Submission 425
Name withheld
Submitter group: Individuals.
Position labels are not shown for individual submitters.
An AI model identified the submitter group and chose the extracts. Counts do not measure importance or merit.
Extracts from the submission, in the submitter’s own words.
Extracts
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.
This points to a looming risk: that AI will soon become fully self-recursive, improving its own code, algorithms and evaluation without any need for human oversight. Already, humans are struggling to evaluate what AI is doing, much less explain it 4. A fully self-recursive AI would carry profound dangers for critical infrastructure and societal institutions.
Given that AI capabilities will only increase in future, the potential for AI to develop more effective methods of human manipulation stands to undermine the wellbeing of Australians and the democratic and institutional systems which act in our nation’s interest.
The highly distributed nature of most frontier AI models with the potential for millions of self-replications 12 means that once advanced AI has been developed, it may be almost impossible to de-commission.
recent reports from the Google Threat Intelligence Group of attempts to use AI for the development of biological and nuclear weapons 13, as well as reports from Anthropic of at least five cases of actors using their models for biological weapons development, 14 call for the highest levels of trans-national regulation.
A program corrected obvious text-recognition errors in this quote for display.
I write as a concerned citizen to highlight concerns about the risks of AI, which I believe will likely preclude Australia from realising the majority of opportunities alluded to in Terms of Reference (a) to (e) of the Inquiry.
AI indeed holds significant promise for profound improvements in Australia’s resilience, productivity growth, economic prosperity and living standards. But it also carries profound societal and existential risks which should not be overlooked in the drive to achieve these benefits.
A program corrected obvious text-recognition errors in this quote for display.
Extracts from recommendations
Extracts chosen by an AI model, in the submitter’s own words. Listed in the same order as on its page in the interactive guide.
Consider existential AI risk as within scope for this Committee.
Regulation and trans-national safety agreements to pause frontier model development are urgently needed and should remain the priority for the Australian Government over any ill- informed roll out of AI capacity built on assumptions that economic and societal benefit will inevitably follow.
A program corrected obvious text-recognition errors in this quote for display.
Place a moratorium on recursive self-improvement of AI models.
Ensure that the forthcoming Australian AI standards legislation place as much emphasis on the safety of the models themselves as on the standards for the physical operation of data centres.
Sufficiently resource the Australian AI Safety Institute to recognise and respond to these risks with enforceable governance rather than simply guidelines or recommendations.
Read the original
In some quotes a program corrected obvious text-recognition errors for display.