I saw this going viral Anthropic CEO goes full techno-optimist in 15,000-word paean to AI.
The author’s predictions are way too optimistic. None of those things he predicts are going to happen, or at least not due to AI or in our lifetimes. It’s not at all going to cure cancer or cure aging, or extend lifespans much. This is because those problems are so much more difficult and multifactorial. Developing a cancer treatment requires not just understanding the biology and the medicine, but also testing it on live subjects. The latter is why it’s so hard to develop new treatments.
The only thing AI is revolutionizing is image/text generation and maybe automating or improving productivity on the margins. It’s otherwise hardly transformative. It is no substitute for top human talent or creativity. AI is revolutionizing text boxes, text/image generation, and text inputs, but not the rest of society. It’s like a smarter version of Google, at best. But instead of SEO spam, it’s AI-generated spam.
AI has led to much more spam. If you think AI spam is bad now, you ain’t seen nothing yet. The reality is the majority of people do not care or cannot notice if content is AI-generated or human-generated, so to save money we can expect more of the former, especially as AI content improves in quality. I have already seen this in news stories where there are weird mistakes and other artifacts suggestive of AI generation. If you are not paying attention then you would probably not notice.
There will be a two-tier sort of internet in which ~99% of content is generated by or with the help of AI, and then artisanal content for those who have more discriminating or discerning tastes or are willing to pay. This is how top content creators, such as on YouTube, Substack, or Twitter will thrive and make money despite a deluge of AI-generated content. It’s similar to how the ability to record music digitally did not put top musicians out of business; the elite content still rose to the top despite more content being produced.
Same for extending lifespan. Humans run up against a wall or hard limit, at around 110-120-years-old, that has not changed in a century despite modernity. It’s like, “We can make a 60-year-old feel like a 40-year-old,” but we cannot make an 80-year-old like a 60-year-old. At that point, things start to fail so rapidly nothing can work. There is little reason to expect AI can raise this limit, as this is a biological limit innate or intrinsic to humans, not replicating human behavior or thinking, which is what AI does best.
Mouse models of aging suffer from the problem that human lifespans respond much less to interventions, either biological or environmental, compared to smaller organisms. Making some tweaks can extend mouse lifespans by 50%; at best for humans you may get 5 months. What makes humans live a long time also makes them more impervious to life extension.
The author argues that AI will make it easier to find new permutations of drugs, but these still must be tested and developed. Overall, I am pessimistic AI can transform society outside of the textbox. AI is great at solving textbox-based problems or content-based problems, not so great at other stuff, like automating jobs or solving ‘big problems’ like cancer or aging. This is not to say there cannot be an AI revolution that puts a dent in those problems, but it’s not this.
One can argue, “This shows a lack of imagination of what AI is capable of doing.” Maybe, but the author has not supplied any better or more more convincing evidence otherwise. Just saying something is possible and making handwaving assumptions about the promises of AI, is no more convincing than saying it is not. If the realm of what is possible is extended to an infinite or undefined timeline, then in theory anything is possible, which not useful either.