With artificial intelligence (AI) tools increasing in sophistication and usefulness, people and industries are eager to deploy them to increase efficiency, save money, and inform human decision making. But are these tools ready for the real world? As any comic book fan knows: with great power comes great responsibility. The proliferation of AI raises questions about trust, bias, privacy, and safety, and there are few settled, simple answers.
As AI has been further incorporated into everyday life, more scholars, industries, and ordinary users are examining its effects on society. The academic field of AI ethics has grown over the past five years and involves engineers, social scientists, philosophers, and others.
To trust a technology, you need evidence that it works in all kinds of conditions, and that it is accurate. "We live in a society that functions based on a high degree of trust. We have a lot of systems that require trustworthiness, and most of them we don't even think about day to day," says Caltech professor Yisong Yue. "We already have ways of ensuring trustworthiness in food products and medicine, for example. I don't think AI is so unique that you have to reinvent everything. AI is new and fresh and different, but there are a lot of common best practices that we can start from."
Today, many products come with safety guarantees, from children's car seats to batteries. But how are such guarantees established? In the case of AI, engineers can use mathematical proofs to provide assurance. For example, the AI that a drone uses to direct its landing could be mathematically proven to result in a stable landing.
This kind of guarantee is hard to provide for something like a self-driving car because roads are full of people and obstacles whose behavior may be difficult to predict. Ensuring the AI system's responses and "decisions" are safe in any given situation is complex.
One feature of AI systems that engineers test mathematically is their robustness: how the AI models react to noise, or imperfections, in the data they collect. "If you need to trust these AI models, they cannot be brittle. Meaning, adding small amounts of noise should not be able to throw off the decision making," says Anima Anandkumar, Bren Professor of Computing and Mathematical Sciences at Caltech. "A tiny amount of noise—for example, something in an image that is imperceptible to the human eye—can throw off the decision making of current AI systems." For example, researchers have engineered small imperfections in an image of a stop sign that led the AI to recognize it as a speed limit sign instead. Of course, it would be dangerous for AI in a self-driving car to make this error.
When AI is used in social situations, such as the criminal justice or banking systems, different types of guarantees, including fairness, are considered.
Dr Yampolskiy has carried out an extensive review of AI scientific literature and states he has found no proof that AI can be safely controlled – and even if there are some partial controls, they would not be enough.
He explains: “Why do so many researchers assume that AI control problem is solvable? To the best of our knowledge, there is no evidence for that, no proof. Before embarking on a quest to build a controlled AI, it is important to show that the problem is solvable.
“This, combined with statistics that show the development of AI superintelligence is an almost guaranteed event, show we should be supporting a significant AI safety effort.”
He argues our ability to produce intelligent software far outstrips our ability to control or even verify it. After a comprehensive literature review, he suggests advanced intelligent systems can never be fully controllable and so will always present certain level of risk regardless of benefit they provide. He believes it should be the goal of the AI community to minimize such risk while maximizing potential benefit.
As capability of AI increases, its autonomy also increases but our control over it decreases, Yampolskiy explains, and increased autonomy is synonymous with decreased safety.
For example, for superintelligence to avoid acquiring inaccurate knowledge and remove all bias from its programmers, it could ignore all such knowledge and rediscover/proof everything from scratch, but that would also remove any pro-human bias.
“Less intelligent agents (people) can’t permanently control more intelligent agents (ASIs). This is not because we may fail to find a safe design for superintelligence in the vast space of all possible designs, it is because no such design is possible, it doesn’t exist. Superintelligence is not rebelling, it is uncontrollable to begin with,” he explains.
“Humanity is facing a choice, do we become like babies, taken care of but not in control or do we reject having a helpful guardian but remain in charge and free.”
He suggests that an equilibrium point could be found at which we sacrifice some capability in return for some control, at the cost of providing system with a certain degree of autonomy.
One control suggestion is to design a machine which precisely follows human orders, but Yampolskiy points out the potential for conflicting orders, misinterpretation or malicious use.
He explains: “Humans in control can result in contradictory or explicitly malevolent orders, while AI in control means that humans are not.”
If AI acted more as an advisor it could bypass issues with misinterpretation of direct orders and potential for malevolent orders, but the author argues for AI to be useful advisor it must have its own superior values.
“Most AI safety researchers are looking for a way to align future superintelligence to values of humanity. Value-aligned AI will be biased by definition, pro-human bias, good or bad is still a bias. The paradox of value-aligned AI is that a person explicitly ordering an AI system to do something may get a “no” while the system tries to do what the person actually wants. Humanity is either protected or respected, but not both,” he explains.
To minimize the risk of AI, he says it needs it to be modifiable with ‘undo’ options, limitable, transparent and easy to understand in human language.
He suggests all AI should be categorised as controllable or uncontrollable, and nothing should be taken off the table and limited moratoriums, and even partial bans on certain types of AI technology should be considered.
Instead of being discouraged, he says: “Rather it is a reason, for more people, to dig deeper and to increase effort, and funding for AI Safety and Security research. We may not ever get to 100% safe AI, but we can make AI safer in proportion to our efforts, which is a lot better than doing nothing. We need to use this opportunity wisely.”
0 Comments