People go to a search engine to accomplish some action (Book a flight, find a good recipe, etc). They are only looking for answers in-so-far as it helps them accomplish that action. The new GPT plugins are poised to short circuit the existing search process by providing — not answers — but rather completing user actions. An example:
Lets say I am going to Austin TX for a marketing conference. I will need air travel, airport transport, a hotel, a few dinner reservations, and a conference itinerary:
Good old Googling produces:
- A search result with 130 flights to wade through.
- A search result with 50 hotels in downtown Austin in a nasty map UX with so many results, it is just the worst. (I dare you to try to find the link to the hotel ‘website’ for the map search in under a minute on Google.)
- A search result with taxi’s, Ubers, public transport, and hotel shuttles.
- A search result with 50 restaurants in downtown Austin on a hideous overlapping cluster of a map that is a meaningless jumble of currency signs.
- A list of pages at the conference site to sift through.
All served with a minimum of 10 ads per search to ignore. Even a seasoned traveler will take an hour – probably two – to work through the options. Google search has become and untenable process to endure. Is it any wonder people go straight to Expedia or Amazon to purchase things?
As a bonus, some of that info is produced with inferences to what Google thinks it knows about your preferences and history.
Meanwhile ChatGPT with plugins produce:
Here are your flights, Uber, hotel, and dinner reservations. I set reminders on your phone calendar, and sent receipts to your email. I put in a few reminders to pack, and get travel supplies if you need them.
That is all done in under a minute. If there are issues with your personal preferences or questions about flights, it will stop and prompt you.
The big Search Engine Era is over.
Welcome to the new era of the Action Engine.
But How do We Get to Action Engine?
It seems like in order to get to the action of the example being real-world ready, we have a few steps to take first – or concurrently. The first of those steps is turning Bing and Google into recommendation engines. This is the point we are at now with GPT4 being bolted onto Bing and the imminent bonding of Bard and Google. They simply don’t know all of our preferences that would go into choosing a flight, hotel and transport. They can make some very educated guesses — recommendations — but they don’t have the data to book it for you. This is why I have found the whole concept of search as an Answer Engine to be a broken construct. Users are not looking for answers – the user is looking for recommendations as a basis for decision and taking actions.
Enter a ChatGPT Plugin from Expedia. So Expedia has indepth flight information, owns hotels.com, and has agreements with all the major car-for-hire/rental services. Given what it knows about people who book similar services, there is a very high degree of probability that they would be able book it for you with confidence of success being high.
Booking a trip, is a fairly easy to understand for most people. A much more difficult idea to wire our brains for, is all the little steps throughout a day that take time an attention. Simple little steps we all take to get through a normal day:
- Check the weather? Not really a thing to do because the real question is, “Do I need an umbrella”. Great, have a plugin remind me of that when I am walking out the door.
- Check the traffic? Not really, a GPT plugin already sent alternate routes to your car.
- Need milk? I don’t know, but a plugin knows the last time I ordered milk, and my average consumption. You can set it to order that milk for you, or just remind you that you probably need milk on the way home.
- Ever use Zapier? It takes quite awhile to wire up a Zap to do anything complicated. A plugin should be able to automate it all.
- You are on your phone, and you start flipping between apps to find something. Then you copy and paste it. A plugin can go find all that for you without having to leave the app.
This is but one from what I call the Plugin Playbook.
OpenAI – All your Base Belongs to Us
If you are still with me, then all this leads to an interesting realization about big data, Google, OpenAI, and the future of the web.
If a plugin is using the OpenAI API, that means all those actions that require AI input are running through OpenAI servers. All the recommendations, purchases, results, receipts, preferences, and personal details related to that action are also rolling over to OpenAI. HELLO!? That is a staggering ton of data that only Google can dare compete with long term. Those services that are featured in the OpenAI plugin/API ecosystem stand to rake-in massive revenue.
One of the behind-the-scenes AI drama-party that has occurred, has been the tit-for-tat back-n-forth dueling press conferences and program releases by Microsoft and Google. Last week we saw Expedia announce a partnership with OpenAI to produce a travel AI plugin. Now this week, we see Google put out a guarantee for travel. Related? Of course they are related.
Part Two Next week