The battle for AI supremacy
ChatGPT, an artificial-intelligence chatbot backed by Microsoft, was the first out of the blocks, but challengers are hot on its heels. What does the future hold? Simon Wilson reports
What is ChatGPT?
It’s the artificial-intelligence (AI) chatbot taking the world by storm since its release in November, drawing a million users within a few days – and a hundred million within two months – and prompting predictions of potentially revolutionary changes in business, culture and society. The technology it uses is called a “large language model”, which assembles a gigantic linguistic database by trawling the web, then uses complex algorithms to extract related words – and arrange them in convincingly human-like, grammatically correct sentences – in response to queries.
Who’s behind it?
ChatGPT was created by a San Franciscobased start-up called OpenAI, but its biggest backer is Microsoft, which has invested $10bn in the project. Earlier this month Microsoft unveiled its long-awaited integration of ChatGPT into its search engine Bing, calling it a “co-pilot for the web”. For Microsoft, long the staid elder statesman of Big Tech, playing catch-up with the likes of Google and Meta, this is a very big deal, with the potential to seize ownership of next-generation web search. Currently, Bing has a paltry 3% share. Google tried to steal its thunder by publishing details of its own chatbot, Bard.
Will chatbots change the world?
Not everyone thinks so. Cory Doctorow, for example, a technology journalist, reckons that “ChatGPT and its imitators have all the hallmarks of a tech fad”. In his view, Google’s web strategy shouldn’t be to “race Microsoft to see who can be first to leap off the peak of inflated expectations” for chatbots, but to focus ruthlessly on “how to exclude (or, at a minimum, fact-check) the confident nonsense of the spammers and SEO creeps” (that is, those who exploit “search engine optimisation” to creep up the search results). Others disagree. Tyler Cowen, the US academic, argues that once the initial wave of excitement has passed, the use of AI assistance is permanently going to change the nature of reading and writing and thinking. The New York Times technology columnist was “deeply unsettled, even frightened” by the quasi-human abilities of the Bing chatbot he’d been asked to test. Ben Thompson of Stratechery called his experience with the Bing bot “the most surprising and mind-blowing computer experience of my life”.
How much do chatbots get wrong?
A lot, says Adam Rogers in Business Insider. AI chatbots are not actually “intelligent”, they’re just a posh version of autocomplete – and there’s a big risk that they could prove more dangerous than useful. First, chatbots don’t understand what they’re saying; they merely recapitulate things they’ve absorbed elsewhere. That means they often get things wrong, regurgitating lies and conspiracy theories – or presenting mainstream opinion as unshakeable science. Researchers call this their tendency to “hallucinate” – producing “highly pathological translations that are completely untethered from the source material”. Second, when it comes to the use of AI in web search, chatbots “elide the sources they’re drawing on, and the biases built into their databases”. But complex subjects and nuanced arguments don’t lend themselves to “one-and-done” answers. Chatbot-driven search means oversimplified and unreliable results – and “the collateral damage in this war of the machines could be nothing less than the obliteration of useful online information forever”.
But that’s not in anyone’s interest?
Indeed. It’s early days – and the billions flowing into generative-AI suggest investors have confidence that the companies shaping the sector are not going to pursue strategies that anger and alienate their customers. Most fundamentally, the age of AI looks like placing the corporation, rather than the academy, back at the heart of innovation, says The Economist. In the late 20th century, corporate research and development (R&D) became “steadily less about the R than the D”. Companies moved away from doing science towards developing existing ideas. In the field of AI, by contrast, “almost all recent breakthroughs in big AI globally have come from giant companies
– Google, Microsoft, Amazon, Meta – because they have the computing power, and because this is a rare area where research results can rapidly be incorporated into products. Currently, Microsoft is in the lead, says Richard Waters in the Financial Times – although Google’s grip on internet search will not be easy to loosen. Either way, the tremor that ran through Alphabet’s stock price earlier this month (the Google-owner’s shares fell 8% in a day after a stumbling demonstration of its chatbot search) shows that “the disruptive potential of generative AI is starting to sink in on Wall Street”.
So what about smaller players?
The capital flowing into generative-AI start-ups, which collectively raised $2.7bn last year in 110 deals, suggests that venture capitalists think that not all the value will be captured by big tech. In all, according to data collated by investment firm NFX, some $12bn of funding has been raised by 450 start-ups hoping to commercialise generative AI technology. Start-ups such as Anthropic and Character AI have built their own ChatGPT challenger, says The Economist, and Stability AI has created a popular opensource model that converts text to images. In the field of “computer vision”, involving the analysis of images, it’s the Chinese laboratories that are pre-eminent, led by the state-backed Beijing Academy of Artificial Intelligence (BAAI). In generative-AI more broadly, Chinese firms such as Baidu, Alibaba and NetEase are racing to match the West’s recent developments, says the FT. “Everyone wants to create ChatGPT now,” says Huan Li, creator of WeChaty, one of China’s most popular chatbot programs, “but it’s very hard, especially for Chinese companies, which can’t get the latest Nvidia chips and have limited data sets for training AI models.”
“The war of machines could obliterate useful online information forever”