My Position on AI

There continues to be a lot of “discussion” concerning AI. I have not taken a stand on this blog, preferring to follow around @apoorplayer and make counter-arguments, which is just sort of trollish. So let me write here and give him a chance to follow me around for once.

Let me start with a personal story. On St Patrick’s Day 1978, my mother, who had just turned 42, died of liver cancer. I was 19 at the time, and dropped out of my freshman year at college to help during her illness, which lasted about 5 months. After her diagnosis, she went through surgery, but they found that the tumor was inoperable. They then treated her with an experimental chemotherapy, which made her extremely sick while leaving her cancer unimproved. We then decided not to go further, and she died a few months later.

Even today, the survival rate for liver cancer is low–37% in the best case scenario. However, the American Cancer Society notes “People now being diagnosed with liver cancer may have a better outlook than these numbers show. Treatments improve over time, and these numbers are based on people who were diagnosed and treated at least five years earlier."

Stating the obvious, she would likely have a better chance of survival, or at least a longer life, if she was diagnosed in 2025 than in 1978. There’s nothing particularly earth-shattering about this generally-accepted fact. Nevertheless, the problem with liver cancer is that it is often not diagnosed until it is advanced beyond the localized stage, which greatly reduces survival rates. One area of research is trying to discover biomarkers that would lead to earlier detection, or to treatment before the cancer developed at all. “Despite the significant advances in cancer biomarker research,” the website for cancer research lab Creative Biomart explains, “several challenges remain.”

One major challenge is the heterogeneity of cancer, where different parts of the same tumor or different tumors in the same patient may have distinct biomarker profiles. This complexity requires comprehensive and dynamic biomarker panels to capture the full picture. Another challenge is the need for standardization and validation of biomarker assays to ensure consistent and reliable results across laboratories and clinical settings.

They then say:

“Looking forward, the integration of advanced technologies such as artificial intelligence and machine learning with biomarker research holds great promise. These technologies can analyze large datasets to identify novel biomarkers and predict treatment outcomes more accurately. In addition, the development of non-invasive biomarker tests, such as liquid biopsies, will further revolutionize cancer detection and monitoring.”

The research to identify biomarkers is largely built on AlphaFold technology. In 2021, Forbes published an article describing the breakthroughs of AlphaFold2 (we are currently on AlphaFold3). “In 1972, in his acceptance speech for the Nobel Prize in Chemistry, Christian Anfinsen made a historic prediction,” that “it should in principle be possible to determine a protein’s three-dimensional shape based solely on the one-dimensional string of molecules that comprise it.” The Forbes author continues:

Finding a solution to this puzzle, known as the “protein folding problem, has stood as a grand challenge in the field of biology for half a century. It has stumped generations of scientists. One commentator in 2007 described it as “one of the most important yet unsolved issues of modern science. AI just solved it.”

Before AlphaFold, they continue, “we knew the 3-D structures for only about 17% of the roughly 20,000 proteins in the human body. Those protein structures that we did know had been painstakingly worked out in the laboratory over the decades through tedious experimental methods like X-ray crystallography and nuclear magnetic resonance, which require multi-million-dollar equipment and months or even years of trial and error. Suddenly, thanks to AlphaFold, we now have 3-D structures for virtually all (98.5%) of the human proteome.”

AlphaFold technology “is widely employed in various aspects of diagnostic research, such as the study of disease biomarkers, microorganism pathogenicity, antigen-antibody structures, and missense mutations.”

All of that was about AlphaFold2; AlphaFold 3 was announced on 8 May 2024. “It can predict the structure of complexes created by proteins with DNA, RNA, various ligands, and ions. The new prediction method shows a minimum 50% improvement in accuracy for protein interactions with other molecules compared to existing methods. Moreover, for certain key categories of interactions, the prediction accuracy has effectively doubled.”

AlphaFold is owned by Alphabet’s AI lab DeepMind, and the source code of AlphaFold 3 was made available for non-commercial use to the scientific community upon request in starting in November 2024. So instead of requiring each lab undertaking such research to purchase multi-milion dollar equipment, Alphabet has made it free for non-commercial use, which means that more researchers have access and the chance of solutions increase accordingly.

Thanks to AI, there is a chance that we will see significant breakthroughs that will prevent some other 19-year-old kid from having to watch their mother die.

So this is why I find it difficult to get worked up about LLMs getting “trained” on your texts. There are bigger fish to fry, and your complaints are negatively affecting public opinion about an technological breakthrough that should be celebrating because it could solve this and many, many other complex problems that threaten humanity.