New Ultrafast Artificial Intelligence Mimicking Brain Dynamics

Using advanced experiments on neuronal cultures and large scale simulations, a group of scientists at Bar-Ilan University in Israel has demonstrated a new type of ultrafast artifical intelligence algorithms — based on the very slow brain dynamics — which outperform learning rates achieved to date by state-of-the-art learning algorithms.

The researchers rebuild the bridge between neuroscience and advanced artificial intelligence algorithms that has been left virtually useless for almost 70 years.

Compared with the existing artificial intelligence algorithms, asynchronous input based biological learning schemes can improve the scaling of learning rates of feedforward networks.

Results suggest that faster learning rates can be achieved with dendritic adaptation in comparison to a traditional synchronous perceptron. In addition, an extension of the fast learning rates to multi-layer networks is also discussed. The paper concludes with guidelines for fundamental questions in the future regarding the development of advanced classes of deep learning algorithms.

Scientific Reports – Biological learning curves outperform existing ones in artificial intelligence algorithms

The simulation results of biological learning algorithms show state-of-the-art optimal learning curves in supervised learning of feedforward networks. The biological learning algorithms comprise asynchronous input signals with decaying input summation, weights adaptation, and multiple outputs for an input signal. In particular, the generalization error for such biological perceptrons decreases rapidly with increasing number of examples, and it is independent of the size of the input. This is achieved using either synaptic learning, or solely through dendritic adaptation with a mechanism of swinging between reflecting boundaries, without learning steps. The proposed biological learning algorithms outperform the optimal scaling of the learning curve in a traditional perceptron. It also results in a considerable robustness to disparity between weights of two networks with very similar outputs in biological supervised learning scenarios. The simulation results indicate the potency of neurobiological mechanisms and open opportunities for developing a superior class of deep learning algorithms.

5 thoughts on “New Ultrafast Artificial Intelligence Mimicking Brain Dynamics”

  1. Incredible. I predicted that 2019 would see researchers build biologically similar artificial neural nets. This has taken us into a new age of AI with ANNs that operate fundamentally like Mamilian brains – by using time scale on inputs instead of feeding through all the information at once like they do at deep mind which is incredibly primitive , Convolutional Neural Nets are 1980s technology and will not advance any further. We actually have the foundations to build a Human Level AGI now. We will have a super intelligence this decade for certain now. This is probably the most significant scientific discovery in human history lol but no one is even paying attention.

    Reply
  2. Fortunately, real success isn’t about what you know, it’s about who you know.

    More seriously, the first country that can build something like this that can grok the world market is going to rule it all (and probably without anyone firing a shot), not least of which because it would be able to figure out where the serious competing efforts were being made and frustrate (or expropriate) every one of them.

    Yes, this is the real arms race nowadays, with everyone thinking, “How terrible if my own country, for all its faults, is not the first.”

    Even if a superior intelligence operated much slower than our human wetware, say experiencing only an hour for each week of ours, it would still have incredible impact as, depending on how much superior it was (and it wouldn’t have to be by much), it could likely get far more done in that hour than most ordinary people do in a lifetime.

    And one of it’s early efforts would likely be closing that time gap.

    Reply
  3. > We just haven’t understood what is really going on in the biological system well enough to emulate them.

    From this article it sounds like “until now” would fit well at the end there. Or perhaps we have understood it partially, but haven’t figured out how to emulate it (until now). Made some progress with this work, but probably still have room to improve on both fronts.

    Reply
  4. Not clock cycles. Upwelling synchrony in medium to large scale local networks that can elicit linked responses in distant networks. Information inducing information on similar scales but with infinitely subtle nuance.

    Anything close to a meaningful description of human brain function is almost indistinguishable from psychobabble. Nothing like what we understand as computing science or “neural networks.”

    Reply
  5. Well, yes, it’s been obvious all along that, given the performance biology gets out of components with “clock cycles” in the tens of Hertz, utilizing the same scheme with semiconductors would produce a brain that would be insanely fast.

    We just haven’t understood what is really going on in the biological system well enough to emulate them.

    Reply

Leave a Comment