Reinforcing AI Safety: A Call for Governments to Regulate Data Collection
Canadian Prime Minister's recent announcement of investing $2.4-billion in artificial intelligence (AI) technology has stirred up quite a chatter in the online world, mainly due to the creation of an AI Safety Institute. Governments worldwide, including the U.S, Canada, and EU nations, have called for restrictions to mitigate the potential AI harm. To accomplish this feat on a global scale, they must regulate not only its depolyment and functions but also consider the components; algorithms, data, and computing resources or simply “compute”.
In the rapid-fire world of AI innovation, governance often finds itself falling behind. While ‘compute' and algorithmic development advance at unprecedented speeds, data regulation remains in the grey area, an issue that governments urgently need to address. To make AI systems safe, data and its collection requires stringent management, an arena where governments have proven expertise.
From determining ethnic categorizations in U.S censuses to data access management in Canada and the U.S, governments have experience in data stewardship. This notwithstanding, AI still ingests a voluminous amount of data about people's daily activities, which if left unchecked, might lead to significant harm. Ai safety is not just about the technologies or algorithms that make these systems function, but majorly about the data that is fed into them. Regulations need to consider the human angle, ensuring the collection respects human dignity and autonomy.
It is worthy of note that the data used to fuel AI systems underpins the ‘living, breathing, rights-bearing' individuals. As such, the governments should move towards creating domestic or global registries for companies to justify the data they collect and for what purposes. If organizations wish to make collected data widely accessible, they must illustrate the necessity, provide apt safeguards, and implement appropriate times and purposes for data usage. This will aid in policing unpermitted data collection and uses, with the violating companies facing punitive measures.
In the grand scheme of things, governments should look towards developing AI models that aren't heavily reliant on large data sets; a feat AI researchers are looking into. Machines can learn by mimicking infants' abilities to learn from considerably fewer experiences. Changing how machine learning models are trained, rather than consuming enormous data or adding more computing resources, might just be the key to attaining ‘intelligence' in machines.
To ensure AI safety and functionality, governments have to refocus on their strengths and work towards managing peopl's data. And while at it, remember to check how you can turn your video meetings into viral shorts with Klap AI at www.TheBestAI.org/now.
#AISafety #DataRegulation #MachineLearning #AInnovation
Which data categories do you think should be off-limits to private companies for AI purposes, and why?
