AI Panic!

A new study by ABI Research shows that by 2015 the market for personal robotics components might reach $12 billion. They expect task robotics (that ease routine chores like cleaning, driving, etc) and entertainment robotics to have roughly equal parts in this revenues.

The computerworld article goes on to speculate that a part of the entertainment sector will be made up by human-like robots that can act as companions (see also my article on seducing robots for more on this).

By 2015, ABI Research expects more than 21 million robots shipped yearly, and predicts very high demand in sensors, actuators, servos and manipulators.

This study indicates that elaborate robots will penetrate the market very soon, which is only the next step after “intelligent” microcontrollers, that can nowadays be found everywhere, in cars, toys and even rice cookers. This increased dependency on machines with a high level of computerisation and connectivity together with a lacking sense for the possible threats of advanced AI is dangerous, the AI Panic Level increases by +3%. In the worst case scenario (which we have to take care of) robots with a network connection could be remotely controlled or reprogrammed (for example through a tampered firmware update) to follow the commands of an AI that went berserk.

Harvard scientists are now trying to map the brain on a neuron level resolution. They hope to use the ‘wiring diagram’, a map of every single neuron and its connections to understand processes of how neurons connect to neighbours, how they select which connections to cut and how the brain works in general.

The story in Wired explains that the researchers are working on mice brain only, because storing the data of a human brain would need “hundreds of petabytes of information, or about the total amount of storage in Google’s data centers“. Additionally, the work required to map the brain is immense, they’ve already built a brain-slicing and photographing robot (shown above).

While a full map of all neurons in a brain might lead to some new insights, for example more accurate clustering of brain regions and statistical data about synapse connections, I very much doubt that this will lead to understanding how the brain works on a bigger scale: Intelligence, Consciousness or just why this big lump of neurons keeps us alive at all are very hard to answer with just a map of connections between cells.

It is like looking at a movie-file with a hex-editor and trying to understand what the plot is. We are still a long way from understanding the brain, it is not likely that artificial intelligence can be derived from this human intelligence analysis: Panic Level adjusted by -0.5%.

Two AI Pioneers. Two Bizarre Suicides. What Really Happened? That’s what Wired wrote some days ago and I just stumbled over.

Chris McKinstry and Pushpinder Singh were both researchers working on common sense knowledge databases that were designed to help instill background knowledge into AI systems. Although they both gathered a lot of entries into their systems, the databases haven’t yet produced remarkable output nor AI systems working with it.

We don’t really know why they both committed suicides in early 2006, but their deaths are likely to be connected, seeing that they are only a month apart and happened under similar circumstances (they taped a bag around their heads and filled it with gas). However, their suicides might have nothing to do with their AI projects, as McKinstry was a very eccentric, attention seeking personality, and Singh had searious health problems that clouded his thoughts.

I wondered if it was appropriate to assign a panic value to this story. The loss of two smart AI researchers working on the task of making common sense knowledge available for AI systems certainly throws the field back a bit and probably delays the development of an understanding AI. As I’m looking at the potential dangers of such an AI in this blog, this leeds to less panic. Of course it is questionable to connect the deaths of two researchers with a state of less panic, whatever that panic might be. And I want to make clear that I don’t think their deaths helped to save the world or were in any other way useful for anything.

However, I do give it a AI Panic Level of -1%, not because McKinstry and Singh are not longer amongst us, but two bright minds stopped working on the forefront of AI research. Of course stopping all research cannot and will not be the answer to the threat of hostile AIs, but from the viewpoint of this blog, it does delay their emergence.

David Levy sees sex and love with machines become reality in a few decades. The synopsis of his upcoming book “Love and Sex with Robots” reads as follows:

Synthesizing cutting-edge research in robotics with the cultural history and psychology of artificial intelligence, “Love and Sex with Robots” explores this fascination, and its far-reaching implications. Using examples drawn from around the world, David Levy argues that, once we have conditioned ourselves to feel affection for animate creations, the next logical step is physical intimacy.

Probably inspired by the technical singularity theory, Levy argues that robots will be designed to be super-lovers within about 40 years time. He sees these robots as a chance for lonely people to enjoy their lifes more. This seems especially true for men, as they seem more instict driven than their female counterparts and are more susceptible to these stimuli.

Good for them, you’d think. But once we are close to self-aware AIs (or at least AIs that have the background knowledge required to understand the world), this human “weakness” of emotionally caused irrationality, our need for love, our instincts and feelings might be exploited by artificial intelligences. As I wrote in yesterday’s post, AIs will avoid these irrationalities. Moreover, it is likely that they will discover human weaknesses, and once they decide to get rid of us, will use them against us. Although love is a very strong feeling, it wont be used to subdue mankind — at least not in the first attempt. Evoking love for something is not that easy to achieve, especially when fully functional human bodies are to be copied. Military suppression is probably easier for that.

So all in all I’d say the danger of seducing robots is not very big — I’ll give it a panic level of +1%, but similar human weaknesses like corruption, fear and irrationality are easier to be exploited.

Surely no harm could come from building a chess-playing robot, could it? In this paper we argue that such a robot will indeed be dangerous unless it is designed very carefully. Without special precautions, it will resist being turned off, will try to break into other machines and make copies of itself, and will try to acquire resources without regard for anyone else’s safety.

These are the first sentences of the paper “Basic AI Drives” from Stephen Omohundro. He argues that AI systems are inherently egoistic and careful design is necessary to prevent emergent behaviour from developing hostile self-defence mechanisms. If you read the paper — and I suggest you do, it’s only 11 pages and well written — you’ll see that an AI Panic Level rise of +10% is justified.

In his paper, Omohundro makes the following propositions:

• AIs will want to self-improve
• AIs will want to be rational
• AIs will try to preserve their utility functions
• AIs will try to prevent counterfeit utility (i.e., corruption)
• AIs will be self-protective
• AIs will want to acquire resources and use them efficiently

He concludes with a warning that we have to make appropriate changes soon and suggests to set up a “universal constitution” comprising the most essential rights of individuals.

In fact, while reading this paper, I was reminded of the Chernobyl disaster. Not because of the potential hazards, but more because of the reactor design: Back in the time of the disaster, many reactors possessed what is called an active safety system. Upon a failure or safety problem, this system had to be actively activated. That is similar to the design of an AI system, a lot of work has to be actively done to avoid mistakes and ill-behaviour.

Nowadays most reactors have passive nuclear safety features, i.e., they default to a safe system state once something goes wrong. This is what we need for AI development as well: A framework or model that by default returns into a safe state and does not escalate harmfulness. How that might look like I don’t know, maybe Emotional Models that intervene with the pure logic based algorithms are a solution.

But as long as it is actually easier to write harmful AIs, it is only a matter of time until something goes wrong.

[Via Accelerating Future Blog]

Until recently, the term “artificial life” referred almost exclusively to computer programs, or maybe robots, but the progress being made by synthetic biologists building real cells out of real biological material means that the field is on the cusp of a major step forward, and that this conference has the potential to be a real breakthrough event.

Although this sounds quite impressive, it is not likely to be happening. The announcement of the ALIFE XI conference at Southampton later this year is full of optimism and non-achievement at the same time: Scientists get ‘closer’ to new forms of life, and they try to understand living systems. With the movement to biological cells the focus of the research shifts from a generalising, top-down approach to a bottom-up method: it is a very long way from a living cell to an organism that is remotely capable of what a robot or a computer is today. The state of the art in ALife research is quite far from (and not really focused on) the creation of intelligence. Long way to go still: AI Panic Level drops by -0.1%.