Tom Snyder: Exploring what adding ChatGPT to Siri could mean | WRAL TechWire

Two weeks ago, I wrote about how big companies don’t conduct their own innovation.

Almost universally, any new feature or product from a big tech company is actually something that was developed by a startup that was acquired by the brand, or that licensed the capability. Little did I know that a few days after I wrote that piece, Apple would announce the integration of OpenAI’s ChatGPT into Siri.

Now OpenAI isn’t your typical startup. At eight and a half years old, OpenAI had the unfair advantage of $1 billion in pledged funding from some of the best-known innovators in the world, including Peter Thiel, Reid Hoffman and Elon Musk. It turns out $1 billion wasn’t needed for OpenAI to get swallowed up by big tech. The original contributors only provided $130 million in fuel before Microsoft took a strong ownership position with a $1 billion investment in 2019 and an additional $10 billion last year.

As interesting as it is to consider the monopoly implications of Microsoft and Apple collaborating under the radar through OpenAI … Today, I’m more interested in exploring what adding generative AI to one of the iconic voice interface platforms could mean. To do that, we must begin with a brief history of search.

Internet search has been stagnant for an extremely long time. The very first search engine was a tool called Archie, developed at McGill University in 1990 to index FTP-file names. By 1993, World Wide Web Wanderer enabled URL and web title searching, and in 1994 WebCrawler first let users search the full text of websites. By the mid 1990’s, Alta Vista, Yahoo and Lycos all had implemented a text-based search bar that has remained the primary search interface for 30 years.

For anything in the technology sector to remain unchanged for a few years is astounding, let alone for 30 years. The algorithms behind our searches have evolved and improved, but the core format – typing a short text query into a search bar – has persevered. Google took the search bar to scale, embedding it into Chrome, YouTube, the Play Store and other applications. Even when I use a voice interface, like speaking a destination to Google Maps when driving, the app interface auto-populates text into the search bar before showing the results.

Will the integration of ChatGPT and other Large Language Models into search demonstratively change how we search? Google has implemented LLM capability into enterprise search and has announced that Bard will be integrated into Google search. Microsoft has implemented Prometheus into Bing. Big tech has decided this is the future, so what does it mean?

So far, the primary difference is in the output of a search. Instead of an immediate list of website URLs, searches are opening with an attempt at contextual, text-based replies, followed by traditional link lists. The inputs are still search bars – albeit ones that accept longer input queries than just a short phrase or question. Based on the research I have been able to conduct, I see no evidence that users are fundamentally shifting to asking questions in a new or more detailed way. When users go directly to a purpose-built generative AI tool like Perplexity, Copilot or Gemini, their prompts may be more elaborate. But for the legacy places we engage search bars, the paradigm of the short-text prompt remains.

My hunch is that as generative AI performance improves, we’ll see more elaborate responses to our search queries than the few sentences of AI-generated text that we get today. Increasingly, search will respond with a multi-media mix of text, images, audio and video. I anticipate that we will shortcut directly from a list of search results to something that looks like a fully designed webpage, automatically generated by AI based on your search inquiry.

Technology platform companies benefit from users remaining on their proprietary platforms. Google, Microsoft, Apple and others will resist sending you off their own websites and to others, now that they can instantly AI-generate a response that is richer than a list of HTML links. Rather than link you to a traditional (existing) website, you’ll be presented with an instantly AI-generated website as a response to each search. There will remain pass-through links to facilitate e-commerce, for example. But the massive data aggregation companies will track online behavior a layer or two deeper in your online interaction than they are doing today by keeping you on their platforms longer.

I’ve noted this “platform grip” recently in a recent behavior change of my smart TV at home. After a software upgrade sometime in the last six months, I am getting multiple streaming platform recommendations directly on my smart TV home screen that are tied to my personal viewing habits. In the past, I’d be presented with generic paid-promotion suggestions on the home screen, but I’d have to select and enter a specific streaming platform to see that platform’s recommendations that were tailored to my own personal data. I looked at the updated terms of service associated with the TV software update and saw that by accepting the upgrade (as if I had a choice), I’ve given the TV manufacturer permission to track what I watch on any streaming platform I watch through that device.

The TV manufacturer is attempting to hold me on their own operating system, so I do not need to dive to the specific streaming sites of each streaming platform individually. That’s a data-expansion strategy – fueled by generating recommendations to keep me on their platform rather than redirecting me to leave it.

If the search platforms make a similar shift, it will drive a fundamental shift in the economics of search. Today, companies spend massive amounts of time and energy building content on their own websites and keeping that content fresh to improve their SEO performance (Search Engine Optimization). High-ranking content is prioritized in search results, gets more clicks, is well cross-linked and is actively engaged by people browsing the internet. SEO-optimization can be a lower-cost method for companies to get “promoted” by search algorithms, and is an alternative path to paid ad placement.

If the new paradigm I describe comes to bear, companies may be disincentivized to put effort into SEO-heavy websites. If generative AI creates a website-like response, consumers may far less regularly end up on any individual company’s website. Like the smart TV attempting to keep me on its home screen through AI-generated recommendations, the search engine will keep me on its platform, minimizing my need to “click-through”. And Google, for example, won’t be as disadvantaged in data collection about me even if I’m browsing on Safari or another competitor’s browser.

The easiest way for companies to make sure their interests are embedded in these AI-generated search responses is likely to be typical pay-for-play advertising, and search companies would likely see a spike in that revenue. But Google and others still need domain-specific data and information to enable accurate AI response. If that information isn’t coming from regularly updated private websites, where will that source data come from? I think companies will shift to creating and maintaining valuable data sets and content API’s to feed the search algorithms instead. Those who do this with the highest quality, and the most up-to-date information will see the equivalent prioritization as the high-ranking SEO of today.

Large data set owners are already aligning with the fundamental changes of search that I have described. They want to make sure that their data gets incorporated into future generative AI responses. The Associated Press has licensed its article library to OpenAI. Shutterstock has deals with Meta and OpenAI. Reddit has licensed its data set to one of the big (as yet unnamed) players. Reuters, Wiley and LeMonde all have new revenue line items on their balance sheets for data licensing. Certainly these partnerships not only include legacy data sets, but continuous feeds of fresh data.

This concept of a generative AI “website-like” search response is a radical shift – and I’ll admit, not one I have a personal awareness of that anyone is working on. But we are overdue for a major shift. The 30-year paradigm of a search bar and “list of links” response had a great run. But that era is coming to a close.

Next week, I’ll return to the ChatGPT integration with Siri and where we will see much more radical changes in how we interact with technology. See you then.