Remove Data

Forget about “Machine Learning” for a Moment. Why “Machine Un-learning” is Essential for Future Freedom if we hope to avoid Orwell’s “1984.”

A Free Society Requires the Constant Pursuit of Truth


The search for truth is an ongoing process that requires continuous refinement and correction. In today’s world, where information is abundant and algorithms constantly improve, it’s easy to assume that the search for truth is a one-time accomplishment.


However, in the pursuit of a free society, we must recognize that machine learning alone is insufficient without machine unlearning. As Andrew Liu, a computer science and statistics Ph.D. from Stanford, highlighted in a recent blog post, unlearning is not just about forgetting what we know, but about reevaluating our understanding of the world and challenging our existing beliefs.


Without this combination of learning and unlearning, we risk being trapped in a cycle of misinformation and misconception, perpetuating biases, oversimplifications, and inaccuracies.


In the context of artificial intelligence (AI), machine unlearning is crucial, as it acknowledges and addresses the limitations of our knowledge, ensuring that our machines and society remain vigilant and adaptable [1].


After dedicating a considerable portion of my existence on planet earth to the humble pursuit of computational enlightenment, I find myself in agreement with many of Andrew’s thoughts.


In a free society, the pursuit of truth is paramount. It’s the foundation upon which our democratic institutions, social structures, and individual freedoms are built. However, with the explosion of data and increasing reliance on AI, verifying the accuracy of information becomes a tiring task.


If we rely solely on machine learning without machine unlearning, we risk perpetuating inaccuracies and misconceptions. The cycle of learning and refining becomes self-reinforcing, making it impossible to arrive at the truth.


For example, the current administration in Washington, D.C., has been accused of presenting various concepts as “facts” to the country without sufficient evidence [2-5]. These claims include downplaying the severity of the US border crisis [3], declaring the COVID-19 vaccines “safe and effective” [2], labeling inflation as “transitory” [3], and understating the threat posed to the United States by China [4].


Furthermore, the administration has promoted the idea that self-perceived gender identity should take precedence over biological differences between the sexes, despite the physical distinctions between men and women being too important to ignore [6].


All of these things have been taught to “artificial intelligence” models as if they were fact. They must be urgently un-learned by the machines.


Machine learning algorithms are designed to refine their performance based on the data they’re trained on, but this approach is not a guarantee of accuracy. These algorithms can amplify biases and perpetuate misconceptions if the data is flawed or biased.


Moreover, as new information emerges, machine learning alone cannot adapt to correct these errors. It requires an equally important mechanism to unlearn and reevaluate existing knowledge.


Unlearning in AI ensures that our machines can recognize and rectify flaws in their understanding. This does not mean discarding all knowledge acquired, but rather acknowledging and refining it.


By incorporating unlearning into AI, we can create a system that’s constantly evolving, self-correcting, and reliant on verifying the accuracy of information.


In a society where truth matters and more information is created each day at an increasing rate, the stakes are high. We cannot afford to dismiss the importance of unlearning in the pursuit of technological advancements.


If we don’t prioritize machine unlearning, we risk being trapped in a never-ending cycle of misinformation and bias, as a judge recently likened the Biden administration’s censorship efforts with big tech to Orwell’s ‘1984’ [7].


In such an environment, our freedom to make informed decisions, hold our leaders accountable, and participate meaningfully in the democratic process is severely weakened, especially as machines participate more in our ongoing dialogues.


A free society requires the constant pursuit of truth. While machine learning is a vital component, we must not underestimate the importance of machine unlearning. By acknowledging and addressing the limitations of our knowledge, we can ensure that our machines and society remain vigilant and adaptable.


We must prioritize the development of AI that not only learns but also unlearns, constantly refining and correcting its understanding of the world, with the objective of arriving at unbiased truth.




1. Liu, K. (2022, January 24). The importance of unlearning. Andrew Liu’s Blog.


2. Washington Stand. (n.d.). The 5 Worst Lies of the Biden Administration.


3. Fact check: Biden makes false and misleading claims in economic speech. (2023, January 28). CNN.


4. Comer: President Biden’s Pattern of Lies, Corruption, and Obstruction Demand Action from Congress. (2023, March 8). House Oversight Committee.


5. Lott, J. R. (2022, March 17). Column: Forget Trump, Hold Biden and His Administration Accountable for Their Lies. Hunt News.


6. National Review. (2023, October 5). Biden’s Ridiculous Gender-Neutral Family Policy.


7. Verity, J. (2023, June 2). Judge Likens Biden-Big Tech Censorship to Orwell’s ‘1984’, Admin’s Answers on Free Speech ‘Prove Him Right’. The Daily Signal.