Ask A Genius 1182: Follow the River Where the AI Lead

Author(s): Rick Rosner and Scott Douglas Jacobsen

Publication (Outlet/Website): Ask A Genius

Publication Date (yyyy/mm/dd): 2024/11/13

*Interview conducted in November, 2024.*

Rick Rosner: Let’s ask ChatGPT: “What are three significant technological advances expected within the next five years?”

Scott Douglas Jacobsen: That’s a reasonable way to phrase it.

“What are three significant technological advances expected within the next five years?” Responses include widespread adoption of artificial intelligence and automation, breakthroughs in quantum computing, and advances in personalized medicine and biotechnology.

Rosner: Let’s focus on the third one, given what we discussed earlier. What do you envision with regards to personalized medicine and biotechnology?

Jacobsen: Personalized medicine will likely involve tailored medical treatments and gene therapy systems connected to CRISPR gene-editing technologies. One aspect could be the development of more targeted cancer therapies. Every type of cancer has unique structures on its surface that affect how easily immune cells can detect, grab, engulf, and kill it. For the immune system to attack cancer, it needs to recognize which cells are cancerous. Some cancers have distinct surface markers that the immune system can learn to identify, enabling the production of the necessary receptors to detect and attack them.

Rosner: Newer gene therapy techniques can expand the range of surface structures that can be targeted by the immune system. This is already occurring but currently only addresses certain types of cancer. Cancer is complex; numerous cellular mutations must align for it to become a fully malignant disease. Thus, various attributes can be targeted for treatment.

Jacobsen: So, you agree with the AI’s assessment that we will see improved cancer treatments?

Rosner: It’s often said that cancer isn’t one disease but hundreds of different ones. We’re likely to develop more methods to disrupt the growth cycles of these cancers, make it harder for them to metastasize, or prevent metastasized cells from embedding in other parts of the body. This will result in more points of attack and a wider range of treatable cancers.

Take kidney cancer, for instance. It’s challenging because, while it doesn’t metastasize frequently when small, even a tumor up to 4 or 7 centimeters may still be considered “small” and manageable. Other cancers, however, can spread at any stage. Kidney cancer also has mechanisms that disable immune cells in its vicinity, making immune therapies less effective and complicating treatment if it spreads

There’s a drug called Keytruda that is effective against many types of cancer, including kidney cancer. However, it’s one of those treatments that, at best, reduces the risk of recurrence or progression by about 40%. It’s not a cure, and it comes with significant side effects. Cancer will continue to be a major issue for several decades because it encompasses a wide range of different diseases. I don’t think we’re at the stage where we have treatments that can reduce the lethality of all cancers by 50%. Some types of cancer are highly treatable, while others remain extremely lethal.

Pancreatic cancer is a prime example of this. If it’s not detected until symptoms appear, the prognosis is often poor.

Jacobsen: So, moving on to the next question. What was the main point you noted from ChatGPT’s response? Let’s discuss the first claim regarding quantum computing breakthroughs.

Quantum computing is expected to achieve significant advancements that will allow for faster processing speeds. This, in turn, could facilitate more complex data analysis, advanced simulations, and potential breakthroughs in cryptographic sciences, material science, and drug discovery.

That’s an intriguing field. What do you see as the main challenges with quantum computing?

Rosner: There are two primary issues. First, building stable quantum systems with more than a few qubits is challenging. Quantum computers are so powerful that even a system with only a handful of qubits can perform substantial tasks. However, to unlock their full potential, you need a system that can maintain a greater number of qubits. The inherent problem is that quantum systems are unstable—you’re trying to sustain an isolated and indeterminate state until the computation is complete.

The second problem lies in structuring tasks so they’re suitable for quantum computation. Quantum computing excels at problems like the traveling salesman problem, but there are numerous other complex problems that need to be adapted to leverage quantum capabilities.

You mentioned ChatGPT’s other comments—could you recap them?

Jacobsen: ChatGPT referenced applications in cryptography, material science, and drug discovery through advanced simulations and data analysis. Cryptography is particularly significant. There’s a classic example from the 1980s: creating an unbreakable encryption key by multiplying two large prime numbers. It was believed that factoring such a product would take millions of years with conventional computing. However, quantum computing, with its ability to process many calculations simultaneously, could potentially crack these codes, making what was once secure, breakable.

That would be revolutionary—and a potential security risk.

Rosner: Didn’t a couple of researchers win the Nobel Prize in medicine for utilizing AI to figure out how to fold proteins precisely as desired?

So, once again, this appears to be a good problem for quantum computing—designing scenarios where you’re running an enormous number of possible combinations simultaneously. In a quantum system, what would traditionally take hundreds of years could potentially be done in mere seconds. That makes sense based on what ChatGPT indicated. Over the past decade, brute-force substance testing has relied more on robotics than AI. Robots can create thousands of miniature petri dishes, each containing problematic cells.

The robots can handle the repetitive task of placing thousands of different substances into those petri dishes, which would take humans an incredible amount of time. This process has essentially reduced the reliance on human intuition to identify potentially effective substances. Instead, every possible substance is tested because automation makes it feasible. ChatGPT’s point suggests that, with quantum systems, instead of physically testing thousands of substances, you could simulate millions of tests. If the substances could be characterized in a way that allows quantum computing to simulate them, then testing a million substances becomes realistic within a reasonable time frame.

Genes essentially code for the creation of proteins with specific shapes, and in biology, structure is everything. If you aim to develop a library of millions of potential protein shapes for various applications, quantum computing could be ideal for this kind of task. ChatGPT’s first prediction was that AI would become universally adopted, which is hard to dispute.

Jacobsen: True, although the term “AI” is often misapplied. There’s a lot of simple autocomplete functionality labeled as AI.

Rosner: One could argue that AI is fundamentally about autocomplete. For example, when you train a graphics AI by inputting a vast number of art pieces, it essentially turns a text prompt into an autocomplete task, providing the most likely artistic rendition of what your words describe.

Jacobsen: Would you say a better term for AI might be “virtually unlimited autofill”?

Rosner: Yes, that’s a fair assessment. It captures the essence of how it functions—essentially an expansive version of autocomplete.

Jacobsen: I see where this is leading. You’re suggesting that even in areas like AI-generated adult content, there are intricacies in user commands that push AI to understand context and commands more deeply.

Rosner: When people pay for AI-generated adult content, they can input highly detailed prompts that cater to specific preferences. An AI-generated scenario might involve one character engaging in an act and another walking in unexpectedly. Clearly, someone programmed these prompts because they’re recurring themes in image sets.

You can see how AI progressively learns how doors, door frames, and perspectives work. At first, characters may appear stuck halfway through a door, or proportions might be incorrect. But as the AI continues training, these details improve. It eventually understands perspective better and can render scenes where the person walking in appears smaller due to distance. AI has repeatedly refined aspects like shadows through what is essentially Bayesian analysis of the most probable scenarios.

Jacobsen: So, it’s like watching the AI learn through trial and error, guided by patterns and probabilities?

Rosner: It’s a continuous learning process based on data-driven refinements. AI doesn’t truly “know” anything—we’re aware of that. It doesn’t think as humans do, but it operates on statistical foundations for how things behave. For instance, it has a statistical basis for understanding the behavior of shadows and the principles of perspective. What makes images appear realistic—or conversely, unnervingly unrealistic—often comes down to details like the eye line. I’ve observed AI in this specific context, and while it manages to get the eye line of the main subject correct, it struggles with the person in the background. In scenarios where someone is caught in an act and reacts with surprise, the foreground figure often looks fine, but the background character frequently ends up with mismatched or misaligned eyes, where one eye might be larger or pointing in the wrong direction. It’s interesting to watch AI slowly improve and start getting these details right.

Jacobsen: So, it’s like a process where the AI is learning to handle nuanced visual details over time?

Rosner: But ultimately, AI is just autofill. And yes, autofill is something we rely on for everything. In a sense, our own brains function with a kind of autofill. When we speak, we don’t think out every single word meticulously; we start the sentence, and our brain fills in the rest. Unless, of course, we’re writing something meticulous, like an essay for The Atlantic, where every word is scrutinized.

So, moving forward, you could say that we’re going to be using this type of “autofill” constantly. And that leads to the point of learning statistics. AI relies heavily on Bayesian statistics, and while I can say that confidently, explaining how feedback across neural networks functions would take more time. But for the sake of simplicity, yes.

Jacobsen: Makes sense. Should we use that as a metaphor for summarizing this?

Rosner: Sure, let’s go with this for now.

Jacobsen: Do we need another call, or are we good for now?

Rosner: No, I still need to head back to the gym one more time. 

Jacobsen: Then I’ll take a quick break and continue writing. Thank you very much. 

Rosner: Thanks again. Bye.

Jacobsen: Bye.

Last updated May 3, 2025. These terms govern all In Sight Publishing content—past, present, and future—and supersede any prior notices. In Sight Publishing by Scott Douglas Jacobsen is licensed under a Creative Commons BY‑NC‑ND 4.0; © In Sight Publishing by Scott Douglas Jacobsen 2012–Present. All trademarks, performances, databases & branding are owned by their rights holders; no use without permission. Unauthorized copying, modification, framing or public communication is prohibited. External links are not endorsed. Cookies & tracking require consent, and data processing complies with PIPEDA & GDPR; no data from children < 13 (COPPA). Content meets WCAG 2.1 AA under the Accessible Canada Act & is preserved in open archival formats with backups. Excerpts & links require full credit & hyperlink; limited quoting under fair-dealing & fair-use. All content is informational; no liability for errors or omissions: Feedback welcome, and verified errors corrected promptly. For permissions or DMCA notices, email: scott.jacobsen2025@gmail.com. Site use is governed by BC laws; content is “as‑is,” liability limited, users indemnify us; moral, performers’ & database sui generis rights reserved.