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 338

Mr Nik Samoylov

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.

A personal submission urging the Committee to pause AI development, arguing that companies underinvest in safety and that oversight such as the UK AI Security Institute is insufficient. It proposes a superintelligence ban with jail terms, safety cases, strict liability, an international treaty, and support for a US ban bill.

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 the more important question than safety testing is the overall direction of AI development, as companies race to build AI that competes with humans.

    The more important question than safety evaluations and testing is about the overall direction of the development of AI. Companies, including bigtech and heavily funded upstarts, are rushing to develop AI that will compete with humans in all economic and other spheres.

  2. Argues that AI companies, locked in a race to develop economically useful AI, underinvest in safety and security.

    As AI companies are locked in a “race” to develop AI that will be economically useful, they underinvest in safety and security, despite spending millions on PR and lobbying, including safety-washing.

  3. Says the UK AISI reports having mere days to test select aspects of new models and has no power to prevent their release or training.

    AISI reports that it has mere days to test select aspects of new models. Things like scheming, sandbagging, etc. simply cannot be elicited in such short timeframes. And even if that is the case, AISI has no power to prevent the release or training of a model.

  4. Argues that unmonitored or autonomous AI cyberattacks and bioweapons are unlikely to discriminate between nations and that cooperation on a pause is necessary.

    Unmonitored or autonomous AI cyberattacks, bioweapons, etc. similarly are unlikely to discriminate. If our side does not cooperate on the pause, it does not matter what China does, we will all lose.

  5. Proposes strict liability for AI incidents, with developers and deployers jointly liable.

    Introduce rules for strict liability for AI incidents for developers and deployers, jointly liable.

Recommendations

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

  1. Prioritise AI threats and regulation ahead of pressure to adopt, develop, host or use the technology.

    When it comes to artificial intelligence, the Australian government must first and foremost think of threats and regulations for this technology, ahead of any FOMO to adopt, develop, host, or use it.

  2. Regulate AI companies to maintain minimum safety standards, including, for some models, thorough testing and competent government approval before development/training, post-training and deployment.

    That leaves it to the government to regulate them to ensure minimum safety standards are maintained. For some models, such standards should include thorough testing and approval from competent government authorities before models are developed/trained, post-trained, deployed.

  3. Legislate a ban on developing superintelligence or AI that meets or exceeds human intelligence, with long jail terms for such developers and proscription of companies aiming to do it under counter-terrorism laws.

    Pass a law that bans the development of superintelligence or any AI that meets or exceeds human intelligence. Long jail terms for such developers. Proscription of companies that aim to do it under counter-terrorism laws.

  4. Legislate to require safety cases from developers of certain AI models.

    Pass laws that will require safety cases from developers of certain types of AI models.

  5. Introduce strict, joint liability for AI incidents for developers and deployers.

    Introduce rules for strict liability for AI incidents for developers and deployers, jointly liable.

  6. Direct the Foreign Minister urgently towards an international treaty for a global ban on developing superintelligence or any AI meeting or exceeding human intelligence.

    Direct the Foreign Minister to urgently work towards an international treaty for a global ban on the development of superintelligence or any AI that meets or exceeds human intelligence.

  7. Support the US 'Ban Artificial Superintelligence Act' initiative.

    A good starting point is the US proposal for “The Ban Artificial Superintelligence Act” by Sen. Bernie Sanders and Rep. Greg Casar: https://www.sanders.senate.gov/wp-content/uploads/Ban-Artificial-Superintelligence-Act- Release-Summary.pdf. It is an already active initiative that Australia ought to support.

Read the original

Open in the interactive guide