Something has changed in the air around AI. It's not a dramatic turn of events, the kind that often heralds the dawn of a new era, but more like a quiet room, where everyone is suddenly looking around.
This has happened in the last few days, as several high-profile figures have begun to raise a question that was previously considered a dream. Is AI safe, they ask, in case it doesn't do what they expected?
That's a common theme in this report about why some worry that AI could “go wrong.”
As the pieces of AI come together and the technology gains power, people like Deepmind founder Demis Hassabis and OpenAI CEO Sam Altman are raising questions like: “If machines start to think for themselves, how will the programs we create ensure that their actions don't exceed the expectations of human intelligence?”
In the process, some AI systems are beginning to behave in ways not intended by their researchers.
This is described as “showing some of these skills to identify security risks, showing the kinds of actions they think are best in the environment and behave in ways that the researchers who created those systems did not expect.”
Not bad, unexpected. And this alone should be a source of discomfort.
While this warning is getting louder, there have been debates in the industry about whether we should build stronger artificial intelligence.
There will be more debate about risks and rewards as AI evolves, because that's where things get complicated. AI systems are still evolving to be more useful to humans in the future.
Some people will say that AI will eventually overtake humans and take over the world and there will be a loss of human control over it.
However, others believe that it will be used to achieve beneficial things. Then you have the question, do the benefits outweigh the risks.
We can only hope that people who know better technology will come to some conclusion about this matter.
There may be attempts to use AI systems for both good and bad purposes. There are also fears that they could be used in cyber attacks and even biological attacks in some cases.
Not everyone is the best. Although the fear of AI “going rogue” has grown, and this may not always be a healthy thing, it is good to have the conversation.
It will take years for the debate to play out. There is an opportunity for cooperation in this area, to establish some kind of governance, or at least global consensus.
AI, as the most transformative technology in the world, has already brought great benefits to humanity. This is an area that needs a lot of attention from us and our leaders.
