Author: Dr Troy Neilson

Executive Summary

The Urgent Crisis of Digital Cultural Colonisation

Australia stands at a critical juncture in its digital future, facing an unprecedented threat to its cultural and linguistic identity that most citizens don’t yet recognise. As large language models rapidly become the invisible mediators of human communication, embedded in everything from email composition to legal drafting, from educational assistance to creative writing, they carry with them a profound and insidious form of cultural imperialism. These models, trained overwhelmingly on American data, reflecting predominantly American cultural assumptions and linguistic patterns, are quietly but persistently reshaping how Australians express themselves, think about problems, and understand their world.

This is not a distant threat or theoretical concern. Every single day, thousands of Australian students submit assignments written with the assistance of AI that subtly guides them toward American spelling, American phrasing, American conceptual frameworks. Every day, Australian lawyers draft contracts using AI trained on American legal precedents, unconsciously adopting American legal thinking. Every day, Australian creative writers seek inspiration from AI models that suggest American narrative structures, American cultural references, American ways of resolving conflicts and expressing emotions. Each interaction seems insignificant in isolation, but collectively they represent a massive, ongoing erosion of Australian linguistic and cultural distinctiveness.

The Technical Achievement: Proof of Possibility

Our research presents both a demonstration of this threat’s reality and a practical path toward digital sovereignty. Through focused effort with deliberately constrained resources, we successfully adapted a large language model to better understand and generate Australian English. This technical achievement, whilst significant in its own right, serves a larger purpose: proving that meaningful cultural adaptation of AI systems is both technically feasible and economically achievable without Silicon Valley-scale resources.

The results exceeded our initial expectations. Through systematic testing and practical validation, we documented clear evidence of the model’s acquisition of Australian linguistic patterns, cultural knowledge, and conceptual frameworks. The model didn’t just learn to recognise “servo” and “arvo”; it began to demonstrate the underlying thought patterns that make Australian English distinctive: the preference for concrete over abstract expression, the instinctive deflation of pretension through humour, the particular way Australians conceptualise social relationships and institutional structures.

The Fundamental Limitation: Why Adaptation Isn’t Enough

However, our success paradoxically revealed a profound limitation that shapes the rest of this discussion. While we successfully modified the model’s behaviour, we were still fundamentally working with a base model trained primarily on non-Australian data. The deep architectural assumptions, the tokenisation patterns optimised for American English, the attention mechanisms trained on American document structures, all remain at the core of the system. This limitation is not merely technical but philosophical. Fine-tuning, no matter how successful, represents a form of cultural overlay rather than cultural foundation. We’ve applied an Australian veneer to an American core, achieved Australian fluency within an American framework, created Australian expression through American architecture. While this represents meaningful progress and offers immediate practical benefits, it cannot be the endpoint of Australia’s digital sovereignty journey.

The Imperative: Foundation Models for True Sovereignty

This research therefore serves a dual purpose: demonstrating what’s immediately possible while advocating for what’s ultimately necessary. Australia needs its own foundation models, trained from scratch on Australian data, designed with Australian architectural choices, optimised for Australian languages (including Indigenous languages), and controlled by Australians. This isn’t technological nationalism or digital protectionism; it’s recognition that in an age where AI increasingly mediates human thought and expression, cultural sovereignty requires computational sovereignty.

The path forward requires coordinated action across multiple sectors. Government must recognise AI as critical cultural infrastructure deserving of national investment comparable to other major infrastructure projects. Universities and research institutions need to prioritise cultural AI research and develop the human capital necessary for sovereign AI development. Industry must recognise the competitive advantages of culturally-aligned AI and invest accordingly. Communities must actively participate in data collection, validation, and governance of these systems.

The Global Significance: A Blueprint for Digital Diversity

Perhaps most importantly, this research provides a reproducible blueprint for any nation or cultural group seeking to preserve its voice in the algorithmic age. We’ve demonstrated that meaningful cultural AI development doesn’t require the resources of tech giants. From small nations like New Zealand and Ireland to Indigenous communities worldwide, from linguistic minorities to regional cultures, our approach offers a practical path toward digital sovereignty.

The alternative – accepting the current trajectory toward a handful of massive, culturally homogenised models dominating global communication – represents a form of algorithmic colonisation that will reshape human cultural diversity within a generation. The choice we make now about cultural AI will reverberate for decades, determining whether future generations inherit a rich ecosystem of culturally-diverse AI systems or a flattened, homogenised digital monoculture where all human expression is filtered through a single, predominantly American lens.

Part I: The Silent Erosion of Our Digital Identity

The Invisible Revolution Reshaping Human Expression

We are living through perhaps the most significant transformation in human communication since the invention of writing itself, yet most people remain entirely unaware of its scope and implications. Large language models have moved from research curiosities to ubiquitous infrastructure with breathtaking speed. They now mediate millions of human interactions daily: composing emails, drafting reports, writing essays, creating marketing copy, generating creative content, assisting with coding, answering questions, providing tutoring, offering therapy, and countless other applications that continue to emerge.

But these models are not neutral tools. They are cultural artefacts, trained on specific data, embedding particular worldviews, optimised for certain forms of expression. And overwhelmingly, they are American. Not deliberately or maliciously American, but American by default, American by data availability, American by the simple fact that most of the internet’s English-language content originates from or caters to American audiences.

This American dominance in training data creates models with deeply embedded biases that go far beyond simple vocabulary differences. These models have learned to think in American. They default to American legal frameworks when discussing rights and obligations. They assume American educational structures when discussing academic progression. They apply American business practices when discussing corporate organisation. They reference American history as common knowledge while treating other nations’ histories as foreign and requiring explanation. They embed American social norms around politeness, directness, appropriate topics of conversation, humour styles, and interpersonal relationships.

The Mechanics of Algorithmic Colonisation

The process by which these models reshape human expression is subtle but pervasive. Unlike traditional forms of cultural influence, which operate through conscious consumption and choice, AI-mediated influence operates below the threshold of awareness. When an Australian student uses an AI writing assistant, they don’t consciously choose to adopt American expression; the model simply makes American phrasing seem more natural, more fluent, more “correct.”

This happens through multiple mechanisms. First, there’s direct suggestion: the model proposes American spellings, American idioms, American sentence structures. Users, especially younger ones or those less confident in their writing, accept these suggestions without questioning their cultural origin. Second, there’s reinforcement learning: when users accept AI suggestions, they internalise these patterns, gradually shifting their own expression toward the model’s preferences.

Third, there’s standardisation pressure: as AI-assisted writing becomes ubiquitous, it creates a new standard of “professional” or “educated” expression that happens to be distinctly American.

The impact extends beyond individual expression to institutional communication. Australian businesses adopting AI tools find their corporate communications drifting toward American business English. Australian legal firms using AI for contract drafting unconsciously adopt American legal terminology and conceptual frameworks. Australian educational institutions using AI for curriculum development find American educational assumptions embedded in their materials. Australian government agencies using AI for citizen communication find their messages losing their distinctive Australian character.

The Australian Context: Unique Challenges and Opportunities

Australia faces a particularly complex challenge in this landscape. As an English-speaking nation, we cannot simply translate models as non-English-speaking countries might do. We must actively assert our distinctiveness within the anglophone AI ecosystem, competing for recognition against the overwhelming dominance of American English in training data and model development.

Yet our English is profoundly different from American English in ways that go far beyond spelling conventions or vocabulary choices. Australian English carries within it the traces of our unique history: the Indigenous substrate that gave us thousands of place names and concepts without European equivalent, the convict origins that created our distinctive relationship with authority, the gold rush era that brought Chinese, German, Italian, and Greek influences, the post-war immigration that added layers of Mediterranean and Asian expression, the ongoing multiculturalism that continues to evolve our language.

More fundamentally, Australian English embodies a distinctive worldview shaped by our geography, our history, our social structures. The harsh landscape that taught us understated resilience. The egalitarian ethos that makes us instinctively deflate pretension. The isolation that created our particular form of self-reliant community. The Indigenous influence that gave us different ways of understanding land, time, and relationships. These aren’t just cultural decorations on a universal English; they’re fundamental differences in how we conceptualise and express our Australian reality.

The Stakes: What We Stand to Lose

The stakes of this digital transformation extend far beyond language preservation. At risk is nothing less than the diversity of human thought and expression in the digital age. Languages and dialects aren’t just different ways of saying the same things; they’re different ways of thinking, different ways of understanding reality, different ways of solving problems. When we lose linguistic diversity, we lose cognitive diversity. When we homogenise expression, we homogenise thought.

Consider how different cultures conceptualise time, space, causation, identity, relationship. These differences, encoded in language, offer different perspectives on universal human challenges and experiences. Australian Indigenous languages, for instance, include concepts of time and place that Western physics is only now beginning to appreciate. The Australian tendency toward pragmatic scepticism offers a valuable counterweight to American optimism in assessing technological risks. The Australian approach to social welfare, encoded in how we discuss fairness and obligation, provides alternatives to American individualism.

If we allow AI to homogenise these differences into a single, predominantly American mode of expression, we impoverish not just Australian culture but human culture. We lose alternative ways of thinking that might prove crucial in addressing future challenges. We sacrifice the cognitive diversity that has always been humanity’s greatest problem-solving asset.

Moreover, we risk creating a form of digital dependency that extends beyond culture to economics and security. If Australian institutions become dependent on foreign AI systems for critical functions, we become vulnerable to external control. The providers of these systems can change terms, restrict access, or prioritise their domestic users at any time. We’ve seen this play out in other technological dependencies; AI dependency would be far more consequential.

Part II: The Experiment – Reclaiming Linguistic Territory

Understanding the Depth of the Challenge

Before we go further, it’s worth understanding the magnitude of the challenge we faced. Modern large language models represent some of the most complex artefacts humans have ever created. Trained on massive amounts of text, using significant computational resources, they encode patterns and relationships across virtually all of human written knowledge. The idea that we could meaningfully shift such a system’s cultural orientation might seem impossibly ambitious.

The challenge goes beyond mere scale. These models don’t simply memorise text; they build complex, multidimensional representations of concepts, relationships, and patterns. They develop what researchers call “emergent capabilities”, or abilities that weren’t explicitly programmed but arise from the complex interactions of billions of parameters. They form internal representations that we don’t fully understand, using mechanisms we can observe but not entirely explain.

The American bias in these models isn’t a simple overlay that can be easily removed. It’s woven into the very fabric of how they process language, affecting tokenisation patterns, attention mechanisms, semantic embeddings, syntactic preferences, and pragmatic assumptions.

Our hypothesis was that despite these deep structural biases, targeted adaptation with high-quality Australian data could shift these patterns enough to create meaningful cultural alignment. We weren’t trying to eliminate American influence entirely – that would require training from scratch. Instead, we aimed to prove that even within the constraints of an American-trained base model, we could achieve sufficient Australian alignment to serve Australian users effectively.

The Process and Results

Through extensive experimentation and careful optimisation, we successfully adapted the model to better understand and generate Australian English. The transformation was remarkable – the model demonstrated clear acquisition of Australian linguistic patterns, cultural knowledge, and conceptual frameworks.

Part III: The Findings – Systematic Validation Through Practical Testing

Reframing “Anecdotal”: The Value of Practical Validation

The characterisation of our testing as “anecdotal” requires careful reconsideration. While we didn’t conduct formal benchmarking against standardised metrics, primarily because no such metrics exist for cultural alignment in an Australian context, our validation was both systematic and comprehensive. We prefer the term “practical validation” as it better captures the rigorous, albeit qualitative, nature of our evaluation process.

Traditional NLP benchmarks measure capabilities like question answering, sentiment analysis, or named entity recognition. These metrics, while valuable for assessing general capability, tell us nothing about cultural alignment. A model could achieve perfect scores on SuperGLUE while completely failing to understand Australian English. We needed different approaches to validation, ones that directly assessed whether our training had achieved a meaningful cultural shift.

Vernacular Mastery: Beyond Vocabulary Lists

The most immediately observable change was in the model’s handling of Australian vernacular. But this went far beyond simple vocabulary recognition. The model demonstrated understanding of the contextual usage, register appropriateness, and cultural connotations of Australian expressions.

When tested with sentences like “I’m flat out like a lizard drinking this arvo, but I’ll swing by the bottle-o on the way home from the servo,” it didn’t just parse the Australian terms; it understood the entire communicative context. It recognised this as someone expressing they’re very busy this afternoon but will stop at the bottle shop on the way home from the service station. More importantly, it understood the register (casual, friendly) and the implied social context.

The model could also generate appropriate responses in the same register, demonstrating genuine understanding of conversational flow in Australian English. It showed awareness of regional variations, understanding that “footy” means different codes in different states, recognising regional variants for swimwear, and understanding contextual clues about location.

Institutional Knowledge: Understanding Australian Systems

Beyond language, the model demonstrated comprehension of Australian institutional structures, legal frameworks, and social systems. This knowledge wasn’t superficial memorisation but showed genuine understanding of relationships and implications.

The model correctly explained the Westminster system, the role of the Governor-General, preferential voting, and the relationship between federal, state, and local government powers. It understood the distinction between barristers and solicitors, the hierarchy of courts, and could apply Australian legal principles rather than American ones when presented with legal scenarios.

The model correctly understood Australian educational structures, the progression through Year levels, and the distinction between universities and TAFEs. This knowledge proved particularly valuable when asked to assist with educational content or advice, as it could provide contextually appropriate information rather than defaulting to American grade levels and SAT scores.

Cultural Conceptual Frameworks: The Deep Structure

Perhaps the most significant evidence of successful cultural adaptation came from the model’s demonstration of distinctively Australian conceptual frameworks. These go beyond vocabulary or factual knowledge to represent different ways of understanding and interpreting the world.

The model not only understood the term “tall poppy syndrome” but demonstrated the conceptual framework it represents. When discussing success or achievement, it would often include caveats about not getting “too big for your boots” or the importance of staying grounded. This wasn’t explicit moralising but a subtle reflection of the Australian cultural tension between celebrating success and maintaining egalitarianism.

The Australian concept of mateship appeared throughout the model’s outputs, understood as involving loyalty without sentimentality, humour as emotional regulation, and shared experience as a bonding mechanism. The model demonstrated the distinctive Australian relationship with authority – respectful but not deferential, compliant but not submissive.

Stylistic Preferences: The Rhythm of Australian English

Beyond vocabulary and concepts, the model demonstrated adoption of stylistic patterns characteristic of Australian English. It showed preference for understatement where American English might exaggerate, describing serious situations as “a bit of a problem” rather than “a disaster.” It demonstrated preference for concrete, practical language over abstract conceptualisation, and naturally incorporated humour, particularly dry, self-deprecating, or ironic humour, into various contexts.

Emergent Behaviours: Unexpected Discoveries

Beyond our targeted testing, we observed several emergent behaviours that suggested deeper cultural adaptation. When explaining Australian concepts to non-Australian audiences, the model would provide culturally appropriate analogies. In some cases, the model created novel expressions that felt authentically Australian while being technically new, suggesting it had internalised generative rules rather than simply memorising fixed expressions.

Part IV: The Broader Implications – From Achievement to Imperative

The Paradox of Success: Why Good Isn’t Good Enough

The success of this adaptation creates a paradox that shapes the remainder of this discussion. We’ve proven that meaningful cultural adaptation is possible with modest resources. The model genuinely acquired Australian linguistic and cultural patterns. Users could interact with it more naturally, finding their expressions understood and their contexts recognised. By any practical measure, the experiment succeeded.

Yet this success illuminates a fundamental limitation that no amount of fine-tuning can overcome. We’ve achieved “Australian fluency within an American framework.” But the framework itself, the architectural assumptions, the tokenisation patterns, the attention mechanisms, the fundamental ways the model processes and understands language, remain irreducibly American.

Consider an analogy: We’ve successfully taught an American to speak Australian English, understand Australian culture, and navigate Australian society. They’ve become fluent, knowledgeable, even intuitive about Australian ways. But at their core, they remain American, seeing Australia through American eyes, interpreting Australian experience through American frameworks, understanding Australian concepts by reference to American baselines. This isn’t a failure of education or adaptation; it’s an irreducible fact of foundation.

The Foundation Model Imperative: Building Rather Than Adapting

These limitations point to an uncomfortable but unavoidable conclusion: true digital sovereignty requires foundation models trained from scratch on Australian data. This isn’t technological nationalism or digital protectionism; it’s recognition that in an age where AI mediates increasing aspects of human thought and expression, cultural sovereignty requires computational sovereignty from the ground up.

A truly Australian foundation model would make Australian-appropriate decisions at every architectural level, from tokenisation design to attention patterns, from positional encodings to model scaling decisions. Training a foundation model entirely on Australian data ensures complete data sovereignty, with Australian content never leaving Australian control.

Dependence on foreign AI models creates economic vulnerabilities through pricing control, access restrictions, capability limitations, and value extraction. Every dollar spent on foreign AI is value extracted from the Australian economy. Building sovereign capability creates Australian jobs, develops Australian expertise, and keeps value within Australia.

The Technical Requirements: Scaling from Proof to Product

Moving from our experimental adaptation to true foundation models requires significant but achievable scaling. While the computational and data requirements are substantial, they are within Australia’s capability. We spend billions on traditional infrastructure and defence. A sovereign AI capability is arguably more critical to our future security and prosperity than many traditional investments.

Foundation model development requires interdisciplinary teams of technical experts, domain experts, and support roles. Australia has this expertise; they’re currently working for foreign tech companies, universities with limited resources, or in adjacent fields. We need to bring them together with appropriate resources and clear mandates.

International Context: The Sovereignty Race Accelerates

While Australia debates, other nations act. France’s participation in the BLOOM project demonstrated that multilingual foundation models are achievable through international collaboration. The UAE’s Falcon models prove that small nations can develop competitive AI. China’s development of multiple foundation models demonstrates the strategic importance nations place on AI sovereignty. India’s efforts show that linguistic complexity isn’t an insurmountable barrier to sovereign AI development.

Part V: The Philosophical Dimensions – What Is Cultural AI? Beyond Technology: The Nature of Cultural Intelligence

As we near the conclusion of this discussion, it’s worth examining the deeper philosophical questions our work raises. What does it mean for an artificial system to be “culturally aligned”? Can a mathematical model truly embody cultural values, or is it merely performing sophisticated mimicry? And perhaps most fundamentally, what is the relationship between language, culture, and thought in artificial systems?

These aren’t merely academic questions. How we answer them shapes how we approach cultural AI development, what we expect from these systems, and how we integrate them into society.

Language as Worldview

The hypothesis that language shapes thought has been debated for decades in human cognition. Our work with AI systems provides a unique new lens through which to examine these questions. When we successfully shifted our model toward Australian English, what exactly changed?

At the surface level, vocabulary and spelling shifted. But our testing revealed deeper changes: the model began exhibiting different preferences for concrete versus abstract expression, different patterns of social interaction, different approaches to authority and hierarchy. These weren’t explicitly programmed but emerged from exposure to Australian language patterns.

This suggests that language models, in learning language, also learn the conceptual structures embedded within it. Australian English, with its preference for understatement, encodes a different relationship to achievement and success than American English with its celebration of individual accomplishment. When the model learns these linguistic patterns, it appears to also absorb these conceptual frameworks.

The Question of Authenticity

This raises challenging questions about authenticity in artificial systems. When our adapted model generates text that sounds distinctively Australian, uses appropriate cultural references, and embodies Australian values, is it being “authentically” Australian? Or is it performing a sophisticated but ultimately hollow imitation?

We find the pragmatic approach more compelling: functional alignment matters more than metaphysical authenticity. If the model serves Australian users effectively, preserves and promotes Australian expression, and helps maintain Australian cultural distinctiveness, then questions of its inner experience are less relevant.

Cultural AI as Cultural Preservation

One of the most profound implications of our work is the role AI might play in cultural preservation. As these systems become more prevalent, they don’t just reflect culture; they actively shape it. A model trained on historical Australian literature preserves not just the words but the patterns of thought, narrative structures, and cultural values embedded in those texts.

This preservation goes beyond static archiving. When the model generates new text in these patterns, it’s actively continuing cultural traditions, creating new expressions within established frameworks. It’s participating in the living evolution of culture rather than merely documenting its past forms.

The Ethics of Cultural AI

The development of culturally-aligned AI systems raises complex ethical considerations. Who gets to decide what counts as “Australian” culture for AI training? How do we balance cultural evolution versus preservation? What rights do communities have over AI systems trained on their cultural materials?

Future development must actively address representation issues, ensuring Indigenous control over Indigenous knowledge and building systems that serve all Australians, not just the digitally privileged. Perhaps most fundamentally, our work suggests that access to culturally-aligned AI should be considered a digital right.

Global Implications: Toward Digital Biodiversity

Our research has implications beyond Australia, suggesting a new model for global AI development that preserves and promotes cultural diversity. This vision faces significant resistance from economic forces, technical momentum, resource disparities, and coordination challenges.

Despite these challenges, the alternative of accepting algorithmic homogenisation, is worse. The loss of linguistic and cultural diversity in the digital realm would impoverish human thought and expression. The concentration of AI power in a few hands would create unprecedented control over human communication. The encoding of a single culture’s values in global AI would constitute a form of digital imperialism.

Conclusion: The Voice We Choose to Preserve

The Journey: From Question to Imperative

This work began with a straightforward technical question: Could we meaningfully shift a large language model toward Australian linguistic and cultural patterns? Through extensive experimentation and testing, we’ve answered that question definitively: yes, meaningful cultural adaptation is both technically feasible and practically valuable.

But in answering our initial question, we’ve uncovered a much larger imperative. Fine-tuning, while valuable, operates within fundamental constraints. True digital sovereignty requires foundation models built from scratch on Australian data, embedding Australian perspectives at every level.

The Achievement: Proof and Blueprint

Our technical achievement stands as both practical tool and existential proof. The adapted model will have immediate applications in educational settings, legal practice, creative writing, and business communication. More importantly, our success proves that cultural AI is achievable, technical barriers are surmountable, resource requirements are within reach, the methodology is reproducible and scalable, and the impact on users is meaningful.

This proof comes at a crucial moment. Around the world, nations are recognising AI’s strategic importance. Those that develop sovereign capabilities will shape the future; those that don’t will have their futures shaped for them. Australia stands at a crossroads: we can choose to be AI producers or consumers, digital sovereigns or digital dependents, cultural preservers or cultural casualties.

The Imperative: Why Australia Must Act Now

The window for establishing Australian AI sovereignty is closing rapidly. Every day of delay has compound consequences through entrenchment effects, data drainage, cultural erosion, and competitive disadvantage.

The cost of delay isn’t neutral; it’s actively harmful. But the opportunity cost may be even greater. By acting now, Australia could establish leadership in cultural AI, build competitive advantages in AI-enabled industries, create high-value employment, preserve and promote our cultural heritage digitally, and shape international AI governance from a position of strength.

The Vision: Australia’s Digital Future

We envision an Australia where AI enhances rather than erodes our cultural distinctiveness, in education, business, government, creative industries, and daily life. This vision isn’t utopian; it’s achievable with current technology and reasonable investment. The barriers aren’t technical but political, economic, and social.

The Call to Action: From Research to Reality

This research should catalyse a national conversation about Australia’s AI future. Every sector of society has a role: government must recognise sovereign AI’s strategic importance, universities must prioritise cultural AI research, industry must recognise sovereign AI’s pragmatic benefits, communities must engage actively in AI development, and individuals must become informed participants in Australia’s AI future.

The Legacy: What We Leave Behind

Ultimately, this work is about legacy, what we leave for future generations of Australians. The decisions we make now about AI will reverberate for decades, perhaps centuries. If we fail to act, we leave future Australians a diminished inheritance. But if we succeed in establishing sovereign AI capability, we leave a richer legacy: AI systems that preserve and evolve Australian culture, that amplify rather than diminish Australian voices, that ensure Australia remains not just a geographic location but a distinctive perspective in global conversation.

The choice is ours, and it must be made now. Not next year, not after further study, not when the technology is “more mature.” Now, while the window remains open, while alternatives remain possible, while Australian culture retains enough distinctiveness to be preserved.

The Final Word: It’s Not Too Much to Ask

In the end, this research is about something both simple and profound: ensuring that when Australians interact with AI, it understands not just their words but their world. That’s not too much to ask. In fact, in the age of artificial intelligence, it’s the absolute minimum we should demand.

We’ve proven it’s technically possible. We’ve demonstrated it’s economically feasible. We’ve shown it’s practically valuable. We’ve established it’s culturally necessary. All that remains is the national will to make it happen.

The future is being written in code today. The question is: will it be written in our voice?

The answer lies not in Silicon Valley or Beijing but in the choices we make as Australians. We can choose to be passive consumers of foreign AI, accepting whatever cultural assumptions and linguistic patterns they embed. Or we can choose to be active creators of Australian AI, ensuring our voice not only survives but thrives in the digital age.

We choose creation. We choose sovereignty. We choose to preserve and evolve the distinctive Australian voice that makes us who we are.

The time for discussion has passed. The time for action is now.