ChatGPT plugins: OpenAI's attempt to kill Google
If this works, the ad model becomes obsolete. If I were Sundar Pichai I'd be freaking out.
ChatGPT is not a search engine. It’s frozen in 2021, and it hallucinates. Despite that, a large number of people insist in using it as one. This generally leads to all sorts of failures. If I were in Sam Altman’s shoes, I’d think “ok, what if we give users what they want?” and try to steer it in that direction. Plugins are obviously one step towards making it ChatGPT more useful and Google-like: they connect it with external resources that deal with information in traditional ways. Once you have used them, it’s obvious why this is a big deal.
When you get access to the alpha version of ChatGPT that supports plugins, you have access to a store. It looks like this:
You can see that I installed Instacart (it has the uninstall button on it). The first thing I did after was ask for a recipe for a cake.
Google can’t prepopulate an instacart shopping cart for me but ChatGPT just did. I’m two clicks away from spending $50. It’s easy to imagine how this could apply to everything. For example, let’s say I want to build a PC and ask it to put together a list of parts. It might generate the shopping list at Amazon or whoever else paid to be featured on the plugin store. This case is not as straightforward as the cake, because GPT may not know the most current components worth recommending. Still, Amazon will be very motivated to sanity check the results and upsell me on better alternatives. All they have to do is pick up where ChatGPT left off.
Now imagine that you are an Instacart competitor (e.g. Amazon Fresh). Creating a plugin for ChatGPT is very easy. During an event I attended this past weekend, Andrey Mischenko from OpenAI demoed how to create and deploy one in less than 60 seconds. This means there is a huge incentive for it, as well as competition to be featured in the store. It’s another customer acquisition channel to try out. And here is where OpenAI has an advantage over Google: because they are not a public company subjected to the tyranny of the quarterly earnings report, they can give early adopters an extremely sweet deal. Google could react with a similar offering on their side, perhaps even better from a purely technological perspective. They might even get the user experience right. However, this poses two problems:
It competes against its own ad business. I can easily measure the ROI of my Google ads against the business I get from the conversational bot leads. If the bot is better (and it would have to be if they are matching OpenAI), I will shift my budget towards it.
Search is way cheaper to operate than LLM inference, probably by an order of magnitude at least. As a result, if Google seriously moves in this direction their profits will suffer and their stock will tank. A textbook example of the Innovator’s Dilemma.
I’m very curious to see how this plays out. If ChatGPT keeps attracting eyeballs, the plugin marketplace will explode. There is a potential for an ecosystem comparable to Apple’s app store.
Bonus: if you have access to the Plugins alpha and are feeling naughty, try playing with the plugin to run arbitrary code on Replit that I built for the hackathon.
