Artificial Intelligence: Regulation
Viscount Camrose Excerpts(1 year, 4 months ago)
Lords ChamberRead Full debate Read Hansard Text Watch Debate Read Debate Ministerial Extracts
Lord Vallance of Balham (Lab)
That is an area that of course comes under several other parts of regulation already. It is also an area where there are massive changes in the way that these models perform. If one looks at GPT-4 versus GPT-3—I know it is not facial recognition, but it gives an indication of the types of advances—it is about twice as good now as it was a year ago. These things are moving fast and there is indeed a need to understand exactly how facial recognition technology is valid and where it has problems in recognition.
Viscount Camrose (Con)
My Lords, the supply chain for the development of the more advanced AI systems is, in almost every case, highly global in nature. That means that it becomes quite straightforward for AI developers to offshore their activities from any jurisdiction whose regulations they might prefer not to follow. This being the case, do the Government agree that the regulations for AI development, as distinguished mostly from use, are going to have to be global in nature? If the Government agree with that, how is it reflected in their plans for AI regulation going forward?
Lord Vallance of Balham (Lab)
The noble Viscount makes an important point. This will be global; there is no question about it. Therefore, there needs to be some degree of interoperability between different regions in terms of the regulations put in place. At the moment, as I said, of the two most advanced, the US is the biggest AI nation in the world and is developing a regulation along similar lines to ours, we believe. The EU is of course the most regulated place in the world for AI and we need to work out, in consultation over the next months, how to make sure that we work out where the areas of interoperability will lie.
Framework Convention on Artificial Intelligence
Viscount Camrose Excerpts(1 year, 4 months ago)
Lords ChamberRead Full debate Read Hansard Text Watch Debate Read Debate Ministerial Extracts
Lord Vallance of Balham (Lab)
The convention sets out activities in the life cycle of AI systems, and they should not infringe our values of human rights, democratic processes and the effectiveness of democratic institutions or the rule of law. It applies to the public sector, to the public sector when using the private sector, and there is an obligation to consider how private sector activities can be taken into account when this is implemented in a national framework.
Viscount Camrose (Con)
My Lords, international bodies currently working on AI safety and regulation include the UN, UNESCO, the ITU, the G7, the G20 and the GPI, among several others. Do the Government agree that although each of these groups is crucial and has a very important role to play in creating safe and well-regulated AI globally, they will be successful only to the extent that they are effectively co-ordinated? If so, what steps are the Government taking to bring that about?
Lord Vallance of Balham (Lab)
We are in active discussion with all those partners. As we consider an AI Act, we will work closely with partners in the US and elsewhere and apply it only to the limited number of companies at the very forefront of AI, to those models of tomorrow which carry particular risk and, again, where guard-rails have been asked for.