Suddenly, it seems like Google has been twiddling its thumbs for the last 10 years, content to rest on the laurels of Google Search’s overwhelming global dominance. In just a few days, the benchmark technology company in artificial intelligence (AI) was abruptly overshadowed by Microsoft CEO Satya Nadella’s surprise announcement – the Bing search engine will incorporate a chatbot developed by OpenAI, the people responsible for the famous ChatGPT.
A day earlier, Google tried to preempt Microsoft by announcing Bard, its own version of a search engine with intelligent chat. But Google wasn’t ready to demonstrate Bard at the hastily organized international press event in Paris two days later. Google’s limited presentation in Paris bombed spectacularly when Bard’s recorded example of an intelligent search included incorrect information about the James Webb Space Telescope. Shares of Alphabet, Google’s parent company, tanked by 8% that day. Global capital markets slapped a heavy penalty on Google’s uncharacteristically improvised parry to the Microsoft offensive.
Bard is an experimental conversational AI service, powered by LaMDA. Built using our large language models and drawing on information from the web, it’s a launchpad for curiosity and can help simplify complex topics → https://t.co/fSp531xKy3 pic.twitter.com/JecHXVmt8l— Google (@Google) February 6, 2023
Why all the sudden interest in AI? Because the whole world has seen ChatGPT and understood its potential. Many people think the tool, which paraphrases content culled from the internet, could revolutionize the search engine experience (with a few modifications) because it’s nicer to get information by dialoguing than by typing keywords. The chatbot’s ability to generate moderately complex text such as summaries, itineraries and essays is also an attractive feature for many users. Large language models (LLM) make this possible, although much doubt remains about their reliability and accuracy.
Some tech pundits are already predicting that the hybridization of generative AI and conventional search engines will be the most significant innovation in consumer technology since Apple released the first iPhone. Bing has always lived in the shadow of Google Search (3% and 90% of global search engine share, respectively). For the first time, it has put Alphabet on the back foot.
The elephant in the room
But the arms race to develop increasingly intelligent search engines has broader implications than the prestige of innovation leadership. Owning the world’s most widely used search engine and web browser enabled Alphabet (with Meta close behind) to dominate the global advertising market for over a decade, bringing in an average annual revenue of $220 billion. Alphabet has used this flood of cash to make strategic company purchases and fund various R&D projects, such as Waymo (autonomous cars) and Calico, the anti-aging biotechnology company.
This bonanza may be coming to an end. For the first time since 2014, Alphabet and Meta accounted for less than 50% of the global advertising market (specifically, 48.4%) in 2022. It is the fifth consecutive year that their combined share has dropped since peaking at 54.7% in 2017, and analysts expect it to continue declining. Why? TikTok is coming on strong and has become the search engine of choice for many young people. Amazon continues to grow, and Apple has cut into Meta’s business since it began allowing app tracking to be blocked.
The abundance of advertising manna may be running out for Google and Facebook. A few years ago, Meta decided that the metaverse was the answer to this problem and its inability to attract young audiences. But Alphabet has been investing in AI for decades and doesn’t seem to have a backup plan, which would explain its hasty reaction to Microsoft’s Bing announcement.
A frenetic race
Nadella has turned Microsoft around in less than a decade. When the executive took the helm in 2014, the company depended almost exclusively on Windows and Office Suite revenue. Nadella decided to bet big on cloud services and AI. Its Azure cloud computing division is already generating a quarter of the group’s revenue. Two years ago, Microsoft invested $1 billion in OpenAI. Shortly after the enthusiastic public response to ChatGPT, Microsoft invested another $10 billion in Open AI to develop the conversational chatbot for Bing.
So, what has Alphabet been up to? As company leaders have taken pains to note, Alphabet was the one who laid the technological foundation upon which chatbots are built. Its Google Brain division and DeepMind, the British company it acquired in 2014, are global leaders in AI. In early February, Alphabet CEO Sundar Pichai reminded everyone that the Transformer network architecture presented in 2017 is the cornerstone upon which the scientific community has built the advanced generative artificial intelligence we are seeing today.
Bard, Google’s bid to make its search engine intelligent, is a pocket-sized version of LaMDA, one of the company’s most advanced language modeling projects. Launched two years ago, LaMDA made international headlines in 2022 when software engineer Blake Lemoine reviewed the model’s ethical posture and said he believed the AI had gained a conscience. DeepMind plans to offer a beta version of its model this year, which it has dubbed Sparrow.
There is no plausible way to deny the impact of ChatGPT on the strategies of big tech companies. And yet that’s what Prabhakar Raghavan, Alphabet’s vice-chairman, did last week. “We have been following our own roadmap in artificial intelligence development for years. ChatGPT has not influenced us at all,” he said in a February 15 press conference in Paris. Nevertheless, the company did rush to announce Bard without offering a launch date. Raghavan himself noted that the company didn’t have an approximate date to offer. “What matters most to us is to achieve the quality we want for the service.”
The technology industry is susceptible to fads, and generative AI is clearly getting all the buzz right now. Aside from Microsoft and Google, Chinese tech giant Baidu recently announced that it is working on its own version of a search engine/intelligent chatbot hybrid. On the other hand, Meta canceled its Galactica project late last year, a language model capable of producing scientific articles based on millions of previously analyzed documents, because it proved to have sexist and racist biases.
To gain a foothold, chat-enabled search engines must prove that they provide reliable information, and this isn’t easy. Examples of ChatGPT’s fabricated content have flooded social media recently, and then there was Bard’s embarrassing and very public error about the James Webb Space Telescope. When pressed, the test version of the chat-enabled Bing also makes up content.
Some leading experts warn of going too fast with this technology. “Large language models should be used as a writing aid, but not much else,” said Yann LeCun, head of AI research at Meta and a renowned expert in the field. DeepMind CEO Demis Hassabis also recommends moving slowly. “It’s okay to be cautious about this,” he said. For the moment, this caution is conspicuous by its absence.
Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition