? Of course, due to various technological limitations, tissue engineers will probably never be able to copy nature exactly. Still, it seems that an “almost perfect” imitation of nature is required in tissue engineering. To maximize the chances that a laboratory-grown transplant will successfully integrate into a human body, the transplant should be an exact match to the tissue or organ that it was grown to mimic.
      • thumb

        Andrew Leader 50+

        • +5
        Mar 6 2012: In my view, the best way to be successful at modeling is first to accept that you can't make a model with infinite accuracy and precision--That's why it's called a model. Next, you have to carefully decide which assumptions you're willing to make, and you have to justify them. Ideally, you should be able to quantify the error: "You should get the right answer within 15%", for example, so that the results are unequivocal, not to be misconstrued as an exact representation, but also not to be disregarded as a mere guess. Of course, it so many disciplines such as biology and economics, the systems are so complex that the error propagation quickly blows up.

        My question regards the areas in which we currently have good models, and the areas in which we can use models to make the jump from science and discovery to engineering and design. In medicine, for example, researchers have been using animal models for years and years. However at this point, and correct me if I'm wrong, few computational models are at the point where we where we would base the treatment of a patient on model results.

        In Electrical Engineering, however, we have fantastic modeling software that will model many, many types of circuits with very good accuracy. In my engineering design classes, I've used models to get a good idea of what a good design might be, and then gone through iterations of trial and error in the lab before making the final product. But some VLSI circuits are not so easy to test in the lab, and designers heavily rely on simulations.

        When, if ever, will we make the leap from pure discovery to design and implementation using computational modeling of complex systems like weather, medicine, ecology, economics....
      • thumb

        Maria Georgescu 50+

        • 0
        Mar 6 2012: As Veronica said, I think that the level of accuracy that we need for a model truly depends on the situation or system we are trying to understand. Sometimes a model does not even need to be accurate to teach an idea. Although the lock and key model of an enzyme reaction has long been disproved, it is still the way that the concept is initially introduced to students because of the relationship and parallel they can draw to everyday life. Once that idea is understood, instructors proceed to inform their students that the model approximates more of an induced fit process rather than a perfect one to one connection.

Showing single comment thread. View the full conversation.