At Fast Company, Gary Smith and Jeffrey Funk apply a helpful bucket of cold water to claims that artificial intelligence will fundamentally revolutionize science.

There is a great deal of hype out there. Smith and Funk cite an article in Wired from way back in 2008, claiming that “The Data Deluge Makes the Scientific Method Obsolete.”

The everyday reality is more like this: AI speeds up many tedious processes but doesn’t replace creative thinking:

Today, AI is being increasingly integrated into scientific discovery to accelerate research, helping scientists generate hypotheses, design experiments, gather and interpret large datasets, and write papers. But the reality is that science and AI have little in common and AI is unlikely to make science obsolete. The core of science is theoretical models that anyone can use to make reliable descriptions and predictions.

Gary Smith, Jeffrey Funk, “Why AI can’t replace science,” Fast Company, June 28, 2024

But hype, of course, gallops faster than progress. For example, former Google CEO Eric Schmidt predicted that Large Language Model (LLMs or chatbots) can write summaries of articles for scientists. Well, sort of:

We now know that LLM literature reviews are unreliable. In May of 2023, two months before Schmidt’s article was published, a credulous lawyer submitted a legal brief that had been largely written by ChatGPT to a Manhattan court. When pressed about fake citations that ChatGPT had included in the filing, ChatGPT obliged by generating fake details of fake cases. The judge was familiar with the relevant precedents and rebuked (and later fined) the lawyer for submitting a brief that was full of “bogus judicial decisions . . . bogus quotes and bogus internal citations.” That, in a nutshell is the problem with relying on LLMs for literature reviews and other factual information. If you know the facts, you don’t need an LLM. If you don’t know the facts, you can’t trust an LLM.

Smith, Funk, “Can’t replace science”

That’s similar to the recent bears in space problem, where a chatbot generated an impressive amount of utter nonsense about the Soviet Union sending bears into space. As Smith and Funk say, “LLMs are prone to generating confident garbage.” It’s not clear that they’ll get better because they are not actually thinking.

They go on to unpack other examples of (“it will change everything”) hype vs. plodding reality in the field of drug discovery, noting “The real test is whether new products and services are developed faster and cheaper with AI than without it.” That part’s still up in the air because only some problems in science are made easier to solve by more powerful number-crunching.

One thing is for sure. Hype has much more practical value in show business than it does in science.

You may also wish to read: AI Is still a delusion. Following instructions and performing fast, tireless, error-free calculations is not intelligence in any meaningful sense of the word. Not knowing what words mean, neither OpenAI’s ChatGPT 3.5 nor Microsoft’s Copilot nor Google’s Gemini can do a simple logic test.