Understanding DeepSeek AI in Context
by John Wunderlich
‘The robots aren’t coming for your jobs, management is’
On Monday, January 27 headlines like DeepSeek tech wipe-out erases more than $1 trillion in market cap as AI panic grips Wall Street popped up on financial and tech press sites. Hardest hit were tech companies like Nvidia a Silicon Valley manufacturer of the kinds of processors used for artificial intelligence (AI) systems behind OpenAI. Also affected were the power companies planning to invest billions in electricity projects. DeepSeek has introduced a significantly more power-efficient and effective AI at a substantially lower cost than the Silicon Valley versions. In addition, the DeepSeek code is open-source, not proprietary, making it available to anyone who cares to build their version.
The latest form of artificial intelligence (AI) has dominated the technology and business news since OpenAI’s ChatGPT launch in 2022. It is the latest, and most significant, boom in capital investment without a clear path to profitability. AI models are based on data obtained on shaky legal, and dubious moral, grounds. The training data and AI model development are based on work done in data sweatshops, generally in the Global South. The data is processed in data centres that consume enough power to make building a nuclear power plant dedicated to the data centre a ‘reasonable’ possibility. If, and it’s a big if, this resource-intensive industry could benefit humanity (as claimed by some) how can this be done?
DeepSeek’s engineering innovation may have reduced AI’s capital and power expenditures. Still, there is no evidence to date that it has anything to reduce the human costs of AI, whether that is the appropriation of human ‘creators’ as training data or the exploitation of data annotators and content moderators, mainly in the Global South. This global ‘extraction machine’ will continue to demand to be fed every digital trail a person makes. “We are constantly being made to serve the needs of two systems: technology and capitalism. Their purposes and goals are almost always prioritised above all others.” Some examples include:
Military AI - The “Joint All-Domain Command and Control,” or JADC2, is a U.S. Department of Defence project to connect sensors from all branches of the armed forces into a unified network powered by artificial intelligence. In Israel, +972 and Local Call reported that the Israeli army has developed an artificial intelligence-based program known as “Lavender,” which uses AI to generate targets for assassination in Gaza.
AI in the workplace - Bossware refers to software tools employers use to monitor and manage employee productivity and behaviour. AI plays a significant role in these tools by enabling more sophisticated and automated tracking methods. Companies similarly use AI to identify candidates or screen resumes - possibly to exclude union or social activists.
AI and advertising - AI plays a significant role in advertising campaigns, enhancing their reach and effectiveness. AI can create convincing fake news articles, images, videos, and even deepfakes. AI-driven bots can automate the spread of misinformation on social media platforms. These bots can like, share, and comment on posts to amplify their reach, creating the illusion of widespread support or belief. AI can analyse vast data to identify and target specific groups or individuals with tailored messages. Some governments use AI to censor critical content and promote their messages. AI systems can automatically detect and remove posts that oppose the government’s narrative while boosting pro-government messages.
Commercial AI is devoted to achieving platform dominance. It depends on exploiting a global supply chain and reifying capitalism. Capital pours into AI companies expecting investment returns based on market domination by ‘platforms’. Platform dominance is based on the kind of ‘network effects’ that made Facebook and Google dominant in social media and search, and depends on a regulatory regime preventing alternatives. Exploiting a global supply chain creates a ‘race to the bottom’ between nations competing for business process outsourcing (BPO) contracts. This hegemonic narrative of AI is built on yet another version of “TINA” or there is no alternative. But there ARE alternatives.
Pushing back against the capitalist agenda of AI as a system of power and control involves the tools activists have always used: solidarity and organizing around clear demands.
Public AI: Public AI encourages community-driven AI projects where local communities can develop and deploy AI solutions tailored to their specific needs. This approach fosters innovation at the grassroots level and ensures that AI technologies are relevant and beneficial to local contexts. This can include AI technologies to be developed and controlled by public institutions rather than private corporations, ensuring that the benefits of AI are distributed equitably. The reduced cost and open-source DeepSeek makes this demand achievable.
Data Cooperatives: Establish and support data cooperatives to create collective or social control structures. Data cooperatives should be member-owned and member-run, ensuring that the members make all data use decisions democratically.
Transparency and Accountability: Demand companies and governments disclose how AI systems are developed, deployed, and used. This includes making algorithms and data sources transparent and holding entities accountable for the impacts of their AI systems.
Worker Rights and Protections: Unions and civil society should push for regulations that protect workers from being unfairly monitored or replaced by AI systems. This includes ensuring fair wages, job security, and organisation rights.
Inclusive AI: Demand that AI development includes diverse voices and perspectives, particularly those from marginalised communities. This helps ensure that AI systems do not perpetuate existing inequalities.
Education and Awareness: Call for public education campaigns to raise awareness about the implications of AI and empower individuals to understand and challenge its use.
Environmental Impact: Demand that AI development and deployment consider and mitigate environmental impacts, promoting sustainable practices in the tech industry.
In almost every instance above, these demands can also be used to build solidarity with workers and marginalised groups in the Global South. These demands can help ensure that AI is developed and used in ways that don’t concentrate power and control.
References
Crawford, Kate. Atlas of AI: power, politics, and the planetary costs of artificial intelligence.
Golumbia, David. Cyberlibertarianism: The Right-Wing Politics of Digital Technology.
Levine, Yasha. Surveillance Valley: The Secret Military History of the Internet.
Merchant, Brian. Blood in the Machine: The Origins of the Rebellion Against Big Tech.
Morozov, Evgeny. The AI We Deserve, Boston Review 2024.02
Muldoon, James and Graham, Mark and Cant, Callum. Feeding the machine: the hidden human labor powering A.I.
Pasquinelli, Matteo. The Eye of the Master: A Social History of Artificial Intelligence.
Sadowski, Jathan, Dr.. The Mechanic and the Luddite: A Ruthless Criticism of Technology and Capitalism.