Australia needs to take advantage of AI's potential, Scott Farquhar says | 7.30

Key Concepts:

• Artificial Intelligence (AI)
• Large Language Models (LLMs)
• Data Centers
• Copyright Law & Fair Use
• Transformative Technology
• Regulation of AI
• Productivity
• Reskilling/Upskilling
• Unions & Workforce Evolution

AI: Opportunities and Concerns

Scott Farquhar acknowledges potential "crazy out there scenarios" with AI but believes the chances are low, focusing on the positives: increased productivity and more enjoyable jobs. He draws a parallel to the introduction of cars, highlighting the need to "ameliate the negatives" while embracing the positives of transformative technologies like AI.

What is AI?

Farquhar provides examples of AI in everyday life:

• Navigation apps (Google Maps, Waze, Apple Maps) routing users around traffic.
• Spotify recommending new songs.
• Chatbots and Large Language Models (LLMs) like ChatGPT.

He notes that LLMs have ushered in a significant shift in how we perceive AI.

Australia's Role in the AI Ecosystem

Farquhar advocates for Australia to move beyond being a passive consumer of AI and become a creator and exporter of technology. He identifies a significant opportunity for Australia to become a regional data center provider.

• Data Centers: He describes data centers as collections of computers in cooled environments connected to the internet. A large data center can represent a $5 billion investment over 10 years, including construction, chips, power, and cooling.
• Competitive Advantages: Farquhar believes Australia can compete due to its system of law, cheap and abundant green power, and expertise in building large projects.
• Regional Demand: He argues that there will be a need for regional data centers serving customers in Southeast Asia (e.g., Singapore, Malaysia, Philippines), citing the high usage of ChatGPT in Indonesia and Vietnam as examples.

Regulation of AI

Farquhar expresses reservations about a single, overarching "AI act" due to the broad nature of AI applications. He suggests a "wait and see" approach for harms that aren't immediately obvious, arguing that existing regulations cover many current concerns.

• Tech Company Responsibility: He acknowledges the need for laws and rules to guide businesses in the AI space, echoing the sentiment of a former Microsoft executive.

Copyright Law and Fair Use

Farquhar proposes changes to Australia's copyright law to allow exemptions for AI companies to mine Australian content for text and data for free.

• Fair Use: He argues that Australia's copyright law lacks a clear "fair use" exemption for transformative technologies like AI, unlike the US. He believes this hinders investment in AI in Australia.
• Transformative Use vs. Theft: He distinguishes between using AI to directly copy an artist's style (probably not fair use) and using AI collaboratively to create something new (potentially fair use).
• Benefits vs. Rights: He suggests that the benefits of LLMs outweigh the concerns of artists regarding the use of their material for training AI models.
• Analogy to iPods: He compares the current situation to the early days of iPods, when it was technically illegal to transfer music from a legally owned CD to an iPod. Copyright law was later updated to permit this.
• Midnight Oil Example: He uses the example of the band Midnight Oil to illustrate the debate, suggesting that they shouldn't have the right to prevent their catalog from being used to create new AI models if the use is transformative. He argues that this is similar to news sites quoting articles under "fair dealing" principles.

Workforce Evolution and Unions

Farquhar discusses the role of unions in the context of AI and its impact on the workforce.

• Union Involvement: He believes unions should be involved in workplace evolution related to AI, including training for new jobs and managing job displacement.
• Veto Power: He doesn't support unions having a veto over the introduction of AI in the workplace, but emphasizes collaboration.
• Retraining: He questions whether companies should be responsible for retraining workers displaced by AI for entirely new industries (e.g., retraining call center workers to build data centers). He suggests exploring stronger safety nets for displaced workers.
• Industrial Revolution Parallel: He acknowledges concerns about the potential for AI to disproportionately harm the poorest workers, drawing a parallel to the Industrial Revolution. He emphasizes the need to actively pursue opportunities related to AI to ensure that Australia benefits as a nation.

Treasurer's Productivity Summit

Farquhar outlines his key messages for the Treasurer's Productivity Summit:

1. Data Centers: Emphasizing the importance of building data centers in Australia.
2. Tech Trades: Advocating for shorter, more focused training programs ("tech trades") to rapidly reskill workers for the AI era (e.g., training electricians for data centers in 6-12 months instead of 4 years).
3. Government Adoption of AI: Highlighting the potential for government to improve efficiency and public services by adopting AI internally (e.g., streamlining development application processes).

Personal Reflections

Farquhar reflects on his time at Atlassian, noting that his skills in setting a vision and getting things done are transferable to other areas. He discusses his involvement with the Tech Council of Australia and his increased focus on philanthropy.

Conclusion

Farquhar is optimistic about the potential benefits of AI for Australia, but emphasizes the need for proactive measures to ensure that the nation captures those benefits. He advocates for strategic investments in data centers, streamlined training programs, and government adoption of AI. He acknowledges the concerns surrounding copyright and workforce displacement, but believes that these challenges can be addressed through thoughtful regulation, collaboration, and a focus on creating new opportunities.