Concept 3: Artificial General Intelligence

• Post1 What is AGI?

Let’s start at the beginning. Why do we even need this term?

60 years ago when the term ‘AI’ was coined, the ambition was to build machines that can learn and reason like humans. Over several decades of trying and failing (badly), the original vision was largely abandoned. Nowadays almost all AI work relates to narrow, domain-specific, human-designed capabilities. Powerful as these current applications may be, they are limited to their specific target domain, and have very narrow (if any) adaptation or interactive learning ability. Most computer scientists graduating after the mid 80’s only know AI from that much watered-down perspective.

However, just after 2000 several of us felt that hardware, software, and cognitive theory had advanced sufficiently to rekindle the original dream. At that time we found about a dozen people actively doing research in this area, and willing to contribute to a book to share ideas and approaches. After some deliberation, three of us (Shane Legg, Ben Goertzel and myself) decided that ‘Artificial General Intelligence’, or AGI, best described our shared approach. We felt that we wanted to give our community a distinctive identity, to differentiate our work from mainstream AI which is unlikely to lead to general intelligence.

The term ‘AGI’ gave a name to this emerging group of researchers, scientists, and engineers who were actually getting back to trying to develop ‘real AI’. This ‘movement’ was officially launched with the publication of the book Artificial General Intelligence , and has since gathered momentum with additional publications and annual AGI conferences. By now, the term has become quite widely used to refer to machines with human, or super-human level capabilities.

Some people have suggested using ‘AGI’ for any work that is generally in the area of autonomous learning, ‘model-free’, adaptive, unsupervised or some such approach or methodology. I don’t think this is justified, as many clearly narrow AI projects use such methods. One can certainly assert that some approach or technology will likely help achieve AGI, but I think it is reasonable to judge projects by whether they are explicitly on a path(however far away it may be) to achieving the grand vision: a single system that can learn incrementally, reason abstractly, and act effectively over a wide range of domains — just like humans can.

Elsewhere I’ve elaborated on what human intelligence entails; here I want to take a slightly different angle and ask “What would it take for us to say we’ve achieved AGI?”. This is my proposed descriptive definition, followed by some elaboration:

A computer system that matches or exceeds the real time cognitive (not physical) abilities of a smart, well-educated human.

Cognitive abilities include, but are not limited to: holding productive conversations; learning new commercial and scientific domains in real time through reading, coaching, experimentation, etc.; applying existing knowledge and skills to new domains. For example, learning new professional skills, a new language (including computer languages), or even novel games.

Acceptable limitations include: very limited sense acuity and dexterity.

Alternative suggestions, and their merits

“Machines that can learn to do any job that humans currently do” — I think this fits quite well, except that it seems unnecessarily ambitious. Machines that can do most jobs, especially mentally challenging ones would get us to our overall goal of having machines that can help us solve difficult problems like ageing, energy, pollution, and help us think through political and moral issues, etc. Naturally, they would also help to build machines that will handle remaining jobs we want to automate.

“Machines that pass the Turing Test”— The current Turing Test asks too much (potentially having to dumb itself down to fool judges that it is human), and too little (limited conversation time). A much better test would be to see if the AI can learn a broad range of new complex human-level cognitive skills via autonomous learning and coaching.

“Machines that are self-aware/ can learn autonomously/ do autonomous reduction/ etc.”— These definition grossly underspecify AGI. One could build narrow systems that have these characteristics (and probably have already), but are nowhere near AGI (and may not be on the path at all).

“A machine with the ability to learn from its experience and to work with insufficient knowledge and resources.”— Important requirements but lacking specification of the level skill one expects. Again, systems already exist that have these qualities but are nowhere near AGI.

Some objections

Why specify AGI in terms of human abilities?— While we’d expect AGI cognition to be quite different (instant access to Internet, photographic memory, logical thinking, etc.), the goal is still to free us from most work. In order to do that it must be able to operate in our environment, and learn interactively via natural language and human interaction.

Why not require full sense acuity, dexterity, and embodiment?— I think that a reasonable relaxation of requirements is to initially exclude tasks that require high dexterity & sense acuity. The reason is that initial focus should be on cognitive ability — ie. a “Helen Hawking” (Helen Keller/ Stephen Hawking)” The core problem is building the brain, the intelligence engine. It can’t be totally disconnected from the world, but its senses/ actuators do not need to be very elaborate, as long as it can operate other machines (tool use).

• Post2 Towards Artificial General Intelligence

Recent success of Deep Learning in complex tasks like image recognition, speech recognition, and game playing has created a wave of discussions and arguments about Artificial General Intelligence (AGI) and its potential threat to humanity. Thanks to the media hype (and over-promising claims in the introductions and conclusions of recent research papers), people have started believing that science fictions will soon become a reality. Are we really close to solving artificial general intelligence? What are the essential characteristics an agent should possess to claim “the AGI agent” title? In this article, I will try to answer these questions from my point of view.

Firstly, what is Artificial General Intelligence? Wikipedia defines AGI as follows:

“Artificial general intelligence (AGI) is the intelligence of a (hypothetical) machine that could successfully perform any intellectual task that a human being can.”

This is also referred to as Strong AI or Full AI. Firstly, do we have a system which has this AGI capability? Answer is NO. Then where is this recent progress in machine learning research leading us to? To answer this question, let us try to think about another related question. What will happen if someone comes up with a learning algorithm that achieves 0% error in ImageNet classification task? Can that learning algorithm be used to create an AGI agent? Now we should go back to the definition of AGI. An AGI agent can successfully perform any classification doesn’t guarantee that the same algorithm can answer questions intellectual task that a human being can. Achieving 0% error in image (Question Answering) or play chess at even a novice chess player level. This immediately sheds some light on the reality. Several recent papers claim that their systems surpass human performance in a particular task. Whether this statement is true is a topic for separate debate. Even if this is true, we need not worry. Such systems will break as soon as you put them on a different job.

Current systems are not sufficient to achieve AGI. Well, then what are necessary to achieve AGI? There might be several ways to approach this problem. But here are my thoughts about one possible way. Before explaining the roadmap towards AGI, let me describe five important characteristics any AGI system should possess.

Ability to multi-task aka multi-task learning:

It is very clear that the general AI agent should be able to multi-task. An average human being can do 100s of tasks without significant effort. It is perfectly reasonable to expect this quality from an AGI agent. A system which calls itself an AGI agent should be able to describe images and at the same time answer your questions or recognize your friends. One naive way to achieve this is to learn multiple tasks independently using multiple models. However we do not exploit the common factors between related tasks in this method. A more elegant solution is where a single system can learn multiple tasks simultaneously. This is known as multi-task learning in the Machine Learning literature. The basic idea of multi-task learning is to design a single general learning algorithm that can learn multiple tasks simultaneously, with the hope that each task would help learning of other related tasks. Rich Caruana’s PhD thesis provides an excellent introduction to multi-task learning.

Recently there has been a fairly decent amount of progress in multi-task learning. For example, in language understanding (which is my primary interest), Collobert et al., 2011 proposed a neural network method to learn multiple NLP tasks simultaneously and achieved near-state-of-the-art performance in all the tasks. Note that their system is not specialized to any single task while the state-of-the-art methods for each task are highly task-specific. Recent success in factoid question answering (Bordes et al., 2015 and previous papers) is also based on multi-task learning where they multi-task question answering with paraphrase detection tasks. But, all these multi-tasking are in same source or very similar sources (It was “text” in these examples). Nevertheless, we are really far from general multi-task learning algorithms which can multi-task with a bunch of heterogeneous source domains (e.g., image, text, and speech signal).

Ability to reuse the knowledge aka transfer learning:

Remember the first time you started playing a video game? You might have taken a few weeks to become a pro-player in that game. But how much time did you take to master the next game? Probably, just a few days! This is an amazing capability of the human mind! We humans can seamlessly transfer our learning experience from one task to another related task and learn the new task much faster. This is known as Transfer learning in the machine learning literature. One can view multi-task learning as a transfer learning problem in the sense that the goal of multi-task learning is to transfer some knowledge between the tasks so that they help each other. However, it is just one type of transfer learning. You can also do transfer learning when you learn multiple tasks sequentially. Very recently Rajendran et al., 2015a and Parisotto et al., 2015 proposed ways to do transfer learning across multiple decision making tasks. Rajendran et al., 2015a highlights a significant problem in transfer learning which is negative transfer between tasks (one task affecting the performance of another task) and proposes methods to avoid negative transfer. Both these work are preliminary steps towards solving general transfer learning problem and solving transfer learning is essential to have a generic AI system.

Learning from few examples aka zero-shot/one-shot learning:

Let us assume we have a system which can do multi-task transfer learning. Is it sufficient? Can it achieve human-level performance? To say that a system has achieved human level performance, it should also be able to learn from similar amount of data (examples) to learn a task. How many examples does a kid require to learn to identify a dog? One or two? May be three. But the current image recognition system requires at least hundreds of examples to learn to identify dogs in the image. What an AGI agent really needs is the ability to pick up new tasks faster. This is known as zero-shot/one-shot learning in machine learning (learning from zero or one example respectively). Very recently, Lake et al., 2015 proposed an algorithm to do one-shot classification. This is a significant step towards the grand goal and more research in this direction is required.

Ability to make use of multiple modalities aka multimodal learning:

As mentioned previously, most of the existing research focuses on multi-task transfer learning across similar sources. For example, Chandar et al., 2015 proposes a framework to do multi-task transfer learning which is restricted to the case where sources are multiple languages and tasks are same but in different languages. But to call an algorithm a general purpose AI algorithm, it should be able to work with multiple modalities of information like text, image, speech, and video. Recently, researchers have started looking into this problem and we still do not have general algorithms which can take arbitrary set of modalities. For recent progress, have a look at Ngiam et al., 2011, Srivastava et al., 2014, Rajendran et al., 2015b.

Ability to learn through feedback aka reinforcement learning:

Again, let us assume that we have a multi-modal multi-task transfer learning algorithm. Can such an algorithm be completely supervised? Probably not. We humans learn a lot of tasks without any direct supervision. Instead we learn them mostly through interacting with environment (i.e., by taking a sequence of actions and having indirect feedback on the actions). Remember how you learned to ride a bicycle? Showing you videos of people riding bicycle would not have worked. Isn’t it right? You really learned to ride only when you tried it yourself and fell down a couple of times (negative reinforcement). This is the idea of reinforcement learning. In reinforcement learning, an agent interacts with the environment and learns through feedback. This is an essential capability for a general purpose AI agent. Reinforcement Learning is one of the oldest fields in machine learning and it has a wealthy literature by itself. However, we still lack RL agents who can learn tasks with very few trials and can take a large number of actions. This requires some serious research.

The Master Algorithm:

Now let us consolidate all the necessary capabilities of an AGI agent. It should be able to perform multi-task learning, transfer learning, zero-shot/one-shot learning, multi-modal learning, reinforcement learning.

Such an algorithm would be a real master algorithm. I am reusing the word “Master Algorithm” from Pedro Domingo’s famous book “The Master Algorithm”. But we take completely different approaches. While his goal was to describe the master algorithm as a unification of different schools of machine learning, my goal is to describe the capabilities such a master algorithm should have. If we can have a system with all these capabilities, then such a system would perform many intellectual tasks that a human being can.

Although I have focused on the aspect of learning algorithms in pursuing what we need to achieve AGI, apart from these, there are also many other problems that need to be solved on the way. e.g., some domain specific problems in computer vision, natural language processing, planning, search, control, etc. However, I believe that the five components that I identified here will play an important role in achieving AGI.

A solution for artificial general intelligence is one of key approaches to understand how human brain works (assuming that the nature is an optimal player). Even if we can achieve inventing such a master algorithm, none of these capabilities would help the algorithm gain self-consciousness (are all humans self-conscious?) and I would say that the master algorithm which doesn’t know its existence is still not a threat to humanity, IF it does not get into the hands of evil people.

I agree that there is an exponential progress in all these sub-fields. But we are not sure if the rate of the exponent is good enough to solve AGI in next few years. I am not being too optimistic and I would expect at least few more decades to have a baby AGI agent. OK. Let me go and work on my baby AGI agent ;)

• Post3 Geoffrey Hinton and Demis Hassabis: AGI is nowhere close to being a reality