Google crisis grows as its ChatGPT competitor flops

For years, Google’s cash cow search business kept it at the forefront of innovation.

From experiments with self-driving cars in California to using artificial intelligence (AI) for drug discovery in London, the company pushed the boundaries of what was possible thanks to the font of profit at its heart.

Yet now its innovation crown is being challenged by an old foe: Microsoft.

Last Friday Microsoft claimed that the newest model of ChatGPT, an AI algorithm backed by the PC pioneer, was showing “sparks” of intelligence that were “strikingly close to human-level performance”.

Artificial general intelligence, as it is known, has been the holy grail of researchers for decades. The prospect that Microsoft may have gotten there first is yet another body blow for Google.

The company has been left scrambling to react to the surprise success of ChatGPT, which launched to the public last November. Google executives have labelled it a “code red” problem and co-founders Sergei Brin and Larry Page have emerged from semi-retirement to hold meetings with top AI execs to thrash out a response.

ChatGPT presents an existential threat to Google's core business.

“Search is as good as how you use it,” says Sachin Dev Duggal, founder of UK start-up Builder.AI. “Most people just type in a bunch of words.

“With a language model [the technology underpinning ChatGPT], you can get a more refined search result. It can infer what those words mean in a wider context.”

Microsoft has already integrated ChatGPT into its search engine, Bing, sparking fears that the new AI-powered search tool could upend the traditional power balance that has seen Google dominate for decades.

Last week Google’s fightback began in earnest as it launched its own competitor, Bard, to the public.

At a staff meeting in the wake of Bard’s initial reveal last month, Jack Krawczyk, a senior director at Google, tried to distance the tool from its search engine.

“I want to be clear: Bard is not search,” he said, according to CNBC. However, he admitted: “We can’t stop users from trying to use it like search.”

The expectation is that, before too long, Google will plug AI into search.

“You’ll see AI-powered features in Search that distill complex information,” chief executive Sundar Pichai said last month.

Speaking to The Telegraph, Krawczyk said: “There is a separate effort for how generative models will look in search; that is not what you see here.

“It is a very early stage of this technology and we really want to make sure right now we are focused on delivering the right amount of quality.”

Bard is a large language model that has been trained on billions of articles and millions of pages of books.

Unlike Google’s ubiquitous search bar, where users type in a query and are returned a list of internet links, dealing with Bard is more akin to a conversation.

The AI can engage in back-and-forths with users in natural language, answering questions with colourful, chatty responses.
It opens conversations with: “I’m Bard, your creative and helpful collaborator. I have limitations and I won’t always get it right.”

Google executives said that Bard is best looked at as a creative aid, providing inspiration or ideas, rather than answers grounded in fact.

“We think of this as being complementary to search,” says Zoubin Ghahramani, a Cambridge academic and Google’s head of research.

However, the product got off to a rocky start. A blog post from Google boss Pichai posted last month lauded the bot. It would provide “fresh, high-quality” responses to questions from internet users, he said. The announcement included a suggested prompt: “What new discoveries from the James Webb Space Telescope can I tell my nine-year-old about?”

Unfortunately, Bard answered the question wrongly.

The AI bot misattributed a key scientific discovery to the Nasa telescope in its answer.

The error wiped 6pc from Google-parent company Alphabet’s share price on February 8, a wipeout worth $120bn.

In an effort to avoid any more embarrassing – and potentially costly – slip-ups, Google has set strict guardrails around Bard in its public trials.

Words such as “genocide”, for instance, will provoke a stock answer: “I’m a language model and don’t have the capacity to help with that.” The bot also steadfastly avoids questions about medicine or possible health conditions.

However, testing this week found Bard still makes the same mistake when asked about the James Webb telescope.

As with ChatGPT, the technology still struggles to differentiate between fact and fiction.

Krawczyk says Google has tried to “design a product where people are not taking the answers to be purely authoritative”.

Unlike Microsoft’s Bing chatbot, Bard does not serve up external links in answers it gives. Experts say that makes it far more limited and less useful.

“[Microsoft chief executive Satya] Nadella and OpenAI [ChatGPT’s developer] are right now miles ahead of Google and its Bard endeavour for AI,” says Dan Ives, an analyst at Wedbush Securities.

Google is already facing mounting pressure from investors and rivals.

Its digital ad sales slipped 4pc in the final three months of 2022, only the second time in its history it has reported a drop.

Shares have fallen by a quarter over the last 12 months.

Tightening economic conditions have forced the company to lay off 12,000 “Googlers” in an effort to cut costs.

As with Facebook owner Meta, Google is hoping that AI can help it turn things around.

Yet observers have been underwhelmed.

“Compared to what Bing has done… Bard seems to be a step backwards,” says Sridhar Ramaswamy, a former Google executive and founder of internet search start-up Neeva.

Ramaswamy says the lack of links as citations puts it a step behind Microsoft.

After ChatGPT’s viral success, Microsoft invested billions in OpenAI. Companies including Stripe, Morgan Stanley and Klarna have all announced deals with the start-up.

Google, meanwhile, has attempted to manage expectations with Bard. It has stressed the limitations of the technology. Deals have not been as forthcoming.

Jerome Pesenti, former head of AI at Meta, says Bard “doesn’t quite seem at the level yet in terms of performance”.

Some insiders are growing nervous. While Google edges forward, Microsoft appears to be racing ahead. On the anonymous tech worker forum Blind, Google staffers are glum. “ChatGPT clearly has an edge over this,” says one.

When one member of the public asked Bard how long it would take for Google to shut it down, the AI last week falsely claimed it had already been shut down because of “lack of adoption”.

Googlers will be hoping this is simply a bug in the code, not an ominous portent.