Abstract
Artificial Intelligence (AI) has recently become a real formal science: the new millennium brought the first mathematically sound, asymptotically optimal, universal problem solvers, providing a new, rigorous foundation for the previously largely heuristic field of General AI and embedded agents. There also has been rapid progress in not quite universal but still rather general and practical artificial recurrent neural networks for learning sequence-processing programs, now yielding state-of-the-art results in real world applications. And the computing power per Euro is still growing by a factor of 100–1,000 per decade, greatly increasing the feasibility of neural networks in general, which have started to yield human-competitive results in challenging pattern recognition competitions. Finally, a recent formal theory of fun and creativity identifies basic principles of curious and creative machines, laying foundations for artificial scientists and artists. Here I will briefly review some of the new results of my lab at IDSIA, and speculate about future developments, pointing out that the time intervals between the most notable events in over 40,000 years or \(2^9\) lifetimes of human history have sped up exponentially, apparently converging to zero within the next few decades. Or is this impression just a by-product of the way humans allocate memory space to past events?
Note: this is the 2012 update of a 2007 publication (Schmidhuber 2007b). Compare also the 2006 celebration of 75 years of AI (Schmidhuber 2006c).
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Balcan, M. F., Beygelzimer, A., & Langford, J. (2009). Agnostic active learning. Journal of Computer and System Sciences, 75(1), 78–89.
Barto, A. (2013). Intrinsic motivation and reinforcement learning. In G. Baldassarre & M. Mirolli (Eds.), Intrinsically motivated learning in natural and artificial systems. Springer (in press).
Behnke, S. (2003). Hierarchical neural networks for image interpretation, volume 2766 of lecture notes in computer science. Springer.
Bishop, C. M. (2006). Pattern recognition and machine learning. NY: Springer.
Bringsjord, S. (2000), ‘A contrarian future for minds and machines’, chronicle of higher education (p. B5). Reprinted in The Education Di-gest, vol. 66(6), pp. 31–33.
Ciresan, D. C., Meier, U., Gambardella, L. M., & Schmidhuber, J. (2010). Deep big simple neural nets for handwritten digit recogntion. Neural Computation, 22(12), 3207–3220.
Ciresan, D. C., Meier, U., Gambardella, L. M., & Schmidhuber, J. (2011a). Convolutional neural network committees for handwritten character classification. In 11th International Conference on Document Analysis and Recognition (ICDAR), pp 1250–1254.
Ciresan, D. C., Meier, U., Masci, J., Gambardella, L. M. & Schmidhuber, J. (2011b). Flexible, high performance convolutional neural networks for image classification. In International Joint Conference on Artificial Intelligence IJCAI, pp 1237–1242.
Ciresan, D. C., Meier, U., Masci, J., & Schmidhuber, J. (2011c). A committee of neural networks for traffic sign classification. In International Joint Conference on, Neural Networks, pp 1918–1921.
Ciresan, D. C., Meier, U., Masci, J., & Schmidhuber, J. (2012a). Multi-column deep neural network for traffic sign classification. Neural Networks, 32, 333–338.
Ciresan, D. C., Meier, U., & Schmidhuber, J. (2012b). Multi-column deep neural networks for image classification. In IEEE Conference on Computer Vision and Pattern Recognition CVPR 2012, pp 3642–3649.
Ciresan, D. C., Meier, U., & Schmidhuber, J. (2012c). Multi-column deep neural networks for image classification. In IEEE Conference on Computer Vision and Pattern Recognition CVPR 2012. Long preprint arXiv:1202.2745v1 [cs.CV].
Darwin, C. (1997). The descent of man, prometheus, amherst. NY: A reprint edition.
Dayan, P. (2013). Exploration from generalization mediated by multiple controllers. In G. Baldassarre & M. Mirolli (Eds.), Intrinsically motivated learning in natural and artificial systems. Springer (in press).
Fedorov, V. V. (1972). Theory of optimal experiments. NY: Academic.
Fernandez, S., Graves, A., & Schmidhuber, J. (2007). Sequence labelling in structured domains with hierarchical recurrent neural networks. In Proceedings of the 20th International Joint Conference on Artificial Intelligence (IJCAI).
Floridi, L. (2007). A look into the future impact of ICT on our lives. The Information Society, 23(1), 59–64.
Fukushima, K. (1980). Neocognitron: A self-organizing neural network for a mechanism of pattern recognition unaffected by shift in position. Biological Cybernetics36(4), 193–202.
Gers, F. A., & Schmidhuber, J. (2001). LSTM recurrent networks learn simple context free and context sensitive languages. IEEE Transactions on Neural Networks, 12(6), 1333–1340.
Gers, F. A., Schraudolph, N., & Schmidhuber, J. (2002). Learning precise timing with LSTM recurrent networks. Journal of Machine Learning Research, 3, 115–143.
Gisslen, L., Luciw, M., Graziano, V., & Schmidhuber, J. (2011). Sequential constant size compressor for reinforcement learning. In Proceedings of Fourth Conference on Artificial General Intelligence (AGI), Google, Mountain View, CA.
Glasmachers, T., Schaul, T., Sun, Y., Wierstra, D. & Schmidhuber, J. (2010). Exponential Natural Evolution Strategies. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO).
Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38, 173–198.
Gomez, F. J., Schmidhuber, J., & Miikkulainen, R. (2008). Efficient non-linear control through neuroevolution. Journal of Machine Learning Research JMLR, 9, 937–965.
Graves, A., Fernandez, S., Gomez, F. J., & Schmidhuber, J. (2006). Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural nets. In ICML ’06: Proceedings of the International Conference on Machine Learning.
Graves, A., Fernandez, S., Liwicki, M., Bunke, H., & Schmidhuber, J. (2008). Unconstrained on-line handwriting recognition with recurrent neural networks. In J. C. Platt, D. Koller, Y. Singer, & S. Roweis (Eds.), Advances in Neural Information Processing Systems 20 (pp. 577–584). Cambridge: MIT Press.
Graves, A., Liwicki, M., Fernandez, S., Bertolami, R., Bunke, H., & Schmidhuber, J. (2009). A novel connectionist system for improved unconstrained handwriting recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 31(5), 855–868.
Graves, A., & Schmidhuber, J. (2009). Offline handwriting recognition with multidimensional recurrent neural networks. In Advances in Neural Information Processing Systems (p. 21). Cambridge: MIT Press.
Hansen, N., & Ostermeier, A. (2001). Completely derandomized self-adaptation in evolution strategies. Evolutionary Computation, 9(2), 159–195.
Hart, S., Sen, S., & Grupen, R. (2008). Intrinsically motivated hierarchical manipulation. In Proceedings of the IEEE Conference on Robots and Automation (ICRA). California: Pasadena.
Hochreiter, S., Bengio, Y., Frasconi, P., & Schmidhuber, J. (2001). Gradient flow in recurrent nets: The difficulty of learning long-term dependencies. In S. C. Kremer & J. F. Kolen (Eds.), A Field Guide to Dynamical Recurrent Neural Networks. NJ: IEEE Press.
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780.
Holland, J. H. (1975). Adaptation in natural and artificial systems. Ann Arbor: University of Michigan Press.
Hutter, M. (2002). The fastest and shortest algorithm for all well-defined problems. International Journal of Foundations of Computer Science, 13(3), 431–443 (On J. Schmidhuber’s SNF grant 20–61847).
Hutter, M. (2005). Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability. Berlin: Springer (On J. Schmidhuber’s SNF grant 20–61847).
Jaeger, H. (2004). Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication. Science, 304, 78–80.
Kaelbling, L. P., Littman, M. L., & Moore, A. W. (1996). Reinforcement learning: A survey. Journal of AI research, 4, 237–285.
Kolmogorov, A. N. (1965). Three approaches to the quantitative definition of information. Problems of Information Transmission, 1, 1–11.
Koutnik, J., Gomez, F., & Schmidhuber, J. (2010). Evolving neural networks in compressed weight space. In Proceedings of the Conference on Genetic and, Evolutionary Computation (GECCO-10).
Krizhevsky, A. (2009). Learning multiple layers of features from tiny images. Master’s thesis: Computer Science Department, University of Toronto.
Kuipers, B., Beeson, P., Modayil, J., & Provost, J. (2006). Bootstrap learning of foundational representations. Connection Science, 18(2).
Kurzweil, R. (2005). The singularity is near. NY: Wiley Interscience.
LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278–2324.
LeCun, Y., Huang, F.-J., & Bottou, L. (2004). Learning methods for generic object recognition with invariance to pose and lighting. In Proceedings of Computer Vision and Pattern Recognition Conference.
Lenat, D. B. (1983). Theory formation by heuristic search. Machine Learning, vol. 21.
Levin, L. A. (1973). Universal sequential search problems. Problems of Information Transmission, 9(3), 265–266.
Li, M., & Vitányi, P. M. B. (1997). An introduction to kolmogorov complexity and its applications (2nd ed.). NY: Springer.
Maass, W., Natschläger, T., & Markram, H. (2002). A fresh look at real-time computation in generic recurrent neural circuits. Institute for Theoretical Computer Science, TU Graz : Technical report.
Mitchell, T. (1997). Machine learning. NY: McGraw Hill.
Moravec, H. (1999). Robot . NY: Wiley Interscience.
Newell, A., & Simon, H. (1963). GPS, a program that simulates human thought. In E. Feigenbaum & J. Feldman (Eds.), Computers and thought (pp. 279–293). New York: McGraw-Hill.
Oudeyer, P. -Y., Baranes, A., & Kaplan, F. (2013). Intrinsically motivated learning of real world sensorimotor skills with developmental constraints. In G. Baldassarre & M. Mirolli (Eds.), Intrinsically motivated learning in natural and artificial systems. Springer (in press).
Rechenberg, I. (1971). Evolutions strategie–optimierung technischer systeme nach Prinzipien der biologischen Evolution. Dissertation, Published 1973 by Fromman-Holzboog.
Robinson, A. J., & Fallside, F. (1987). The utility driven dynamic error propagation network. Technical Report CUED/F-INFENG/TR.1, Cambridge University Engineering Department.
Rosenbloom, P. S., Laird, J. E., & Newell, A. (1993). The SOAR papers. NY: MIT Press.
Schaul, T., Bayer, J., Wierstra, D., Sun, Y., Felder, M., Sehnke, F., et al. (2010). PyBrain. Journal of Machine Learning Research, 11, 743–746.
Scherer, D., Müller, A., & Behnke, S. (2010). In International Conference on Artificial Neural Networks.
Schmidhuber, J. (1990). Dynamische neuronale Netze und das fundamentale raumzeitliche Lernproblem. Dissertation: Institut für Informatik, Technische Universität München.
Schmidhuber, J. (1991a). Curious model-building control systems. In Proceedings of the International Joint Conference on Neural Networks (vol. 2, pp. 1458–1463). Singapore: IEEE press.
Schmidhuber, J. (1991b). A possibility for implementing curiosity and boredom in model-building neural controllers. In J. A. Meyer & S. W. Wilson (Eds.) Proceedings of the International Conference on Simulation of Adaptive Behavior: From Animals to Animats, pp. 222–227. MIT Press/Bradford Books.
Schmidhuber, J. (1991c). Reinforcement learning in Markovian and non-Markovian environments. In D. S. Lippman, J. E. Moody, & D. S. Touretzky (Eds.), Advances in neural information processing systems 3 (NIPS 3) (pp. 500–506). NY: Morgan Kaufmann.
Schmidhuber, J. (1992a). A fixed size storage \(O(n^3)\) time complexity learning algorithm for fully recurrent continually running networks. Neural Computation, 4(2), 243–248.
Schmidhuber, J. (1992b). Learning factorial codes by predictability minimization. Neural Computation, 4(6), 863–879.
Schmidhuber, J. (1997). Discovering neural nets with low Kolmogorov complexity and high generalization capability. Neural Networks, 10(5), 857–873.
Schmidhuber, J. (1999). Artificial curiosity based on discovering novel algorithmic predictability through coevolution. In P. Angeline, Z. Michalewicz, M. Schoenauer, X. Yao,& Z. Zalzala (Eds.), Congress on evolutionary computation (pp. 1612–1618). Piscataway: IEEE Press.
Schmidhuber, J. (2002a). Hierarchies of generalized Kolmogorov complexities and nonenumerable universal measures computable in the limit. International Journal of Foundations of Computer Science, 13(4), 587–612.
Schmidhuber, J. (2002). The speed prior: A new simplicity measure yielding near-optimal computable predictions. In J. Kivinen& R. H. Sloan (Eds.), Proceedings of the 15th Annual Conference on Computational Learning Theory (COLT 2002) (pp. 216–228). Lecture Notes in Artificial Intelligence Sydney, Australia: Springer.
Schmidhuber, J. (2003a). Exponential speed-up of computer history’s defining moments. http://www.idsia.ch/juergen/computerhistory.html
Schmidhuber, J. (2003b). The new AI: General & sound & relevant for physics. Technical Report TR IDSIA-04-03, Version 1.0, arXiv:cs.AI/0302012 v1.
Schmidhuber, J. (2004). Optimal ordered problem solver. Machine Learning, 54, 211–254.
Schmidhuber, J. (2005). Completely self-referential optimal reinforcement learners. In W. Duch, J. Kacprzyk, E. Oja, & S. Zadrozny (Eds.), Artificial neural networks: Biological inspirations–ICANN 2005 (pp. 223–233), LNCS 3697. Springer: Berlin Heidelberg (Plenary talk).
Schmidhuber, J. (2006a). Developmental robotics, optimal artificial curiosity, creativity, music, and the fine arts. Connection Science, 18(2), 173–187.
Schmidhuber, J. (2006b). Gödel machines: Fully self-referential optimal universal self-improvers. In B. Goertzel& C. Pennachin (Eds.), Artificial general intelligence (pp. 199–226). Heidelberg: Springer (Variant available as arXiv:cs.LO/0309048).
Schmidhuber, J. (2006c). Celebrating 75 years of AI–history and outlook: The next 25 years. In M. Lungarella, F. Iida, J. Bongard,& R. Pfeifer (Eds.), 50 years of artificial intelligence (vol. LNAI 4850, pp. 29–41). Berlin/Heidelberg: Springer (Preprint available as arXiv:0708.4311).
Schmidhuber, J. (2007a). Gödel machines: Fully self-referential optimal universal self-improvers. In B. Goertzel& C. Pennachin (Eds.), Artificial general intelligence (pp. 199–226). Springer Verlag (Variant available as arXiv:cs.LO/0309048).
Schmidhuber, J. (2007b). New millennium AI and the convergence of history. In W. Duch& J. Mandziuk (Eds.), Challenges to computational intelligence (vol. 63, pp. 15–36). Studies in Computational Intelligence, Springer, 2007. Also available as arXiv:cs.AI/0606081.
Schmidhuber, J. (2009). Ultimate cognition à la Gödel. Cognitive Computation, 1(2), 177–193.
Schmidhuber, J. (2010). Formal theory of creativity, fun, and intrinsic motivation (1990–2010). IEEE Transactions on Autonomous Mental Development, 2(3), 230–247.
Schmidhuber, J. (2011). PowerPlay: Training an increasingly general problem solver by continually searching for the simplest still unsolvable problem. Technical Report arXiv:1112.5309v1 [cs.AI].
Schmidhuber, J. (2012). Philosophers& futurists, catch up! response to the singularity. Journal of Consciousness Studies, 19(1–2), 173–182.
Schmidhuber, J., Ciresan, D., Meier, U., Masci, J., & Graves, A. (2011). On fast deep nets for AGI vision. In Proceedings of Fourth Conference on Artificial General Intelligence (AGI), Google, Mountain View, CA.
Schmidhuber, J., Eldracher, M., & Foltin, B. (1996). Semilinear predictability minimization produces well-known feature detectors. Neural Computation, 8(4), 773–786.
Schmidhuber, J., Wierstra, D., Gagliolo, M., & Gomez, F. J. (2007). Training recurrent networks by EVOLINO. Neural Computation, 19(3), 757–779.
Schmidhuber, J., Zhao, J., & Schraudolph, N. (1997). Reinforcement learning with self-modifying policies. In S. Thrun& L. Pratt (Eds.), Learning to learn (pp. 293–309). NY: Kluwer.
Schraudolph, N. N., Eldracher, M., & Schmidhuber, J. (1999). Processing images by semi-linear predictability minimization. Network: Computation in Neural Systems, 10(2), 133–169.
Schwefel, H. P. (1974). Numerische optimierung von computer-modellen. Dissertation, Published 1977 by Birkhäuser, Basel.
Siegelmann, H. T., & Sontag, E. D. (1991). Turing computability with neural nets. Applied Mathematics Letters, 4(6), 77–80.
Sims, K. (1994). Evolving virtual creatures. In A. Glassner (Ed.), Proceedings of SIGGRAPH ’94 (Orlando, Florida, July 1994), Computer Graphics Proceedings, Annual Conference (pp. 15–22). ACM SIGGRAPH, ACM Press. ISBN 0-89791-667-0.
Singh, S., Barto, A. G., & Chentanez, N. (2005). Intrinsically motivated reinforcement learning. In Advances in Neural Information Processing Systems 17 (NIPS). Cambridge: MIT Press.
Sloman, A. (2011a, Oct 23). Challenge for vision: Seeing a Toy Crane. Retrieved June 8, 2012, from http://www.cs.bham.ac.uk/research/projects/cosy/photos/crane/
Sloman, A. (2011b, June 8). Meta-morphogenesis and the creativity of evolution. Retrieved 6 June 2012, from http://www.cs.bham.ac.uk/research/projects/cogaff/evo-creativity.pdf
Sloman, A. (2011c, Oct 29). Meta-Morphogenesis and Toddler Theorems: Case Studies. Retrieved 8 June 2012, from http://www.cs.bham.ac.uk/research/projects/cogaff/misc/toddler-theorems.html
Sloman, A. (2011d, Sep 19). Simplicity and Ontologies: The trade-off between simplicity of theories and sophistication of ontologies. Retrieved June 8, 2012, from http://www.cs.bham.ac.uk/research/projects/cogaff/misc/simplicity-ontology.html
Smil, V. (1999). Detonator of the population explosion. Nature, 400, 415.
Solomonoff, R. J. (1964). A formal theory of inductive inference. Part I. Information and Control, 7, 1–22.
Stanley, K. O., & Miikkulainen, R. (2002). Evolving neural networks through augmenting topologies. Evolutionary Computation, 10, 99–127.
Storck, J., Hochreiter, S., & Schmidhuber, J. (1995). Reinforcement driven information acquisition in non-deterministic environments. In Proceedings of the International Conference on Artificial Neural Networks, Paris, vol. 2, pp. 159–164. EC2& Cie, 1995.
Strehl, A., Langford, J., & Kakade, S. (2010). Learning from logged implicit exploration data. Technical, Report arXiv:1003.0120.
Sun, Y., Wierstra, D., Schaul, T., & Schmidhuber, J. (2009a). Efficient natural evolution strategies. In Genetic and Evolutionary Computation Conference.
Sun, Y., Wierstra, D., Schaul, T., & Schmidhuber, J. (2009b). Stochastic search using the natural gradient. In International Conference on Machine Learning (ICML).
Sutskever, I., Martens, J., & Hinton, G. (2011). Generating text with recurrent neural networks. In L. Getoor& T. Scheffer (Eds.), Proceedings of the 28th International Conference on Machine Learning (ICML-11) (pp. 1017–1024). ICML ’11 New York, NY, USA: ACM.
Sutton, R., & Barto, A. (1998). Reinforcement learning: An introduction. Cambridge: MIT Press.
Turing, A. M. (1936). On computable numbers, with an application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, Series, 2(41), 230–267.
Utgoff, P. (1986). Shift of bias for inductive concept learning. In R. Michalski, J. Carbonell,& T. Mitchell (Eds.), Machine learning (Vol. 2, pp. 163–190). Los Altos, CA: Morgan Kaufmann.
Vapnik, V. (1995). The nature of statistical learning theory. New York: Springer.
Vinge, V. (1984). The peace war. Inc. : Bluejay Books.
Vinge, V. (1993). The coming technological singularity. VISION-21 Symposium sponsored by NASA Lewis Research Center, and Whole Earth Review, Winter issue.
Werbos, P. J. (1988). Generalization of backpropagation with application to a recurrent gas market model. Neural Networks, 1.
Wierstra, D., Foerster, A., Peters, J., & Schmidhuber, J. (2010). Recurrent policy gradients. Logic Journal of IGPL,18(2), 620–634.
Wierstra, D., Schaul, T., Peters, J., & Schmidhuber, J. (2008). Natural evolution strategies. In Congress of Evolutionary Computation (CEC 2008).
Williams R. J., & Zipser, D. (1994). Gradient-based learning algorithms for recurrent networks and their computational complexity. In back-propagation: Theory, architectures and applications. Hillsdale, NJ: Erlbaum.
Yao, X. (1993). A review of evolutionary artificial neural networks. International Journal of Intelligent Systems, 4, 203–222.
Yi, S., Gomez, F., & Schmidhuber, J. (2011). Planning to be surprised: Optimal Bayesian exploration in dynamic environments. In Proceedings of Fourth Conference on Artificial General Intelligence (AGI), Google, Mountain View, CA.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendices
Aaron Sloman on Schmidhuber’s “New Millennium AI and the Convergence of History 2012”
I have problems both with the style and the content of this essay, though I have not tried to take in the full mathematical details, and may therefore have missed something. I do not doubt that the combination of technical advances by the author and increases in computer power have made possible new impressive demonstrations including out-performing rival systems on various benchmark tests.
However, it is not clear to me that those tests have much to do with animal or human intelligence or that there is any reason to believe this work will help to bridge the enormous gaps between current machine competences and the competences of squirrels, nest-building birds, elephants, hunting mammals, apes, and human toddlers.
The style of the essay makes the claims hard to evaluate because it repeatedly says how good the systems are and reports that they outperform rivals, but does not help an outsider to get a feel for the nature of the tasks and the ability of the techniques to “scale out” into other tasks. In particular I have no interest in systems that do well at reading hand-written characters since that is not a task for which there is any objective criterion of correctness, and all that training achieves is tracking human labellings, without giving any explanation as to why the human labels are correct. I would be really impressed, however, if the tests showed a robot assembling Meccano parts to form a model crane depicted in a picture, and related tests here (Sloman 2011a).
Since claims are being made about how the techniques will lead beyond human competences in a few decades I would like to see sample cases where the techniques match mathematical, scientific, engineering, musical, toy puzzle solving, or linguistic performances that are regarded as highly commendable achievements of humans, e.g. outstanding school children or university students. (Newton, Einstein, Mozart, etc. can come later.) Readers should see a detailed analysis of exactly how the machine works in those cases and if the claim is that it uses non-human mechanisms, ontologies, forms of representation, etc. then I would like to see those differences explained. Likewise if its internals are comparable to those of humans I would like to see at least discussions of the common details.
The core problem is how the goals of the research are formulated. Instead of a robot with multiple asynchronously operating sensors providing different sorts of information (e.g. visual, auditory, haptic, proprioceptive, vestibular), and a collection of motor control systems for producing movements of animal-like hands, legs, wings, mouths, tongue etc., the research addresses:
... a learning robotic agent with a single life which consists of discrete cycles or time steps \(t = 1, 2, . . .\) , \(T\). Its total lifetime \(T\) may or may not be known in advance. In what follows, the value of any time-varying variable \(Q\) at time \(t(t(1\le t\le T) )\) will be denoted by \(Q(t)\), the ordered sequence of values \(Q(1),{\ldots }.,Q(t)\) by Q(\(<\)t), and the (possibly empty) sequence \(Q(1),{\ldots }., Q(t - 1)\) by \(Q(<t).\)
At any given t the robot receives a real-valued input vector \(x(t)\) from the environment and executes a real-valued action \(y(t)\) which may affect future inputs; at times \(t < T\) its goal is to maximize future success or utility....
As far as I am concerned that defines a particular sort of problem to do with data mining in a discrete stream of vectors, where the future components are influenced in some totally unexplained way by a sequence of output vectors.
I don’t see how such a mathematical problem relates to a crane assembly problem where the perceived structure is constantly changing in complexity, with different types of relationships and properties of objects relevant at different types, and actions of different sorts of complexity required, rather than a stream of output vectors (of fixed dimensionality?). I would certainly pay close attention if someone demonstrated advances in machine learning by addressing the toy crane problem, or the simpler problem described in (Sloman 2011d)
But so far none of the machine learning researchers I’ve pointed at these problems has come back with something to demonstrate. Perhaps the author and his colleagues are not interested in modelling or explaining human or animal intelligence, merely in demonstrating a functioning program that satisfies their definition of intelligence.
If they are interested in bridging the gap, then perhaps we should set up a meeting at which a collection challenges is agreed between people NOT working on machine learning and those who are, and then later we can jointly assess progress. Some of the criteria I am interested in are spelled out in these documents (Sloman 2011b, c).
However, all research results must be published in universally accessible open access journals and web sites, and not restricted to members of wealthy institutions.
Selmer Bringsjord, Alexander Bringsjord and Paul Bello on Schmidhuber’s “New Millennium AI and the Convergence of History 2012”
Hollow Hope for the Omega Point
We have elsewhere in the present volume shown that those who expect the Singularity (or, using Schmidhuber’s term, \(\Omega \)) are irrational fideists. Schmidhuber’s piece doesn’t disappoint us: while in recounting what seems all of intellectual history it reflects the brain of a bibliophage, it’s nonetheless long on faith, and short on rigorous argument.
Does it follow from the fact that “raw computing power” continues to Moore’s-Law-ishly increase, that human-level machine intelligence will arrive at some point, let alone arrive on the exuberant timeline Schmidhuber presents? No. The chief challenges in AI, relative to the human case, consist in finding the right computer programs, not faster and faster computers upon which to implement these programs (Bringsjord 2000). This is why automatic programming, one of the original dreams of AI (in which a human writes a computer program \(P\) that receives a non-executable description of an arbitrary Turing-computable function \(f\), and to succeed must produce a computer program \(P^{\prime }\) that verifiably computes \(f\)), is wholly and embarrassingly stalled. What class of being produces all the ingenious programs that increasingly form the lifeblood of the—to use Floridi’s (Floridi 2007) term—infosphere? Machines? Ha.
Does it follow from the myriad neural-network-based advances and prizes Schmidhuber cites that \(\Omega \) will ever be reached, let alone reached by 2040? No. Character/handwriting recognition is neat as far as it goes, but such low-level computation has nothing to do with what makes us us: phenomenal consciousness, free will, and natural-language communication. Taking just the latter in this brief note, character recognition has positively nothing at all to do with the fact that, say, human toddlers are vastly more eloquent than any machine. When a computing machine can not only checkmate the two of us, but debate us extemporaneously and non-idiotically in real time, we’ll take notice (or more accurately, our like-minded ancestors will). As of now, 2012, over a decade from the year Turing predicted human-machine linguistic indistinguishability, the best conversational AI is Apple’s SIRI: cute, but not much more..
Does it follow from the fact that such-and-such “breakthroughs” have happened in the past at such-and-such intervals that the Singularity will occur in accordance with some pattern Schmidhuber has magically divined? No. After all, the advances he cites are tendentiously picked to align with the kind of AI he pursues. Without question, the greatest AI achievement of the new millennium, an example of noteworthy and promising new-millennium AI if anything is, is the Watson system, produced by IBM researchers working on the basis of a relational approach found nowhere in the kinds of AI technologies that Schmidhuber venerates. Humans aren’t numerical; humans are propositional. The knowledge and abstract reasoning capacity that separates Homo sapiens sapiens from Darwin’s (Darwin 1997) “problem-solving” dogs are at their heart at once deliberative and propositional. The kind of AI that buoys Schmidhuber is neither; it’s steadfastly syntactic, not semantic. Whence his unbridled optimism?
Schmidhuber closes in a spate of humility that borders on a crestfallen concession. He raises the possibility that many of those who believe they see \(\Omega \) drawing nigh are driven by desire—desire to see the wonders of great machine intelligence. Here we commend him for his insight. What the fantast sees isn’t really there, but that he “sees” it nonetheless brings him intoxicating joy.
Rights and permissions
Copyright information
© 2012 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Schmidhuber, J. (2012). New Millennium AI and the Convergence of History: Update of 2012. In: Eden, A., Moor, J., Søraker, J., Steinhart, E. (eds) Singularity Hypotheses. The Frontiers Collection. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-32560-1_4
Download citation
DOI: https://doi.org/10.1007/978-3-642-32560-1_4
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-32559-5
Online ISBN: 978-3-642-32560-1
eBook Packages: EngineeringEngineering (R0)