Google today introduced a new tool at its I/O developer conference that will make it easy for users who “aren’t the best at searching” (in the words of Google) to find all the content they or their company created using AI models.
The feature is fittingly named “AI Detect” and should begin getting rolled out across Google’s platforms in the months ahead as a significant effort to battle the growing problem with AI-generated disinformation and the legitimacy of online content.
In a keynote address at the event King said the announcement underscores the increasing sophistication of AI models that can generate realistic looking text, images, audio and even video.
Pichai said that Google is committed to developing and deploying AI safely, and that “as AI becomes more integral to our products, we are committed to sharing as much information as possible about how we develop algorithms.”
The system works with an array of proprietary algorithms behind the scenes and runs machine learning models trained on massive data sets of both human-produced and AI-generated content.
It examines different features, such as linguistic patterns, stylistic inconsistencies, and small contradictions that are sometimes found in the product of AI systems.
Details of the underlying technology are closely-held, though Google researchers suggested the system recognizes certain telltale signs that distinguish AI-created art from human expression.
AI Detect successfully spotted a number of AI-generated articles and images with a high degree of confidence during a live demo.
The app will offer a probability score to indicate the likelihood the article or content is AI generated, giving people a way to judge for themselves how trustworthy an unverified image actually is.
The AI Detect system will be launched as part of Google search engine, flags or indicators may be tagged alongside search results that are determined as probably AI-generated.
The company is also planning to incorporate the technology into Google Docs and other productivity software to make it easier to know when AI is responsible for writing a portion of a document.
Also there is still a talking to offer an API for developers to use AI Detect inside their apps and product.
AI Detect has been received with guarded optimism by industry commentators and policymakers. And, although they understand the value of the tool to fight misinformation, some say they doubt the accuracy of such a tool and, in the future, the possibility for AI models to change and bypass detection.
Google experts explained that AI Detect will be a work in progress, to be updated and optimized in response to changes in AI generation.
While it sounds as if these are articles covering a span of time that is yet to occur, the truth however is that this heralds Google’s proactive stance that will help it tackle the issues in an agenda that is post damage control.
As such AI-generated content becomes widespread, solutions such as AI Detect could become an important factor in maintaining transparency and trust within the digital space. Now will come some fateful months on seeing how well the new tool really works out in the field.