Artificial intelligence is worse than humans in every way at summarising documents and might actually create additional work for people, a government trial of the technology has found.
Amazon conducted the test earlier this year for Australia’s corporate regulator the Securities and Investments Commission (ASIC) using submissions made to an inquiry. The outcome of the trial was revealed in an answer to a questions on notice at the Senate select committee on adopting artificial intelligence.
The test involved testing generative AI models before selecting one to ingest five submissions from a parliamentary inquiry into audit and consultancy firms. The most promising model, Meta’s open source model Llama2-70B, was prompted to summarise the submissions with a focus on ASIC mentions, recommendations, references to more regulation, and to include the page references and context.
Ten ASIC staff, of varying levels of seniority, were also given the same task with similar prompts. Then, a group of reviewers blindly assessed the summaries produced by both humans and AI for coherency, length, ASIC references, regulation references and for identifying recommendations. They were unaware that this exercise involved AI at all.
These reviewers overwhelmingly found that the human summaries beat out their AI competitors on every criteria and on every submission, scoring an 81% on an internal rubric compared with the machine’s 47%.
We know the performance of L2-70b to be on par with L3-8b, just to put the difference in perspective. Surely they models continue to improve and we can only hope the same improvements will be found in L4, but I think the point is that models have improved dramatically since this study was run and they have put in way more attention in the fine-tuning and alignment phase of training, specifically for these kinds of tasks. Not saying this means the models would beat the human summaries everytime (very likely not), but at the very least the disparity between them wouldn’t be nearly as large. Ultimately, human summaries will always be “ground truth”, so it’s hard to see how models will beat humans, but they can get close.