Smart Software: More Novel and Exciting Than a Mere Human

September 17, 2024

green-dino_thumb_thumb_thumbThis essay is the work of a dumb humanoid. No smart software required.

Idea people: What a quaint notion. Why pay for expensive blue-chip consultants or wonder youth from fancy universities? Just use smart software to generate new, novel, unique ideas. Does that sound over the top? Not according to  “AIs Generate More Novel and Exciting Research Ideas Than Human Experts.” Wow, I forgot exciting. AI outputs can be exciting to the few humans left to examine the outputs.

The write up says:

Recent breakthroughs in large language models (LLMs) have excited researchers about the potential to revolutionize scientific discovery, with models like ChatGPT and Anthropic’s Claude showing an ability to autonomously generate and validate new research ideas. This, of course, was one of the many things most people assumed AIs could never take over from humans; the ability to generate new knowledge and make new scientific discoveries, as opposed to stitching together existing knowledge from their training data.

Aside from having no job and embracing couch surfing or returning to one’s parental domicile, what are the implications of this bold statement? It means that smart software is better, faster, and cheaper at producing novel and “exciting” research ideas. There is even a chart to prove that the study’s findings are allegedly reproducible. The graph has whisker lines too. I am a believer… sort of.

The magic of a Bonferroni correction which allegedly copes with data from multiple dependent or independent statistical tests are performed in one meta-calculation. Does it work? Sure, a fancy average is usually close enough for horseshoes I have heard.

bonferroni graph

Just keep in mind that human judgments are tossed into the results. That adds some of that delightful subjective spice. The proof of the “novelty” creation process, according to the write up comes from Google. The article says:

…we can’t understate AI’s potential to radically accelerate progress in certain areas – as evidenced by Deepmind’s GNoME system, which knocked off about 800 years’ worth of materials discovery in a matter of months, and spat out recipes for about 380,000 new inorganic crystals that could have revolutionary potential in all sorts of areas.  This is the fastest-developing technology humanity has ever seen; it’s reasonable to expect that many of its flaws will be patched up and painted over within the next few years. Many AI researchers believe we’re approaching general superintelligence – the point at which generalist AIs will overtake expert knowledge in more or less all fields.

image

Flaws? Hallucinations? Hey, not to worry. These will be resolved as the AI sector moves with the arrow of technology. Too bad if some humanoids are pierced by the arrow and die on the shoulder of the uncaring Information Superhighway. What about those who say AI will not take jobs? Have those people talked with an accountants responsible for cost control?

Stephen E Arnold, September 17, 2024

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta