Technical Deep Dive
The core of the left's AI critique is technically valid but strategically incomplete. Let's dissect the actual mechanisms at play.
Algorithmic Bias: The Technical Reality
Bias in AI systems is not a bug; it's a feature of training data. Models like GPT-4, Claude, and Gemini are trained on vast internet corpora that reflect historical inequalities. Research from the Algorithmic Justice League (led by Joy Buolamwesi) has shown that facial recognition systems from companies like Amazon Rekognition and IBM Watson have error rates as high as 34% for darker-skinned women compared to 0.8% for lighter-skinned men. The technical fix—curating balanced datasets, applying fairness constraints during training, and conducting post-hoc audits—is well understood but rarely implemented at scale.
Labor Replacement: The Economic Architecture
The left correctly identifies that AI is automating cognitive labor, not just manual tasks. However, the mechanism is not about total job elimination but about task disaggregation. For example, in customer service, AI handles tier-1 queries, reducing demand for entry-level agents. In journalism, tools like Jasper and ChatGPT generate first drafts, cutting the need for junior writers. The technical reality is that large language models (LLMs) are becoming cheaper and more capable by the month. The cost per million tokens for GPT-4o has dropped from $10 in early 2024 to $5 now, while open-source models like Meta's Llama 3.1 405B rival proprietary systems at a fraction of the cost.
| Model | Parameters | MMLU Score | Cost/1M tokens (input) | Open Source |
|---|---|---|---|---|
| GPT-4o | ~200B (est.) | 88.7 | $5.00 | No |
| Claude 3.5 Sonnet | — | 88.3 | $3.00 | No |
| Llama 3.1 405B | 405B | 87.3 | $0.30 (via Groq) | Yes |
| Mistral Large 2 | 123B | 84.0 | $2.00 | Yes |
| Gemma 2 27B | 27B | 75.2 | $0.20 | Yes |
Data Takeaway: The open-source ecosystem is collapsing the cost of AI inference. Llama 3.1 405B delivers 98% of GPT-4o's benchmark performance at 6% of the cost. This democratization is a double-edged sword: it lowers barriers for good actors but also for malicious ones. The left's silence on how to harness open-source for public good is a missed opportunity.
Power Concentration: The Infrastructure Layer
The left's critique of Big Tech's AI monopoly is accurate but lacks technical nuance. The real bottleneck is not model architecture but compute infrastructure. Training frontier models requires clusters of 10,000+ NVIDIA H100 GPUs, costing $100 million+. This creates a natural monopoly. However, initiatives like the Petastorm project (open-source data pipeline) and the Hugging Face ecosystem (over 500,000 models shared) are lowering barriers. The left could advocate for public compute banks—government-funded GPU clusters available to researchers, startups, and labor unions—but this idea is absent from their discourse.
Key Players & Case Studies
The Critics: Diagnosis Without Prescription
- Bernie Sanders: His 2023 bill to tax AI companies for worker retraining is a rare constructive proposal, but it remains a single piece of legislation. His broader rhetoric frames AI as an existential threat to jobs, without acknowledging its potential to reduce drudgery or create new categories of work.
- Cory Doctorow: His concept of "enshittification"—platforms degrading quality to extract value—is a powerful critique of Big Tech. But his proposed solutions (interoperability, data portability) are regulatory, not constructive. He does not propose alternative AI systems or funding models.
- Emily Bender: Her work on stochastic parrots (with Timnit Gebru) correctly identifies the limits of LLMs. Yet her stance is entirely cautionary. She has not engaged with the technical community on how to build safer models, only how to critique existing ones.
The Builders: Right-Wing Techno-Libertarians
- Sam Altman (OpenAI): Pushes for rapid deployment with minimal regulation, arguing that slowing down would cede leadership to China.
- Marc Andreessen (a16z): Openly advocates for "accelerationism," viewing any regulation as a threat to innovation.
- Elon Musk (xAI): While warning about AI risks, his Grok model is designed to be "anti-woke," deliberately rejecting safety guardrails.
| Player | Stance | Constructive Proposals | Impact on AI Direction |
|---|---|---|---|
| Sanders | Critical | Tax for retraining fund | Low (no technical engagement) |
| Doctorow | Critical | Interoperability mandates | Low (no alternative models) |
| Bender | Critical | Moratoriums | Low (no engineering input) |
| Altman | Accelerationist | Gradual deployment | High (shapes product roadmap) |
| Andreessen | Accelerationist | Deregulation | High (funds startups) |
| Musk | Accelerationist | Anti-woke AI | High (shifts cultural norms) |
Data Takeaway: The left's influence on actual AI development is near zero. They shape public discourse but not engineering roadmaps. The right-wing builders are the ones deciding what gets built, how it's deployed, and who benefits.
European and Asian Contrasts
- EU AI Act: A risk-based regulatory framework that bans certain uses (social scoring, real-time biometric surveillance) while allowing innovation. It was shaped by progressive policymakers who engaged with technologists.
- Singapore's AI Verify: A voluntary governance framework that provides testing tools for companies. It was developed by the government in collaboration with industry, not in opposition.
- South Korea's AI Hub: A public-private partnership that provides compute resources to startups and researchers. It explicitly aims to democratize AI access.
Industry Impact & Market Dynamics
The left's absence is reshaping the market in three ways:
1. Regulatory Vacuum: Without a credible progressive voice, regulation is being written by industry lobbyists. The US has no federal AI law; instead, we have voluntary commitments from companies. This is a direct consequence of the left's refusal to engage in the legislative process constructively.
2. Labor Market Polarization: AI is creating a bifurcated job market. High-skill workers (prompt engineers, ML engineers) see wage increases; low-skill workers (data entry, customer service) see wage stagnation. The left could advocate for universal basic income or sectoral bargaining, but these ideas are not tied to AI policy.
3. Open Source vs. Proprietary: The open-source movement is the left's natural ally—it democratizes access, reduces monopoly power, and enables community oversight. Yet the left has not embraced it. The Hugging Face platform, which hosts over 500,000 models, is a potential vehicle for progressive AI, but it remains apolitical.
| Metric | 2023 | 2024 | 2025 (est.) |
|---|---|---|---|
| Global AI market size | $142B | $196B | $267B |
| US AI regulation bills passed | 0 | 0 | 0 |
| EU AI Act enforcement | — | — | Start 2025 |
| Open-source models on Hugging Face | 300,000 | 500,000 | 700,000 |
| AI-related job postings (US) | 1.2M | 1.5M | 1.8M |
Data Takeaway: The US is the world's largest AI market with zero federal regulation. The EU, with a smaller market, is leading governance. This is a direct result of the left's failure to translate critique into policy. The open-source ecosystem is booming, but without progressive stewardship, it risks being co-opted by the same forces it seeks to challenge.
Risks, Limitations & Open Questions
What Could Go Wrong
- Self-Fulfilling Prophecy: The more the left frames AI as an existential threat, the more it alienates technologists who might otherwise be allies. This deepens the ideological divide, making collaboration impossible.
- Regulatory Backlash: If the left continues to push for moratoriums rather than nuanced regulation, they risk a populist backlash that rejects all oversight, leading to a Wild West scenario.
- Lost Generation: The left's negativity discourages young progressives from entering AI research. The field becomes dominated by those who see it as a tool for profit, not public good.
Unresolved Challenges
- How to fund public AI? The left has no answer to the question of who pays for democratized compute. Taxing AI companies is a start, but the amounts are trivial compared to the investment needed.
- How to define fairness? The left's critique of bias assumes a shared definition of fairness, but in practice, different communities have different priorities (e.g., accuracy vs. representation).
- How to handle dual-use? AI can be used for good (medical diagnosis) and harm (surveillance). The left's blanket opposition to AI ignores this nuance.
AINews Verdict & Predictions
The American left is making a historic strategic error. By remaining purely critical, they are ceding the most transformative technology of the 21st century to forces that explicitly reject their values. This is not a call for the left to abandon critique—the critiques are valid—but to evolve from diagnostician to surgeon. The left must start building: open-source models trained on ethically sourced data, public compute infrastructure, labor-friendly automation tools, and governance frameworks that balance innovation with protection.
Predictions:
1. By 2026, a progressive AI coalition will emerge, modeled on the EU AI Act, but with a US-specific focus on worker ownership of AI tools. This will be driven not by traditional politicians but by a new generation of left-leaning technologists.
2. By 2027, the first public compute bank will be proposed in a state like California or New York, funded by a tax on AI inference. It will face fierce opposition from Big Tech but will pass due to grassroots pressure.
3. By 2028, the left's silence on open-source will be broken. A major progressive figure (possibly a former tech worker turned politician) will champion open-source AI as a public good, reframing the debate from "AI is dangerous" to "AI should belong to everyone."
What to Watch: The next 12 months are critical. Watch for any progressive politician or organization that launches an actual AI project—a model, a dataset, a tool—rather than just a report or a speech. That will be the first sign that the left is finally building.