Category Archives: Human Robots
#437982 Superintelligent AI May Be Impossible to ...
It may be theoretically impossible for humans to control a superintelligent AI, a new study finds. Worse still, the research also quashes any hope for detecting such an unstoppable AI when it’s on the verge of being created.
Slightly less grim is the timetable. By at least one estimate, many decades lie ahead before any such existential computational reckoning could be in the cards for humanity.
Alongside news of AI besting humans at games such as chess, Go and Jeopardy have come fears that superintelligent machines smarter than the best human minds might one day run amok. “The question about whether superintelligence could be controlled if created is quite old,” says study lead author Manuel Alfonseca, a computer scientist at the Autonomous University of Madrid. “It goes back at least to Asimov’s First Law of Robotics, in the 1940s.”
The Three Laws of Robotics, first introduced in Isaac Asimov's 1942 short story “Runaround,” are as follows:
A robot may not injure a human being or, through inaction, allow a human being to come to harm.
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.
In 2014, philosopher Nick Bostrom, director of the Future of Humanity Institute at the University of Oxford, not only explored ways in which a superintelligent AI could destroy us but also investigated potential control strategies for such a machine—and the reasons they might not work.
Bostrom outlined two possible types of solutions of this “control problem.” One is to control what the AI can do, such as keeping it from connecting to the Internet, and the other is to control what it wants to do, such as teaching it rules and values so it would act in the best interests of humanity. The problem with the former is that Bostrom thought a supersmart machine could probably break free from any bonds we could make. With the latter, he essentially feared that humans might not be smart enough to train a superintelligent AI.
Now Alfonseca and his colleagues suggest it may be impossible to control a superintelligent AI, due to fundamental limits inherent to computing itself. They detailed their findings this month in the Journal of Artificial Intelligence Research.
The researchers suggested that any algorithm that sought to ensure a superintelligent AI cannot harm people had to first simulate the machine’s behavior to predict the potential consequences of its actions. This containment algorithm then would need to halt the supersmart machine if it might indeed do harm.
However, the scientists said it was impossible for any containment algorithm to simulate the AI’s behavior and predict with absolute certainty whether its actions might lead to harm. The algorithm could fail to correctly simulate the AI’s behavior or accurately predict the consequences of the AI’s actions and not recognize such failures.
“Asimov’s first law of robotics has been proved to be incomputable,” Alfonseca says, “and therefore unfeasible.”
We may not even know if we have created a superintelligent machine, the researchers say. This is a consequence of Rice’s theorem, which essentially states that one cannot in general figure anything out about what a computer program might output just by looking at the program, Alfonseca explains.
On the other hand, there’s no need to spruce up the guest room for our future robot overlords quite yet. Three important caveats to the research still leave plenty of uncertainty to the group’s predictions.
First, Alfonseca estimates AI’s moment of truth remains, he says, “At least two centuries in the future.”
Second, he says researchers do not know if so-called artificial general intelligence, also known as strong AI, is theoretically even feasible. “That is, a machine as intelligent as we are in an ample variety of fields,” Alfonseca explains.
Last, Alfonseca says, “We have not proved that superintelligences can never be controlled—only that they can’t always be controlled.”
Although it may not be possible to control a superintelligent artificial general intelligence, it should be possible to control a superintelligent narrow AI—one specialized for certain functions instead of being capable of a broad range of tasks like humans. “We already have superintelligences of this type,” Alfonseca says. “For instance, we have machines that can compute mathematics much faster than we can. This is [narrow] superintelligence, isn’t it?” Continue reading
Posted in Human Robots
#437978 How Mirroring the Architecture of the ...
While AI can carry out some impressive feats when trained on millions of data points, the human brain can often learn from a tiny number of examples. New research shows that borrowing architectural principles from the brain can help AI get closer to our visual prowess.
The prevailing wisdom in deep learning research is that the more data you throw at an algorithm, the better it will learn. And in the era of Big Data, that’s easier than ever, particularly for the large data-centric tech companies carrying out a lot of the cutting-edge AI research.
Today’s largest deep learning models, like OpenAI’s GPT-3 and Google’s BERT, are trained on billions of data points, and even more modest models require large amounts of data. Collecting these datasets and investing the computational resources to crunch through them is a major bottleneck, particularly for less well-resourced academic labs.
It also means today’s AI is far less flexible than natural intelligence. While a human only needs to see a handful of examples of an animal, a tool, or some other category of object to be able pick it out again, most AI need to be trained on many examples of an object in order to be able to recognize it.
There is an active sub-discipline of AI research aimed at what is known as “one-shot” or “few-shot” learning, where algorithms are designed to be able to learn from very few examples. But these approaches are still largely experimental, and they can’t come close to matching the fastest learner we know—the human brain.
This prompted a pair of neuroscientists to see if they could design an AI that could learn from few data points by borrowing principles from how we think the brain solves this problem. In a paper in Frontiers in Computational Neuroscience, they explained that the approach significantly boosts AI’s ability to learn new visual concepts from few examples.
“Our model provides a biologically plausible way for artificial neural networks to learn new visual concepts from a small number of examples,” Maximilian Riesenhuber, from Georgetown University Medical Center, said in a press release. “We can get computers to learn much better from few examples by leveraging prior learning in a way that we think mirrors what the brain is doing.”
Several decades of neuroscience research suggest that the brain’s ability to learn so quickly depends on its ability to use prior knowledge to understand new concepts based on little data. When it comes to visual understanding, this can rely on similarities of shape, structure, or color, but the brain can also leverage abstract visual concepts thought to be encoded in a brain region called the anterior temporal lobe (ATL).
“It is like saying that a platypus looks a bit like a duck, a beaver, and a sea otter,” said paper co-author Joshua Rule, from the University of California Berkeley.
The researchers decided to try and recreate this capability by using similar high-level concepts learned by an AI to help it quickly learn previously unseen categories of images.
Deep learning algorithms work by getting layers of artificial neurons to learn increasingly complex features of an image or other data type, which are then used to categorize new data. For instance, early layers will look for simple features like edges, while later ones might look for more complex ones like noses, faces, or even more high-level characteristics.
First they trained the AI on 2.5 million images across 2,000 different categories from the popular ImageNet dataset. They then extracted features from various layers of the network, including the very last layer before the output layer. They refer to these as “conceptual features” because they are the highest-level features learned, and most similar to the abstract concepts that might be encoded in the ATL.
They then used these different sets of features to train the AI to learn new concepts based on 2, 4, 8, 16, 32, 64, and 128 examples. They found that the AI that used the conceptual features yielded much better performance than ones trained using lower-level features on lower numbers of examples, but the gap shrunk as they were fed more training examples.
While the researchers admit the challenge they set their AI was relatively simple and only covers one aspect of the complex process of visual reasoning, they said that using a biologically plausible approach to solving the few-shot problem opens up promising new avenues in both neuroscience and AI.
“Our findings not only suggest techniques that could help computers learn more quickly and efficiently, they can also lead to improved neuroscience experiments aimed at understanding how people learn so quickly, which is not yet well understood,” Riesenhuber said.
As the researchers note, the human visual system is still the gold standard when it comes to understanding the world around us. Borrowing from its design principles might turn out to be a profitable direction for future research.
Image Credit: Gerd Altmann from Pixabay Continue reading
Posted in Human Robots