In our submission we provided feedback to two terms of reference of the Select Committee.

b. risk and harms

We shared our feedback that the key risk is users trusting the outputs of AI and generative AI. Such users believe all outputs are well researched, reasoned, and true and do not apply a human sense check.

Key to mitigating this risk is education. We need to raise awareness that tools built from generative AI do not ‘think’, in fact, such tools simply generate a response based on complex algorithms applied against the sequence of words in a prompt.

Coupled with educating users is the need to hold the developers and product owners accountable for their products. We seek for regulation on developers to nominate an accountable person for harm arising from toxic content generated by their products.

g. environmental impacts of AI technologies

The exponential growth in demand for and use of tools using generative AI means more devices are needed to store data, an exponential increase in the data there is to interrogate meaning it takes longer to train a model which creates a need for more graphic processing units.  For example, the parameters to train data for the original ChatGPT took around 100 million parameters in 2018, this has exponentially increased to 500 billion in 2023 for Google’s PaLM model.

We recognise that growth in the use of AI is inevitable and consider government has a role to hold developers and designers accountable for their impact on the global environment.

We seek for regulations to require developers of foundation models, large language models and generative AI tools to publish their source of energy and their greenhouse gas emissions.

Conclusion

We look to the government to provide a safety net and guardrails so the tools available to all Australians are safe to use and have been developed in a responsible and environmentally sustainable way.

We appreciate our members sharing how they are using AI and generative AI  tools in thier practices which helped inform our submission.