Press "Enter" to skip to content

How AI search engines became the next big tech arms race

From the original Windows with its revolutionary graphical icons, to the iPhone with its touch and pinch, to the algorithmic ranking of Google Search, there have only been a handful of products over the years that have truly upended the way we as consumers access information through technology. We may now be on the cusp of another — and as with the examples above, its rise could bring significant changes to which companies rule the tech landscape.

In November, artificial intelligence research lab OpenAI released ChatGPT, a way to interact with its large language models using casual speech. Ask a complicated question, and it could pull resources from the internet and give a lengthy answer that read like human prose. Describe a programming function, and it could write the code. Ask for a text to be re-written in a different style or language, and it would oblige.

A dose of AI smarts could give Bing an advantage over Google, but not for long.

The chatbot ignited the collective imagination of the internet, as screenshots of replies that sounded just like human speech spread on social media.

It also reportedly triggered a “code red” at Google, summoning founders Sergey Brin and Larry Page back to the company. And that’s because the ultimate application for ChatGPT as a consumer product was obvious: it would make the perfect search engine.

These language models have been trained on enormous amounts of data to understand how to consume and create language in a natural way, and they’re good at collating, summarising and modifying information in response to prompts. The current version of Google Search gives you a bunch of links and ads — as well as some contextual information — in response to a search string, and, generally, your query is vague, so you get many links to look through.

But a search engine based on a generative model could take your very specific inquiry and give you a tailored set of instructions, advice or information. You could list the ingredients you have and get a custom recipe read out, or ask for a bespoke rhyming poem about any subject. You could have a conversation about what kind of dog you should get using any criteria, or find a restaurant that makes your favourite dish within walking distance from your hotel.

Google Search can do similar things now, but it’s not so easy. And the company has the capability to make a chatbot thanks to its LaMDA (“Language Model for Dialogue Applications”), but it thus far has resisted doing so. Chatbots are designed above all else to sound convincing, so they’re infamous for given wrong information with total confidence, or offending people by saying awful things sincerely. The technology can also easily be abused for impersonation, scams or cheating at school.

Besides, Google didn’t need to improve Search with LaMDA. According to StatCounter, it currently commands a 93 per cent share of the enormous search market. ChatGPT’s public release was a clear threat to that, and so what happened next, happened fast.

In January, Microsoft said it would invest $US10 billion ($14.4 billion) in OpenAI, on top of a prior investment, promising new AI tools for consumers and enterprise. Then on February 1, it announced the first of these: models powered by GPT that would sit in on Teams video calls, keeping track of who said what to whom, generating summaries, mailing out to-do lists and making personalised recaps. It also said AI features were coming soon to virtually all its productivity software, including Office.

Google responded on February 6, saying it was leveraging LaMDA to create the next version of Google Search; an agent called Bard that would sit on top of regular search results and respond to your inquiry in conversation.

The next day, Microsoft confirmed it was making an AI search bot of its own, based on a next-generation version of GPT, as part of its web browser Edge and its search engine Bing, and there were preview versions available immediately.

Microsoft chief executive Satya Nadella clearly sees an opportunity here to make a surprise play for Google’s biggest crown, and he hasn’t bothered mincing words.

“We are grounded in the fact that Google dominates this space,” he told the Wall Street Journal this week.

“A new race is starting with a completely new platform technology. I’m excited for users to have a choice, finally.”

He went even further in an interview with The Verge.

“[Google are] the 800-pound gorilla in this. That is what they are. And I hope that, with our innovation, they will definitely want to come out and show that they can dance,” he said.

“And I want people to know that we made them dance, and I think that’ll be a great day.”

With technology from the company that kicked off this chatbot craze, and versions of its product already accessible by members of the public, Microsoft currently enjoys a perceived advantage over Google. It has heaps of resources, and is attacking a massive market (in search) where it’s long had a comparatively small market share.

Any users it can convert over from Google to Bing (which admittedly is a big ask from a brand familiarity perspective alone) could be hugely valuable, especially if the interest in AI boosts increases demands for ads on the search engine. And in response, Google can only develop better AI and features, which actually benefits everyone in the space as the current issues with generative search are collectively figured out.

On the other hand, those issues are not inconsequential.

Beyond the potential for people to use the technology for nefarious purposes, the entire point of an AI searchbot is that responses are generated on demand, meaning nobody actually knows what it might say until it says it.

Companies will need to build in automatic content moderation, the ability to detect abuse, and a way to explain to users what’s actually happening and why. That includes reminding them that delivering 100 per cent factual information is not something chatbots can currently do — as demonstrated by Google on Wednesday, when a factual error was spotted in its own promotional video for Bard, prompting a $US100 billion sell-off of its parent Alphabet’s shares.

It’s clear that chatting to language model agents is the future of asking for information, and it’s easy to see how it could be adapted for verbal conversations. But given the amount of time it will still take before it can be used reliably and safely, there’s no way to know which tech giant will come out on top.

Be First to Comment

Leave a Reply

Your email address will not be published.