Reading Time: 2 minutes

In a fascinating post, Ian Mulvaney of BMJ compared his predictions for AI from several months ago, with what actually happened.

I’m not usually one for predictions, but I can’t deny it was fascinating to see an expert’s predictions, and too see which of them came true. So, being emboldened and foolhardy, here are five predictions for generative AI in scholarly communications. For the most part, they are innovations that already exist, but which I think will become more prominent:

  • Generative AI will have rapid success providing methodological checks and suggestions, rather than full writing.
  • Generative AI might cause a temporary flurry of increased sales of smart phones and personal devices equipped with “AI chips”, but cloud-based tools will remain by far the most effective way to deploy AI.
  • Generative AI will have an entirely new front end, and will move towards competing directly with search engines (as I write this, OpenAI has just launched a Google competitor, SearchGPT). Search engines and Gen AI tools will exchange ideas and innovations in a frantic attempt to capture and/or maintain dominance in the search engine market.
  • Generative AI will make more use of your profile when delivering results.
  • Generative AI will move towards more subject-specific sites, using smaller, single-domain corpora.

What won’t happen is:

  • I don’t think the practice will change of evaluating generative AI tools by running them against a standard test. While the comparison of several tools against a standard test looks valuable, I think such testing ignores more fundamental issues around the corpus employed. We have known for years that unless you examine and are aware of the bias in your corpus, your results may be suspect.
  • There should be more attention paid to the corpus, rather than simply (as the State of California is attempting) expecting a simplistic commitment from AI providers not to “cause harm” – an impossible task.

I look forward to seeing what turns out. It won’t be the first time I have been proved wrong!

Post Views: 231