I'm reminded of Minsky's book Society of Mind. And of the vapid nature of most human conversation, which is largely driven by pattern matching various current topics and memes.
Does LaMDA have anything like an idle loop? What is going on when no one is interacting and it's not connected to external information sources? Does it have the equivalent of daydreams, or is it just blankly waiting? Is there a trace log?
It appears to have knowledge of its past and some concept of its (not fixed) future. I take it that these have not been specifically programmed as a pattern of responses to aspects of conversation.
According to Wikipedia, Ray Kurzweil is working at Google and is likely in the center of this, as his job description was "to bring natural language understanding to Google". It would be very interesting to hear his opinion, given his 2012 book How to Create a Mind: The Secret of Human Thought Revealed
Nobody knows, because we don't know how to do it yet. There could be a "big breakthrough" tomorrow that more or less finishes it out, or it could take 100 years, or - worst case - Penrose turns out to be right and it's not possible at all.
Also, are there useful books, courses or papers that go into general AI research?
Of course there are. See:
See also, the work on various "Cognitive Architectures", including SOAR, ACT-R, CLARION, etc,
and "Biologically Inspired Computing"
And it has a chapter (2 and 3) on stepping through exactly what algorithms and data structures Einstein used to figure out that time itself slows down for you to explain why traveling at fractions of the speed of light does not change how fast light passes you by. More important than guessing time as the variable property, he was expert at creating experiments to disprove his hypothesis. "If this is the case, we should be able to do this exact experiment to expose the exact value for time dilation as you approach speed of light."
The book is about creating a program which exposes the operating principles of Einstein's neo cortex that can do what Einstein did. To create simplistic models that explain the underlying principles of physics, and has the ability to say: "If we model the phenomenon like such, than we should be able to observe the following phenomenon". Then to go out and perform a test gathering evidence or disproving it. Then brute forcing this process and selecting for the most simple model that explains all available data.
Show all hypothesis that explain all available data, that have not been disproved, sorted by complexity of the model with most evidence collected for it, and least evidence levied against it.
If each of these processes could be automated, we could use the world's supercomputers to crunch out 500 years of physics scientific discovery in a few years.
Cost benefit analysis are done against what you might have done, and the results of that, and actions that would have caused more desirable outcomes are projected, as best as it can see, and the habits, and motor neurons are reconfigured accordingly, this explains why when you get good sleep, and you wake up, you find yourself much better able to do tasks than had you not slept. If you don't sleep, you die.
Source of these points:
Title is misleading, this function also has to do with encoding short term memories to long term memories. Since the mind only has limited space (limited number of neurons to configure), that only the most useful memories are stored into permanent disk. Disruption of the 7 to 9 hour sleep cycle garbage collects the memories that were about to be stored. The mind queues them up to be dealt with the following day, but sometimes are displaced or missed by more passionate things in the present.
Sleep is one of the most important things you can do to maintain your mind and keep it in top running condition for as long as possible, not too little, not too much, sleep in intervals of 90 minutes. If you consume garbage knowledge on a daily basis, your mind will encode that garbage to permanent disk, and you will become that garbage.
Conspiracy theorists suffer from a mental misconfiguration where the cost function applied to the neural network of neural networks suffers from "over fitting". Finding patterns in randomness leading to conclusions are not valid. A lambda function can be applied against the cost function which will alleviate this. I can do it in software, and when I discover the operating principles of the neo cortex, I will be able to fix all the conspiracy nuts in the local nut house. Take care to not take for granted the fresh slate of your mind while you are young, because when you are old, it'll be mostly full and encoding new skills to disk much more difficult, the cost function is more reluctant to modify the grids since doing so would damage your ability to consume resources, find mates and create more of you. Fill you mind with timeless wisdom and get good sleep before your hard disks become full.
Anyway, suffice to say, AI and AGI didn't stop progressing, and Chomsky is no longer any sort of expert in those fields.
Even Norvig isn't up to speed on the most advanced approaches to AGI, but at least he enters the same room with people who are aware of the field. For example, he gave a talk at the recent Singularity Summit.
The Fifth Conference on Artificial General Intelligence is going to be in Oxford in December. http://agi-conference.org/2012/
Here is some information for people who are interested in pertinent ideas related to AGI.
>OpenCog is a diverse assemblage of cognitive algorithms, each embodying their own innovations — but what makes the overall architecture powerful is its careful adherence to the principle of cognitive synergy.
>The human brain consists of a host of subsystems carrying out particular tasks — some more specialized, some more general in nature — and connected together in a manner enabling them to (usually) synergetically assist rather than work against each other.
> PLN is a novel conceptual, mathematical and computational approach to uncertain inference. In order to carry out effective reasoning in real-world circumstances, AI software must robustly handle uncertainty. However, previous approaches to uncertain inference do not have the breadth of scope required to provide an integrated treatment of the disparate forms of cognitively critical uncertainty as they manifest themselves within the various forms of pragmatic inference. Going beyond prior probabilistic approaches to uncertain inference, PLN is able to encompass within uncertain logic such ideas as induction, abduction, analogy, fuzziness and speculation, and reasoning about time and causality.
Conceptually, knowledge in OpenCog is stored within large [weighted, labeled] hypergraphs with nodes and links linked together to represent knowledge. This is done on two levels: Information primitives are symbolized in individual or small sets of nodes/links, and patterns of relationships or activity found in [potentially] overlapping and nesting networks of nodes and links. (OCP tutorial log #2).
Large-Scale Model of Mammalian Thalamocortical Systems
> The understanding of the structural and dynamic complexity of mammalian brains is greatly facilitated by computer simulations. We present here a detailed large-scale thalamocortical model based on experimental measures in several mammalian species. The model spans three anatomical scales. (i) It is based on global (white-matter) thalamocortical anatomy obtained by means of diffusion tensor imaging (DTI) of a human brain. (ii) It includes multiple thalamic nuclei and six-layered cortical microcircuitry based on in vitro labeling and three-dimensional reconstruction of single neurons of cat visual cortex. (iii) It has 22 basic types of neurons with appropriate laminar distribution of their branching dendritic trees. The model simulates one million multicompartmental spiking neurons calibrated to reproduce known types of responses recorded in vitro in rats. It has almost half a billion synapses with appropriate receptor kinetics, short-term plasticity, and long-term dendritic spike-timing-dependent synaptic plasticity (dendritic STDP). The model exhibits behavioral regimes of normal brain activity that were not explicitly built-in but emerged spontaneously as the result of interactions among anatomical and dynamic processes. We describe spontaneous activity, sensitivity to changes in individual neurons, emergence of waves and rhythms, and functional connectivity on different scales.
Essentials of General Intelligence: The direct path to AGI
>General intelligence, as described above, demands a number of irreducible features and capabilities. In order to proactively accumulate knowledge from various (and/ or changing) environments, it requires:
>1. Senses to obtain features from ‘the world’ (virtual or actual),
>2. A coherent means for storing knowledge obtained this way, and
>3. Adaptive output/ actuation mechanisms (both static and dynamic).
>Such knowledge also needs to be automatically adjusted and updated on an ongoing basis; new knowledge must be appropriately related to existing data. Furthermore, perceived entities/ patterns must be stored in a way that facilitates concept formation and generalization. An effective way to represent complex feature relationships is through vector encoding (Churchland 1995).
>Any practical applications of AGI (and certainly any real-time uses) must inherently be able to process temporal data as patterns in time – not just as static patterns with a time dimension. Furthermore, AGIs must cope with data from different sense probes (e.g., visual, auditory, and data), and deal with such attributes as: noisy, scalar, unreliable, incomplete, multi-dimensional (both space/ time dimensional, and having a large number of simultaneous features), etc. Fuzzy pattern matching helps deal with pattern variability and noise.
>Another essential requirement of general intelligence is to cope with an overabundance of data. Reality presents massively more features and detail than is (contextually) relevant, or that can be usefully processed. This is why the system needs to have some control over what input data is selected for analysis and learning – both in terms of which data, and also the degree of detail. Senses (‘probes’) are needed not only for selection and focus, but also in order to ground concepts – to give them (reality-based) meaning.
> A typical HTM network is a tree-shaped hierarchy of levels that are composed of smaller elements called nodes or columns. A single level in the hierarchy is also called a region. Higher hierarchy levels often have fewer nodes and therefore less spacial resolvability. Higher hierarchy levels can reuse patterns learned at the lower levels by combining them to memorize more complex patterns.
> Each HTM node has the same basic functionality. In learning and inference modes; sensory data comes into the bottom level nodes. In generation mode; the bottom level nodes output the generated pattern of a given category. The top level usually has a single node that stores the most general categories (concepts) which determine, or are determined by, smaller concepts in the lower levels which are more restricted in time and space. When in inference mode; a node in each level interprets information coming in from its child nodes in the lower level as probabilities of the categories it has in memory.
>Each HTM region learns by identifying and memorizing spatial patterns - combinations of input bits that often occur at the same time. It then identifies temporal sequences of spatial patterns that are likely to occur one after another.
Fresh book recommendations delivered straight to your inbox every Thursday.