Whether this will be completely overwhelmed by all the slop and all the other negative implications of AI, I don't know.

But it's been a long time since I've had something really positive to say about the technology.

The nice thing is that, for once, using AI may actually have a positive effect in the long term. If people can do better, more exhaustive literature research, then the citation graph will become better connected.

Old papers will become more likely to be discovered and in general, less research disappears in the flood.

Now, GPT is not as good as understanding my field as I am, and it gets many things slightly wrong. But it's usually much better at understanding adjacent fields (like statistics) where I just don't have the training.

This works in many domains. Maybe GPT is not as good at searching as you are, but it speaks tens if not hundred of languages, so it can just retry the same search in another language to see if, say, French people have said something relevant.

It takes hours at least to understand a single paper and half the time the field is so different from yours, there's no hope of getting anywhere.

Moreover, because everybody does literature research poorly, even following the citation graph exhaustively doesn't always throw up everything you need to consider.

The agent-like models, like ChatGPT's deep research can really help out here. To summarize, you ask a question in natural language and GPT goes off and searches the web for your.