Google and Baidu have both revealed that they will bring their own competitors of OpenAI鈥檚 ChatGPT artificial intelligence into their search engines, as an AI arms race between the world鈥檚 largest tech companies explodes into the public sphere.
The news comes just days after revealed that Microsoft is planning to integrate ChatGPT into its Bing search engine to provide results.
Google says it is testing an AI called Bard with a select group of users and will release it to the wider public in coming weeks. The model is based on , a language AI that a Google engineer claimed last year had become sentient, which was met with widespread scepticism from academics.
Advertisement
Sundar Pichai, Google鈥檚 CEO, said in a that Bard will feature in Google search results soon, and that it will 鈥渄istill complex information and multiple perspectives into easy-to-digest formats鈥�.
Baidu, which is sometimes called the 鈥淐hinese Google鈥�, has announced a similar project that it calls Ernie Bot, or Enhanced Representation Through Knowledge Integration. This AI is reportedly undergoing internal testing that should be complete next month. One source that Baidu鈥檚 intention is to launch it as a standalone tool, but then to gradually merge it into its search engine.
The announcements from these two technology giants came soon after Microsoft sparked speculation that it was integrating AI into its search engine Bing by announcing a further , the company behind ChatGPT.
Screenshots taken by some users last week reportedly showed that Microsoft was testing AI responses to search queries in its Bing search engine. The company鈥檚 press conference on 7 February is .
鈥淕oogle has always had a claim of being the dominant search engine just because of coverage and speed, but this could be a way of finally giving Bing a market advantage,鈥� says at the University of Birmingham, UK.
Large language models like ChatGPT are able to write poems, scripts and essays and carry out a wide range of tasks from written prompts. Such tools could replace search engines by writing concise, direct and bespoke answers to queries rather than pointing people to websites that might contain answers to their question, although one issue with current models is that they often make up accurate-sounding but false information.
鈥淭hey鈥檙e also giving you things that maybe are false. ChatGPT is great at making things up, it鈥檚 great at inventing references,鈥� says Lee.
The pace at which AI technology is being rolled out because of strong competition seems like an arms race and could lead to other problems, says at the Alan Turing Institute, UK.
It is well known that bias, sexism and racism exists in AI models because they are trained on human-generated content online, and that problem will still be there if they are integrated into search engines, she says.
鈥淥ne of my big concerns is that, as we use these systems in search engines, it makes us increasingly passive in how we access information about the world,鈥� says Aitken. When you search and you find two or three pages that tell you different things, you choose what to trust, she says, but an AI presenting its answers will make that choice for you.
Aitken says this creates the risk that certain factual inaccuracies or political, racial or gender biases become cemented as fact rather than critically analysed.
Lee has similar concerns.聽鈥淎sk ChatGPT聽 to write a limerick about Donald Trump and it refuses because Donald Trump has been flagged as a political hate figure, but ask it to write a limerick about Biden and it鈥檚 fine,鈥� he says. 鈥淎nd so there is this kind of implicit bias to these things. That鈥檚 going to be a real concern.鈥�
Google, Baidu and OpenAI didn鈥檛 respond to requests for comment.
