Google DeepMind is holding back from publishing GenAI research over competition fears

Google’s DeepMind is reportedly holding back from publishing AI research papers over fears of losing its competitive edge. Co-founder Sir Demis Hassabis has made the vetting process to get research published tougher, per a report by the Financial Times, citing unnamed sources. 

The group was pulling back from sharing any progress in public so their research isn’t exploited by rivals. 

The decision is a shift from Google’s usual route which led to the release of the seminal research paper on Transformers in 2017 which eventually played a big part in the GenAI boom and the creation of ChatGPT by AI firm OpenAI.

Now, Google DeepMind will implement a six-month embargo before “strategic” research papers on GenAI are released. The source also said that the change would ultimately help researchers who were frustrated that the work they had spent time on would not be approved for these reasons.

The report revealed that this increase in bureaucracy had led to several researchers complaining that it had become near impossible to get research around GenAI published. In one instance, a paper that showed Gemini was comparably less safe and capable compared to rival AI models. 

However, papers that were based on security vulnerabilities were still being published as under “responsible disclosure policy,” where companies are given the chance to fix any flaws before the paper is released.