When OpenAI released its ChatGPT tool in November 2022, it was a shot across the bows of Google, with generative artificial intelligence promising a new way to access the world鈥檚 information beyond search engines. Since then, the rivalry between these firms has only grown, with both announcing new services this week. While there are signs that OpenAI is winning this race, is either company aiming for a future anyone actually wants?
On 13 May, at a live demonstration event, OpenAI unveiled GPT-4o, a new AI model capable of more human-like interactions. The AI appeared able to converse with company representatives live on stage, as well as analyse live video footage streamed to it using smartphone cameras, responding to what it 鈥渟aw鈥�.
Google unveiled similar capabilities the following day, showing off how its Astra project could review computer code and fix errors, as well as find a missing pair of glasses in an office, all through the camera on a Google smartphone.
Advertisement
The firm it will be further integrating AI-generated results into its search engine, summarising content from web pages that previously would have been linked to from a search results page, reducing the need for users to click through to websites.
These competing announcements, made just one day apart, had echoes of a similar situation in early 2023, when Google unveiled its Bard chatbot (now rebranded as Gemini) to try to head off the early success of ChatGPT, whose release appeared to have taken the search giant by surprise. 鈥淕oogle has been in this space for quite a long time and had a lot of models that were almost ready for the marketplace before OpenAI opened up the market with ChatGPT,鈥� says at the Alan Turing Institute in London.
Of the two offerings, Google鈥檚 has generally been seen as less significant. 鈥淭he [recent] announcements do help cement OpenAI鈥檚 position at the top in terms of generative AI,鈥� says at the University of Oxford. 鈥淲hen you think 鈥榞enerative AI鈥�, you think GPT-4, and now you鈥檙e going to think about this new model.鈥�
This lead is borne out in data: according to market intelligence firm , while the ChatGPT app was downloaded onto smartphones 3.2 million times this February, Google鈥檚 Gemini was only downloaded 630,000 times that month. The web versions of theses AIs show the same trend, with ChatGPT attracting 1.8 billion visits to its website in April, according to analysts SimilarWeb, while Gemini had less than a quarter of that number.
OpenAI also seems more confident about the abilities of its products. In of GPT-4o, OpenAI claims that its new model can outperform all competitors, including Google鈥檚 Gemini, while Google鈥檚 demonstration only claimed that its AI could match the performance of other models.
But even if OpenAI is pulling ahead, we should keep in mind that the success of the generative AI revolution is far from certain. The field is so new that companies aren鈥檛 competing for market share or customer loyalty, but merely to have the latest products, says Katell.
And there are suggestions that this product invention is stalling. Ahead of the GTP-4o announcement, OpenAI had to quell gossip that it was about to unveil GPT-5, a long-rumoured, purportedly more powerful model capable of reasoning. Instead, the technology it presented may make the issues of anthropomorphising its AI system even more pronounced by giving it human-like interactivity, without necessarily improving its basic capabilities or preventing it from making things up.
Despite the bluster of the past few years, it is also worth noting that only , according to surveys by the Pew Research Center. 鈥淚t鈥檚 still very unclear to what extent [this technology] matters,鈥� says V茅liz. 鈥淭he majority of people haven鈥檛 actually used generative AI tools, and there are some reasons to think that it might end up being less helpful than some people thought it would be.鈥�
