AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 511

Ms Laura Leighton

Submitter group: Individuals.

Position labels are not shown for individual submitters.

An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.

Open in the interactive guide

Summary

Read the original submission for the full text.

Argues for caution about AI despite its potential economic, scientific and cultural benefits, focusing on catastrophic and loss-of-control risks, biological and chemical weapon misuse, and concentrated economic power. Recommends considering extreme-consequence risks, adequately resourcing the AI Safety Institute with bipartisan commitment, and international coordination. The scientist writes in a private capacity.

Key points

Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.

  1. Argues that any policy framework addressing the capabilities of current AI systems will almost definitionally be obsolete by the time it can be enacted.

    The first is that any policy framework that aims to address the possible opportunities and harms posed by the capabilities of extant AI systems will almost definitionally be obsolete by the time it can be enacted.

  2. Argues that the Australian government must acknowledge catastrophic risks from highly capable AI, drawing an analogy with under-preparation for pandemics.

    While many governments acknowledged pandemics as a theoretically serious threat to the normal functioning of society, few undertook an appropriate level of preparation, and the result was a devastating disruption to the economy and society. We should learn from our mistakes. The Australian government must acknowledge the catastrophic societal risks posed by highly capable AI systems.

  3. Says a cited study describes a drug-discovery AI system designing 40,000 lethal molecules in under six hours.

    For instance, an AI system designed to find new drugs instead designed 40,000 lethal molecules in less than 6 hours[11], including the de novo rediscovery of nerve agents previously used as chemical weapons; some of these molecules could be synthesised by a competent garage chemist.

  4. Argues that AI is a force multiplier for beneficial and harmful uses.

    This feature of AI is a force multiplier – both for those wanting to do good (I have benefited enormously from AI chatbots in my own research) but unfortunately also for those wanting to engage in hacking, scamming, drug manufacture, weapons manufacture, etc etc.

  5. Says discussion of AI's economic benefits gives too little focus to power concentration and economic inequality driven by AI.

    I also wish to share my view that in the conversations about the potential benefits of AI to productivity and the economy, I am dismayed to see less focus on power concentration and economic inequality being driven by AI.

Recommendations

Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.

  1. Acknowledge the catastrophic societal risks posed by highly capable AI systems.

    We should learn from our mistakes. The Australian government must acknowledge the catastrophic societal risks posed by highly capable AI systems.

  2. Account for low-probability events with extreme consequences in regulatory assessments.

    Low-probability events with extreme consequences must be taken into account by regulatory assessments such as this one

  3. Consider the potential economic impact of the consolidation of wealth and power in a few large, mostly US-based AI companies, and of the AI 'bubble bursting'.

    Australia needs to consider the potential economic impact of the consolidation of wealth and power in the hands of a few large, mostly US-based companies, as well as the potential impact if ‘the bubble bursts’.

  4. Staff and resource the AI Safety Institute to test new models, support regulators responding to emerging risks and shape safe AI development in Australia’s interests, with bipartisan commitment.

    The announcement of an Australian AI Safety Institute is an extremely important step forward. If appropriately staffed and resourced to meet its 3 stated goals (testing of new models, supporting regulators in responding to emerging risks, and shaping safe AI development in Australia’s interests), this institute has the potential to become a world leader in critical areas of technical AI safety research, in ways legible to Australia’s allies and to frontier AI companies. I fervently hope that bipartisan commitment to this will be achieved.

  5. Coordinate with other middle powers to reduce AI risks and mitigate power concentration.

    As a middle power which is well-respected on the international stage and maintains a broad range of economic and strategic alliances, Australia is well-positioned to coordinate with other middle powers to reduce AI risks and mitigate the impact of power concentration.

  6. Act as a leader or mediator in reaching an international agreement to verify and pace frontier AI development.

    The AI industry is open to international coordination to verify and pace frontier AI technologies to slow down the pace of development. This may buy us time to close the gap between advanced AI capabilities and our ability to control them. Australia could act as a leader or mediator to assist such an agreement to be reached.

Read the original

Open in the interactive guide