ChatGPT has been the talk of the town — as Madhive CEO Adam Helfgott puts it, “We’re at the peak Pokemon Go stage of ChatGPT.”
The attention isn’t unwarranted. It’s pretty amazing to ask the tool a question and watch it spit out text that's difficult to distinguish from human writing. This has led to some interesting applications, such as human-ish chatbots, and AI-generated articles that are similar in quality to those written by humans.
For some internet searchers, the tool may even have the potential to steal some of Google’s traffic, quickly answering questions like:
But ChatGPT's potential doesn't stop there.
There are already interesting use cases in industries like marketing, where the AI-powered tool can be used to generate personalized content for customers. And in the future, we might see ChatGPT used in fields like healthcare, where it could be used to analyze patient data and generate personalized treatment plans.
What sort of impact could tools like ChatGPT have on the CTV advertising industry?
Some in the space predict that ChatGPT could be integrated with existing DSPs to improve decisioning in the bid process. For example, the tool could possibly provide additional context around what sort of ads will perform best during the exact content that’s streaming and who’s watching it.
But according to Helfgott, it just won’t work:
“There's no way ChatGPT can be used by a DSP in real time to say, ‘Hey, here's all the data for an open RTB request, and here's all my campaigns. Where is this one going to have most success?’”
Why not? It’s a matter of speed and compute power.
Programmatic advertising moves fast — you have on average 20 milliseconds to make a decision. To show an ad at the right time and place, in the right way, an advertiser needs to understand:
And on and on. Plus, all of this data needs to be collected in a privacy-protected fashion without accessing personally-identifiable information.
To make this happen, let’s say an adtech platform needs to record 500MB of data, plus a healthy stream of updates to keep that data from going stale. Multiply that by the 7 billion people on the planet, and that’s a couple of exabytes just to make a basic, good adtech prototype work.
(To put in perspective, an office of 100 people would have to surf the web for 57,077 years to reach just one exabyte of activity.)
ChatGPT doesn’t fit the bill — it’s just not fast enough and doesn’t have that kind of compute power.
As CTV platforms work to crack the speed and compute power code, an AI race is underway.
It’s not the generative AI making marketing headlines but rather one that quietly sifts through massive troves of data in fractions of a second. Adtech needs an infusion of scalable, AI-driven analytics to process massive amounts of audience and inventory data in real time.
This technology will enable advertisers to make better decisions around limited CTV impression opportunities quickly enough to compete for them. Advanced computing power will enable DSPs to run the most efficient CTV campaigns and deliver the results that advertisers seek. DSPs like Madhive are evolving to use AI to make smarter purchasing decisions based on huge swaths of data in milliseconds.
As we build out these technologies, the industry will start to realize the true promise of CTV — swiftly securing premium inventory and reaching audiences for a given brand, wherever those audiences are.
Journalist Cleo Abrams explains how AI can improve the performance — and experience — of CTV advertising:
Learn more about TV advertising from Cleo: TV Advertising Explained
So, ChatGPT won’t be integrated into any adtech platforms any time soon, but the tool still has potential in the industry. As it evolves and people get more familiar with the user experience, we’ll continue to see new use cases.
Helfgott explains:
“It is definitely the start of something big, but I think it's still far away. Just like Pokemon GO showed the possibilities of what AR can be, ChatGPT is getting people to think about AI in new ways and find new applications.”
