Why Meta’s Lattice Architecture Matters

Why Meta’s Lattice Architecture Matters

by Taylor Hart Artificial Intelligence Meta

Meta continues to demonstrate resilience. In 2022, the company took at $10 billion hit to its bottom after Apple’s privacy controls diminished Meta’s ability to target users with advertising that relies on third-party cookies. Since then, Meta has been developing better ad products that help the company develop targeted ads via first-party data, or the information that customers share with Meta on its apps such as Facebook and Instagram. Meta recently announced a potential breakthrough with Meta Lattice, which uses AI to improve ad targeting.

What Is Meta Lattice?

Meta Lattice is a new model architecture developed by Meta AI that improves the performance and efficiency of Meta’s ads systems. Meta Lattice is a high-capacity architecture that allows the ads system to understand new concepts and relationships more broadly and deeply in data. Meta says this benefits advertisers through joint optimization of a large number of goals.

Meta Lattice is also capable of generalizing its learnings across domains and objectives. This means that it can be used to improve the performance of ads across a variety of apps, such as Facebook, Instagram, and WhatsApp. Meta also says that Lattice can be used to improve the performance of ads even as the way people use Meta’s products changes.

The Potential Impact of Meta Lattice

Here’s a good before/after way to demonstrate the impact of Lattice.

  • Before: Meta had segmented advertising data. Meta maintained distinct datasets for ads presented in feeds, stories, Facebook and Instagram video reels, and other formats. This was also true for objectives, as ads targeting conversion, traffic, and video views were kept separate.
  • Soon: Meta will unify all this information using AI. With Meta Lattice, the platform will decipher patterns that result in enhanced ad performance across various ad formats, objectives, and types. This unification presents Meta’s machine learning algorithms with a larger dataset for learning. According to Meta, this implies better predictions for identifying the most suitable audience for your advertisements, resulting in more conversions through your Facebook Ads Manager — if Lattice does its job.

Meta Lattice is still under development, but it has already shown results. Meta says that in one test, Lattice was able to improve the click-through rate of ads by 10 percent. Meta Lattice is expected to be used to improve the performance of Meta’s ads systems in the future.

Potential Benefits

Here are some of the potential benefits of Meta Lattice:

  • Improved performance: Meta Lattice is able to improve the performance of ads by understanding new concepts and relationships in data.
  • Improved efficiency: Meta Lattice is able to generalize its learnings across domains and objectives, which makes it more efficient to use.
  • Faster adaptability: Meta Lattice is able to adapt to the fast-evolving market landscape, which makes it more effective at delivering relevant ads to users.

It is important that advertisers understand how Meta is developing new ad products. For instance Meta’s Advantage+ ads provide performance marketers advancing ad targeting tools.

As these AI-based platforms evolve, they may become more significant drivers of response, which could help you target the right audience for your offerings without needing to manually set the parameters of each campaign. However, we caution against any advertiser turning over the keys to AI. Advertising campaigns require human supervision to ensure that they continue to meet your objectives. AI advertising tools require human oversight to ensure that its output is accurate, free of bias, and consistent with your brand voice, among other needs.

Contact True Interactive

To succeed in the ever-changing world of online advertising, contact True Interactive. Read about some of our client work here.

Image source: https://pixabay.com/users/knfind-22056618/

How Google Is Transforming Advertising with AI

How Google Is Transforming Advertising with AI

by Héctor Ariza Artificial Intelligence Google

The Big Tech firms continue to change the online advertising landscape with AI. For example, on May 23 at Google’s Marketing Live event, Google announced new ways that the company is incorporating AI into advertising online. Google has been integrating AI into advertising for quite some time, as we have blogged. Google’s latest announcements make AI an essential tool now for any advertiser that works with Google. Here are some highlights:

Simplifying Campaign Management through Conversational Interfaces

Google said that it is introducing a new, natural-language conversational experience within Google Ads. This feature is designed to streamline the process of creating campaigns and to simplify the management of search ads.

The feature works this way: an advertiser adds a preferred landing page from your website, and Google AI summarizes its content. From there, Google will generate a range of (presumably more relevant) elements for your campaign, including keywords, headlines, descriptions, images, and other assets. Before deploying the generated suggestions, you have the flexibility to review and make edits. And, you can engage in a conversation with Google AI to enhance your campaign’s performance.

In the near future, Google says it will enhance automatically created assets (ACA) for search ads, which use existing ad content and landing pages to generate headlines and descriptions.

ACA also applies generative AI to create and adapt search ads more effectively based on the specific context of a query. For instance, let’s say a user searches for “skin care for dry sensitive skin.” With the assistance of AI, Google can analyze the content from your landing page and existing ads to generate a headline that aligns even more closely with the user’s query, such as “Soothe Your Dry, Sensitive Skin.” Google says this approach improves the relevance of your ads while staying true to your brand identity.

Helping Advertisers with Performance Max

Performance Max is a goal-based campaign type that allows performance advertisers to access all of their Google Ads inventory from a single campaign. According to Google, advertisers who use Performance Max have experienced an average increase of over 18 percent in conversions, all while maintaining a similar cost per action. However, it must be noted that at True Interactive, we’ve not witnessed these kinds of returns with Performance Max and remain very cautious about its value. That said, to improve the capabilities of Performance Max, Google is introducing generative AI technology. Google says that this addition will make it easier for advertisers to create customized assets and expand their reach with just a few clicks. Google AI will learn about your brand and populate your campaign with relevant text and other assets. Moreover, Google will suggest unique and tailored images exclusively generated for your brand.

This feature will also be integrated into the new conversational experience within Google Ads.

Introducing Enhanced Ad Experiences through Generative AI

During Google’s recent conference for developers (Google I/O – which is separate from its May 23 marketing event), Google unveiled new capabilities in generative AI that will promise to change search. This feature, known as search generative experience (SGE), uses generative AI to share answers to queries in the form of complete “snapshots” of content instead of providing a few rich snippets and links to websites for more information. Search can unfold as a series of questions and follow-up questions that the searcher refines (similar to ChatGPT prompt).

Earlier, Google had demonstrated how ads would appear both above and below this immersive new experience. On May 23, Google said it will conduct extensive experimentation to integrate search and shopping ads into the AI-powered snapshot and conversational mode. Additionally, Google will explore ad formats native to SGE, using generative AI to create relevant and high-quality ads tailored to every step of the user’s search journey (although how well Google will do this remains to be seen).

For instance, let’s consider a scenario where someone is searching for “outdoor activities to do in Maui” and then further narrows down their search to include “activities for kids” and ‘surfing.” In this case, they may encounter a fully personalized ad from a travel brand promoting surfing lessons specifically designed for children.

When search ads are displayed, they will include ad labels, with the “Sponsored” label presented in bold black text. This ensures a clear distinction between ads and organic search results, prioritizing a user-centric approach.

Three Implications for Advertisers

We advise advertisers to embrace AI (there really is no choice now – Google is bringing AI to your ad campaigns whether you like it or not). But do so carefully.

1. We Need Visibility and Control

Given that this year’s event was focused on artificial intelligence and automation (as expected), I would argue that both things empower advertisers to deliver incremental results, yet we need to have at least some degree of control over these, and more importantly, we need visibility into performance so we can clearly see what is working and what is not (they shouldn’t be mutually exclusive) and adjust our strategy accordingly.

2. Be Careful with ACA

I mentioned earlier that Google is enhancing automatically created assets (ACA) for search ads, which use existing ad content and landing pages to generate headlines and descriptions.

Incorporating generative AI into ACAs can potentially improve the relevance of advertisements. By using data from other ads and landing pages to improve query matching, ads can become more dynamic and effective, provided you are comfortable with relinquishing control to Google’s AI. this technology holds the potential to significantly boost the relevance of your advertisements, yet it may not be the ideal choice for heavily regulated industries or brands that adhere strictly to compliance standards. By fully handing over control to AI, Google gets the final say in your ad’s content. It’s important to remember that advertisements are subject to FTC regulations, and with ACAs, you will not have assurance over the message that will be displayed.

3. Treat Performance Max with Caution

At True Interactive, we are not quite ready to call ourselves Performance Max enthusiasts considering the limited control advertisers have over these campaigns and lack of visibility into performance. Although we would certainly encourage everyone to try it (if they have the money to spare), we would also recommend exercising special caution, as this campaign may cannibalize traffic from other campaigns being run under the same ad account (such as Search Brand campaign), and therefore hurt the overall performance of Google Ads.

Contact True Interactive

At True Interactive, we advocate on behalf of our clients. We are monitoring these developments closely and assessing how to incorporate conversational AI. Contact us to learn how we can help you succeed in all forms of digital advertising.

Is a ChatGPT Killer Coming from Google?

Is a ChatGPT Killer Coming from Google?

by Mark Smith Artificial Intelligence Google

It looks like the AI arms race is heating up.

Google is expected to announce soon the launch of a competitor to ChatGPT, the generative AI tool that has shaken the technology and business world.

ChatGPT is the product of OpenAI, the company that produced Dall-E, which uses AI to create images. ChatGPT is one of many chatbots designed to respond to queries from people by providing richer, more detailed, and more human-sounding answers than their predecessors. The incredibly slick bot uses AI to do everything from write copy to answer search queries to write code.

Some technology/business watchers have speculated that ChatGPT is a threat to Google Search. That’s because ChatCPGT responds to queries with a single answer that synthesizes information, which could upend how Google Search provides answers with links to information. Moreover, OpenAI is receiving deep funding from Google competitor Microsoft, which is incorporating the tool in its products, including, reportedly, Bing Search.

Well, Google has not taken the rise of ChatGPT lightly. Google’s parent Alphabet announced its quarterly earnings recently, and Alphabet CEO Sundar Pichai said that Google is working on its own form of smart search. He said that “very soon people will be able to interact directly with our newest, most powerful language models as a companion to Search in experimental and innovative ways.”

Apparently “very soon” is almost here. On February 8, Google is hosting an event on YouTube, which will revolve around “using the power of AI to reimagine how people search for, explore and interact with information, making it more natural and intuitive than ever before to find what you need.”

Releasing an answer to ChatGPT is not far-fetched. Google has contended that it has been developing AI-powered search technology for quite some time but is not ready to share it publicly. Examples of Google’s AI-driven products include a chatbot language model called LaMDA (Language Model for Dialogue Applications), an image-generation AI called Imagen, and MusicLM, which translates text to music.

But when OpenAI seized the narrative about generative AI by releasing ChatGPT in November 2022, reportedly Google went into “Code Red” mode and began fast tracking the development of various AI products.

Google has reportedly asked employees to test potential ChatGPT competitors, including “Apprentice Bard,” which makes it possible ask questions and receive detailed answers similar to ChatGPT. Details about Apprentice Bard’s functionality were leaked to CNBC, which reported:

Apprentice Bard looks similar to ChatGPT: Employees can enter a question in a dialog box and get a text answer, then give feedback on the response. Based on several responses viewed by CNBC, Apprentice Bard’s answers can include recent events, a feature ChatGPT doesn’t have yet.

Meanwhile, recently it was reported that Google has invested $300 million in AI startup Anthropic, which is testing a rival to OpenAI’s ChatGPT. Anthropic’s language model assistant, Claude, hasn’t yet been released to the public, but the startup told Bloomberg it planned to expand access to the chatbot “in the coming months.”

One way or another, Google is gearing up to respond – although the impact of a chat-powered AI tool on Google’s paid search business remains unclear.

At True Interactive, we are following these developments closely. We already use Google’s AI-powered ad products. Based on our experiences with AI, we strongly advise that businesses experiment with these tools carefully.

We also recommend that businesses keep people involved in managing AI (or any technology). People are needed more than ever to ensure that AI does its job well. For instance, our experience has consistently shown that automated ads powered by AI underperform without people involved to monitor and modulate them when necessary. The same is true of generative AI. These tools sure sound confident when they present information, but they make mistakes, and they are notoriously biased. They are nowhere near the point of being self-sufficient.

We’ll follow the developments from Google and report back on our blog.

Contact True Interactive

To succeed in the ever-changing world of online advertising, contact True Interactive. Read about some of our client work here.

What Is Machine Learning?: Advertiser Q&A

What Is Machine Learning?: Advertiser Q&A

by Mark Smith Artificial Intelligence

Machine learning is affecting the way businesses operate – including how they advertise. Google, for instance, uses machine learning to help businesses optimize the performance of their search ads. But not everyone understands exactly what machine learning is. We thought we’d take a moment to break the topic down and answer some common questions.

What is machine learning?

Machine learning is a type of artificial intelligence (AI) in which computers literally have the ability to learn, and subsequently make increasingly more intelligent decisions. The learning happens when a computer program accesses and analyzes data — data in amounts generally too vast for humans to read through quickly or accurately. The computer looks for patterns in the data and learns automatically, without human assistance. Spotify, for instance, uses machine learning to understand the musical tastes of its subscribers in order to recommend songs that are more likely to match their interests.

Is machine learning the same as AI?

Machine learning is one aspect of AI. AI in fact encompasses many things, including:

  • Natural language processing, or NLP, a technology that equips machines to interpret what people say in words or text. Advanced NLP not only deciphers speech, it teases out context and detects nuances like sarcasm.
  • Chatbots, the programs operating inside messaging apps or on websites, which allow consumers to accomplish simple tasks like e-commerce transactions.
  • Neural networks, which are AI programs that use the human brain as a model. Neural networks incorporate aspects of AI like NLP in order to perform duties such as recognizing handwriting.
  • Dynamic pricing, in which programs use consumer data to set prices that are most likely to ensure a sale given various factors.
  • Content Curation, an aspect of AI that can be used to figure out what specific goods to recommend to consumers based on data about that consumer.

In short, AI is a supercategory describing and encompassing the many ways computers emulate the way people think and act. In the vast universe of AI, machine learning is one subcategory.

How are businesses using machine learning in online advertising?

Machine learning can help businesses test the effectiveness of different forms of advertising. Not only can it prioritize ads that are doing better, machine learning can even pinpoint what ads are performing best at certain times.

Google’s efforts to stay competitive in a field shared by powerhouses such as Amazon and Facebook exemplify an effective use of machine learning. As we recently discussed, Google’s responsive search ads and responsive display ads make it possible for advertisers to enter multiple headlines, descriptions, and logos. Machine learning figures in by automatically testing various combinations of these factors and highlighting the ads that perform best.

What are some tips for succeeding with machine learning?

Like any tool, machine learning can be used wisely—or abused. To glean the best results, keep in mind that:

  • Machine learning is not a replacement for humans. Use it as a complement to human judgment.
  • Machine learning is a dynamic field. This is not a topic you can learn once and then consider it mastered. Machine learning tools evolve quickly, with new ones coming along seemingly on a weekly basis. Stay on top of the changes.

Contact True Interactive

Curious as to how machine learning can help you with your digital advertising? Contact us.

Photo by Franck V. on Unsplash

Get Ready for AI Everywhere

Get Ready for AI Everywhere

by Mark Smith Artificial Intelligence

In 2019, artificial intelligence (AI) will make digital advertising more targeted, thanks in part to the efforts of Google. But marketers will need to invest more time and effort to make AI pay off.

It’s clear that AI is essential to Google’s growth. In February 2018, CEO Sundar Pichai said AI is more profound than electricity or fire. A few months later, he published a statement of AI principles in which he outlined seven ways Google will use AI (and ways that Google will not). The post focused on the importance of using AI for social good. Pichai did not mention using AI for advertising, but Google is certainly applying AI to make advertising smarter.

For instance, in 2018, Google launched a number of products that use machine learning (a form of AI) to improve online advertising performance. I recently blogged about one such product, responsive search ads. As I noted, responsive search ads make it possible for advertisers to enter multiple headlines (up to 15) and descriptions (up to four) when creating a search ad. Then Google Ads applies machine learning to automatically test different combinations and learn which combinations perform best. In addition, per Google, advertisers can add a third headline and second description to your text ads, and your descriptions can have up to 90 characters.

2018 was just a warm-up for what’s to come in 2019. Businesses demand more accountability and ROI from their online ad spend, and AI does just that. I expect Google will focus more on using AI to make YouTube more effective. Google has already injected AI into YouTube with features such as maximize lift, which is a smart bidding tool that automatically adjusts bids at auction time to maximize the impact a company’s video ads have on brand perception. Maximize lift is supposed to help businesses reach people who are most likely to consider their brand after seeing a video ad.

One concern we often hear from advertisers is that YouTube is not as useful for direct-response campaigns as it is for brand building. In 2019, we’ll see the emergence of tools that do a better job targeting video ad content to people who are in shopping mode and ready to buy as Google makes YouTube more of a lower-funnel platform.

AI will make online advertising better. But AI will also require marketers to invest more time and energy to make it pay off, as I discussed in my post about responsive search ads. It’s important that businesses understand its uses and requirements. For more insight, contact True Interactive. We help businesses maximize the value of their online advertising and understand where the industry is headed.

Image source: https://pixabay.com/en/artificial-intelligence-robot-ai-ki-2167835/