Twitter | Search | |

Summary of #AIChat (Tuesday, May 1st, 2018)

This is a summary of #AIChat with @pchalasani, Chief Scientist @mediamath. Talking #AI #programmatic #marketing #ads #adtech
Nick Tang May 1
Snack and learn! 28 minutes until . Join us at 2PM to 3PM EST/11AM - 12 PM PST/6PM - 7PM GMT Learning with , Chief Scientist at Talking
Reply Retweet Like
Nick Tang May 1
It’s 2PM EST. 11AM PST. 6PM GMT. Welcome to . Our guests cut through the noise. To help you make smart decisions.
Reply Retweet Like
Nick Tang May 1
Let’s start with Q1 with from : Why should marketers use AI in a programmatic setting?
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
A1: In a climate in which we increasingly need to put the consumer first in , delivering to the right people in the right context is paramount. This requires gleaning patterns from massive data-sets + only / techniques can do this
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
Publishers, consumers and marketers will all benefit from -powered that delivers highly relevant, personal, timely messages.
Reply Retweet Like
Nick Tang May 1
Replying to @mediamath @pchalasani
A1. Is this the key to using ? Personalization?
Reply Retweet Like
Ben Sailors May 1
I think so. AI is supposed to just "get" you. It's supposed to understand you. Obviously it can be used in other ways, but if at least the illusion of personalization isn't there, the disconnect will make it mechanical and unapproachable for people.
Reply Retweet Like
Ben Sailors May 1
A1. It's also a way of working smarter not harder. It's important to focus on the right customers with the right message as said. It's also a way to get faster feedback, make tweaks and strengthen targeting with lower investment of time needed.
Reply Retweet Like
Mitch Lieberman May 1
A1: Not personalization, precision.
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
One more thought for A1 - Manually-designed rules are often wrong, and sub-optimal for all parties. techniques enable real-time bidding (RTB) platforms to learn to optimally bid for the right ad-opportunities.
Reply Retweet Like
Mitch Lieberman May 1
Replying to @SailorsBen @mediamath
Question: It is not yet 'easy' to use it requires lots of (clean data) proper process, good questions. This is more about working smarter towards the future, but it is still hard, no?
Reply Retweet Like
Ben Sailors May 1
Huh, that's fascinating. I totally get the point human made rules would be wrong. What's fascinating to me is the RTB part. What's the currency used and what determines the value? And bidding for what? Are the resources really limited?
Reply Retweet Like
MediaMath May 1
Replying to @SailorsBen @nickhtang
Real-time bidding refers to the buying and selling of online ad impressions through real-time auctions - plays a major role here in terms of efficiency and reliability. We should have some additional background on our blog we can share with you shortly!
Reply Retweet Like
Ben Sailors May 1
Replying to @mjayliebs @mediamath
I think this depends where in the stream you are. If you're accessing someone else's data and AI as a tool, it's relatively easy. If you need to collect the data, verify it, massage it, etc. then build an AI framework to make use of it, I'd say no, it's not easy.
Reply Retweet Like
Nick Tang May 1
As we continue discussing Q1, let’s head to Q2 for and Where are you focusing your AI efforts?
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
A2: Our top focus areas include: (a) Predict the probability that a consumer will respond to an ad impression (b) Predict the market clearing price of an ad impression on the ad exchanges (c) Predict consumers’ propensities to respond based on historical behavior
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
We’re deploying cutting-edge (Neural Network) models that leverage large amounts of historical to predict consumer behavior, whether in response to ad impressions, or inherent propensities to exhibit desired behaviors (purchase, video completion etc).
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
Yes, but it's a worthy challenge. Making marketing consumers love requires putting them in control of their experiences. Marketers must think consumer-first, and can help make advertising more valuable and personalized.
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
We are leading for . Our real-time bidding system receives over 300 billion ad opportunities daily from 70+ ad exchanges + we respond with ML-driven bid decisions within 10 milliseconds.
Reply Retweet Like
Nick Tang May 1
As you continue to discuss Q2, let’s move to for and Q3) Where do you see the roadblocks when marketers try to use AI to optimize their programming advertising?
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
A3: A key hurdle is to collect a sufficient volume of high-quality , of ad-impressions as well as consumers’ engagement with marketer sites and apps.
Reply Retweet Like
BrainBlender🤔🌐 May 1
This is an interesting question as now can not only create personalized advertising content as seen from Adobe system but will now move to targeted but will be faced by filters🌐
Reply Retweet Like
Nick Tang May 1
Replying to @mediamath @pchalasani
A3) Follow up. Do you think that this has gotten more difficult now?
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
Connecting human consumers across cookie and device changes can also be challenging, especially in light of concerns.
Reply Retweet Like
Roxana Nasoi May 1
If it's used in the correct transparent manner, so that people understand it is indeed serving them and not using them.
Reply Retweet Like
Roxana Nasoi May 1
search is already using . Even before AI, there was the semantic web (3.0) mentioned 10yrs in research prior to actual implementation. All working into directing the message to the end user in the right context. Clearly not performing 100% efficiently atm
Reply Retweet Like
MediaMath May 1
For sure - it is constantly evolving. The volume of high-quality data is key.
Reply Retweet Like
Nick Tang May 1
And the last question for at Q4) What impact do you believe AI will have on marketing over the next few years?
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
such as ’s will leverage the as well as technologies to bring ever closer to the holy grail of sending the right message to the right user in the right context.
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
A4: A confluence of two exponential trends will radically transform in general, and real-time bidding in particular, over the next five years.
Reply Retweet Like
MediaMath May 1
Replying to @nickhtang @pchalasani
These trends are: the surge in availability of about users interacting with media, ads and products + the rapid innovations in AI hardware, tools and frameworks, making them highly scalable and production-friendly.
Reply Retweet Like
BrainBlender🤔🌐 May 1
A4 will be huge as not only in personalization & predictive but also in construction of chain silos🤔
Reply Retweet Like
Nick Tang May 1
Finally, join us on sometime in June for (the time is the same. 2PM-3PM EST. 11AM-12PM PST. 6PM - 7PM GMT). Our guest? @roxansoi the CCO from . The topic and day? TBA. Thanks for joining us today everyone! Have a great week!
Reply Retweet Like