This article is more than 1 year old
Google now won't black-hole all AI-made pages as spam
Plus: Microsoft's nu-Bing can't answer question on exoplanets correctly either
Web content automatically generated by AI will be ranked according to its quality by Google Search, the internet titan confirmed this week, after previously suggesting it would blanket down-rank computer-created pages.
Last year, Google Search Advocate John Mueller said using machine-learning tools to produce text is "essentially the same as just shuffling words around" to game Google's bots and artificially pump up a page's rank in search results. AI-generated content, thus, would be treated as spam as it would be assumed to be low quality or useless; websites hosting that kind of lazy material would appear lower in results or not at all.
But Google has since changed its mind and has updated its rules. "It's important to recognize that not all use of automation, including AI generation, is spam," it said in a memo to web developers on Wednesday.
"Automation has long been used in publishing to create useful content. AI can assist with and generate useful content in exciting new ways."
According to Google, software has been used in the past by publishers and website owners to output things like "sports scores, weather forecasts, and transcripts." And that is true: the Associated Press, for instance, has automated the production of articles about corporations' financial results, sports matches, and similar stuff that can be produced from templates.
"AI has the ability to power new levels of expression and creativity, and to serve as a critical tool to help people create great content for the web," Google added.
The Chrome maker will now continue to rank content that demonstrates four qualities it dubbed E-E-A-T – expertise, experience, authoritativeness, and trustworthiness – regardless of how that material was created. Using AI to generate content to exploit search engine optimization and manipulate search results, however, is prohibited, just like with human-written spam.
"Appropriate use of AI or automation is not against our guidelines. This means that it is not used to generate content primarily to manipulate search rankings, which is against our spam policies," Google confirmed.
- Google's AI search bot Bard makes $120b error on day one
- Google shows off upcoming AI search features, leaves Bard waiting in the wings
- Conversational AI tells us what we want to hear – a fib that the Web is reliable and friendly
- China's Baidu reveals generative AI chatbot based on language model bigger than GPT-3
The change comes after the search giant plans to start using its LaMDA AI technology to answer people's search queries and automatically generate summaries of information and other text. Hot on the heels of Microsoft, Google announced plans to integrate this model into its search systems on Tuesday.
Effectively, Google is going to offer its LaMDA-based chatbot Bard as a conversational interface for searching the web and looking up info. And if it's OK for Bard to emit text automatically, maybe shooting down all AI-made pages as spam would look a teeny bit hypocritical.
Instead of clicking on different websites and sifting through information, Bard will answer input queries by summarizing webpages. Along those same lines, Microsoft also teased the world with a sneak peek of its AI-powered search: the Windows giant has integrated a large language model built by OpenAI, said to be more powerful than ChatGPT and based on GPT-3.5, into Bing. New search and text generation capabilities are also coming to its Edge web browser, too.
All of this is expected to roll out gradually, and only then we can judge how well it works.
The technology is nascent. Developers are still figuring out how to prevent these models from spewing false information as demonstrated by Google's Bard. In a demo, the model incorrectly claimed NASA's James Webb Space Telescope (JWST) "took the very first pictures of a planet outside of our own solar system," a feat that was actually achieved by the European Southern Observatory's Very Large Telescope years before the JWST was launched. Over $120 billion of Google's market cap was wiped shortly after the error.
Microsoft's upgraded Bing also seems to struggle with this question. When we asked whether the JWST took the first exoplanet pictures, it said yes and linked to two sources that didn't quite support this. While it's true the telescope has taken the first pictures of a never-before-seen exoplanet, it isn't the first 'scope to snap the first-ever picture of exoplanets.
How AI-powered search will change the face of the internet remains to be seen. Some are already questioning whether it could spell doom for publishers that rely on banner ads. Instead of clicking on websites to find information, the new Bing and Bard will just summarize it, reducing the amount of eyeballs on pages.
Publishers and site owners may be left in the dust, as Harvard's journalism foundation Niemen Lab noted. Netizens won't visit pages, sapping publishers' abilities to record ad impressions and collect revenue.
Will these models actually be useful and we can all spend less time browsing the internet? Or will they lead us astray because they cannot generate accurate information, we wonder. ®