Can Google Bard dethrone ChatGPT?
Google’s perception as an innovator is at risk.
UPDATE 3/23/2023: Google Bard is now available for public testing. Google launched it on March 21.
A lingering scent of desperation descended upon the tech hub of Mountain View, California on Monday. Alphabet (formerly Google) announced it intended to add a new generative AI feature to its core search product.
Like ChatGPT, this product, Bard produces written responses to user-submitted questions. This data is synthesized from the vast quantities of human-made content on the Internet.
Conceptually, it breaks little new ground.
Following a short closed beta, Google intends to release Bard as a standalone product for the general public in the coming weeks and eventually integrate elements of the tool into the main search engine.
Despite the usual Silicon Valley pomp and confidence, it’s clear that Alphabet is operating from a place of sheer panic.
ChatGPT — its capabilities and virality — took the company by surprise. A second body blow followed when Microsoft agreed to invest $10bn into the AI’s developer, OpenAI.
Why did this terrify Alphabet? Because it’s hugely dependent on search. It was responsible for over half of the company’s revenue in the quarter ending December 31, 2022.
The search product is, in essence, an information retrieval product. People go to Google because they want to find something or learn something.
ChatGPT arguably offers a better way to do that, and it was created by a company with (at the end of 2022) fewer than 400 employees.
OpenAI wasn’t even a company until recently. It began life as a non-profit.
OpenAI changed models after realizing the following:
- Developing AI is expensive
- AI is hugely computationally expensive to run
- Investors, by their very nature, expect to receive a profit.
It’s a David and Goliath tale, and Alphabet’s panic only amplified after Microsoft dumped $10bn into the company.
It’s not just about the money. Microsoft has a bigger and more capable cloud computing offering than Google.
They’ve invested heavily into the hardware necessary to run large-scale computationally-expensive AI applications. And so, OpenAI has a clear path to popularizing its current and future AI tools.
For the sake of transparency, I haven’t used Bard yet. But I have spent most of my decade working in technology journalism.
I’m intimately familiar with the tell-tale signs of a company in a state of panic. One desperately trying to catch up with a rival. Existential fear has a distinctive smell, and Alphabet reeks of it.
I’m pretty skeptical about Bard’s future chances. Last night, I sat at my computer and came up with a list of why Google is unlikely to reverse the tides now in motion.
For the sake of fairness, I also tried to come up with a list of reasons to be cheerful about Alphabet’s future. Naturally, one list was longer than the other.
Why Google’s Bard could potentially fail
Google is preternaturally bad at commercializing new products. While its engineering prowess is second-to-none, the company quickly loses interest in the months following launch day.
And that’s why genuinely groundbreaking products — like Stadia, Android Things, and countless others — all suffered untimely deaths.
Incidentally, this is also why Alphabet is so vulnerable to ChatGPT. As its 2022 financials reveal, its primary sources of profit are Google’s core services.
Google’s search product has declined in quality in recent years. I’m not the only person to make that observation.
What’s to say that Bard won’t simply produce results based on content that, while effective from an SEO perspective, is otherwise low-quality and inaccurate? Garbage in, garbage out.
And then we get to the thorny issue of plagiarism. Bard will use content from third parties. Newspapers. Blogs. Media outlets. The issue of who owns an AI-generated derivative work is still unresolved.
We can expect pushback from news organizations over this. Australia already requires Alphabet (and other large social media platforms) to compensate publishers for content that appears in timelines and search results.
We imagine they’ll be even more pissed off when they realize Google is creating unauthorized knock-offs of their articles.
Bard will inevitably reduce the traffic sent to websites — particularly those centered around content. Google users won’t need to click through. Again, this will likely provoke some pushback from creators.
Big tech isn’t popular — either in the US or abroad.
It isn’t hard to imagine Congress — or, more likely, the European Commission — taking some tough punitive actions against Alphabet, especially if it’s seen to be hurting news and media organizations.
Alphabet’s business is heavily centered around advertising. As Microsoft’s disastrous first forays into the chatbot space showed, AI algorithms can easily produce incorrect or offensive content.
Advertisers are heavily sensitive to “brand safety.” It’s why YouTube routinely demonetizes controversial content.
Finally, AI is hugely expensive. By one estimation, OpenAI pays $0.01 for every 30 words generated.
Scaling Bard to Alphabet’s global customer base will inevitably raise the company’s operating costs. Servers, GPUs, and power aren’t cheap.
Google almost certainly created Bard while in a fit of flight-or-fight panic. And so, these thorny issues were likely set aside.
A problem ignored is a problem deferred. Eventually, they’ll have to reconcile with these issues.
Reasons to be optimistic
Of course, Alphabet has a few tricks up its sleeve. Not many, sure, but they’re strong.
- Google is still the leading search engine. While Alphabet has to catch up technologically, it doesn’t have to build a user base. That already exists.
- Alphabet is one of the few companies that build AI hardware and software.
- This means two things: First, Alphabet can build purpose-built hardware designed for the workflows it intends to run. Microsoft has to compete with other vendors for Nvidia A100 cards.
- Finally, LaMDA is a really powerful bit of software. So powerful it convinced a Google Engineer it was sentient.
- Combine that with Google’s knowledge of the wider Internet — something built over two decades — and we have the potential for something amazing.
In short, while it’s technically possible for Alphabet to deploy Bard at scale, doing so will inevitably ignite conflicts with content creators, newspaper owners, and regulators.
It poses a risk to its core advertising business. And it’s still of an unproven usefulness.
Even ChatGPT — which is impressive — struggles with accuracy. It produces content that seems plausible but often isn’t. Auto-generated code that doesn’t work. Blog posts that are filled with factual errors.
While this may improve over time, it’s not something you can use commercially — at least, not without some serious human-powered editing. And I’m yet to learn how Bard will be any different.
Race to second place
At the time of writing (Tuesday, Feb 7), Microsoft is reportedly hours away from announcing the latest fruits of its collaboration with OpenAI.
Meanwhile, Chinese search giant Baidu is working on its own ChatGPT alternative called Ernie Bot.
When Bard is unleashed to the public later this month (or, perhaps in March), it will enter an increasingly crowded market.
And while I don’t believe Google’s search dominance is threatened, it’s entirely possible that we could see a gust of momentum behind its biggest rival, Bing.
With ChatGPT now behind a paywall, Bing could have a temporarily-unique selling point. People would have a reason — beyond Microsoft Rewards — to use it.
And that could be enough to start shifting consumer habits. In the long run, that’s probably a good thing. Monopolies are bad, especially when they’re centered around access to information.
But perhaps more importantly, Google’s perception as an innovator is at risk.
Rather than being the company that makes cool products (even if it eventually kills them), it will now become the company playing catch-up.
- Gmail’s creator says ChatGPT will destroy Google in two years
- Google Stadia ends today but the Stadia controller lives on
- Google reportedly taking on Apple with its own AirTag-like device
- Summari is building AI tools that people can trust