Michelle Donelan, the secretary of technology, believes the safety summit will assist identify technology’s early warning indications.
Ministers have cautioned that harmful artificial intelligence systems require a “smoke alarm” to prevent serious threats including mass casualties, cyberattacks, and out-of-control AI development.
Michelle Donelan, the technology minister, expressed her hope that a safety conference to be held in the UK soon would contribute to the creation of a system of early warning where tech companies search for problems in the artificial intelligence products they are developing and are prepared to address them.
We need to set up something akin to a smoke alarm, she continued, so that businesses may not only look for threats but also respond to them.
Donelan said there were “incredible opportunities” with AI, but “we will only really grasp those opportunities if we’re gripping the risks” when he spoke at Bletchley Park, the location of the two-day meeting in November.
The government announced on Monday that the meeting’s main topics will be the exploitation of AI systems to develop bioweapons or cyberattacks, as well as the incapacity to manage the most cutting-edge technologies.
It is claimed that Rishi Sunak firmly believes that as tech companies use more computer power, technological advancements, and increasing investment to create ever-more potent models, there is a limited amount of time left to reach a worldwide consensus on what the most significant AI threats are and how to address them.
The meeting is anticipated to identify the range of significant threats that AI systems potentially pose and steps to reduce them, rather than producing a worldwide agreement on AI development in the form of a nuclear weapon. The emphasis will be on “frontier” AI models, which are innovative machines whose power equals or exceeds that of the most sophisticated models now in use and may pose a threat to life for humans.
The meeting, according to the administration, aimed to map out the current and potential futures of AI research. Global leaders, AI businesses, academics, and civil society organizations will attend the event at Bletchley Park in Buckinghamshire, the site of codebreakers like Alan Turing during WWII.
The GPT-4 model, which drives OpenAI’s ChatGPT, is one example of a model that is “many times” more potent than those that are already in use, according to a statement from the government on Monday.
By default, these models might be rendered accessible to a wide range of people, including those who could do us harm. The abilities of these models are exceedingly difficult to forecast, perhaps even to those who are constructing them.
AI scientists worry that cutting-edge systems might escape human control. These concerns are focused on the potential for advancements in artificially intelligent general intelligence, which is the name for an AI that has human or above-human capacities for intelligence and which, in theory, might overcome any barriers put in their way.
Systems evading will be a topic of emphasis, according to the government’s announcement announcing the meeting on Monday. According to the statement, “advanced technologies that we would look for to be consistent with our principles and intentions” could provide “loss of control risks” that could arise.