Disrupting Digital Marketing: Generative AI's Power and Privacy Implications

EVOLUTION OF DIGITAL MARKETING AND EMERGING ROLE OF AI

Over the years, digital marketing has seen a remarkable evolution reshaping how companies engage with their online audiences. Recently a new player has entered the field as technology develops at an unheard-of rate: generative artificial intelligence (AI)

Generative AI, a subset of AI, involves training machines to independently create original content based on a combination of existing recorded knowledge and work, spanning text, images, and videos. This cutting-edge technology is causing a paradigm shift across various sectors, particularly marketing, by unlocking unprecedented potential for strong content creation, personalization, and customer interaction.

In this blog, we delve into the transformative impact of generative AI in digital marketing, exploring its applications alongside the already well-established AI technologies that have been empowering marketers for years. From chatbots to predictive analytics, AI has revolutionised marketing by offering deeper insights into customer behaviour and preferences. As we uncover how generative AI is transforming the marketing landscape, we also address the crucial topic of data privacy concerns, ensuring marketers understand the ethical implications that come with harnessing the power of AI in their campaigns.

MACHINE LEARNING: HOW DOES IT TRULY WORK 

At its core, generative AI, exemplified by tools like ChatGPT or Midjourney, operates on the neural network framework. As a subset of machine learning, neural networks are modelled by the intricate connections of the human brain, simulating the interactions of biological neurons. Within the AI model, an input triggers the neural network to generate an appropriate response. For instance, ChatGPT has undergone extensive training to predict the next word in a sequence, honing its grasp of syntax, semantics, and contextual subtleties present in human language. Tools that create elaborate surrealistic or realistic images, profile pictures from your selfies or even entire websites, work the same way. However the input parsed and digested by the network will of course be visual.

To ensure an accurate response to Chat GPT user prompts, a training process was employed, using human intervention as one of the steps:

Step 1: Supervised Fine-tuning Model - This phase involves training the model using labelled examples provided by human annotators, mimicking appropriate responses to user prompts.

Step 2: The Reward Model - The AI is then trained to predict the usefulness of various responses generated for different user inputs, ranking them based on their predicted value.

Step 3: Reinforcement Learning Process - Building upon the initial training, the AI model is further refined to maximise rewards anticipated by the Reward Model, updating its predictions for future responses.

(Source: Scaling ChatGPT)

Through these meticulous steps, AI tools like ChatGPT achieve remarkable results in language generation. This prowess empowers businesses to engage customers, craft compelling content, bolster customer support, and create highly targeted campaigns that yield impressive returns on investment.

It is important to recognise that their development doesn’t end here. Interacting with users and receiving feedback on the generated responses allows these tools to continuously adapt and change. This is however one of the risks associated with generative AI tools. The quality of output can only be as good as the input. Large language models trained on Wikipedia, Reddit and other online user-generated sources will inevitably mirror existing biases in cyberspace. Users evaluating the responses of generative AI tools will introduce their own grammatical or contextual errors into the system, causing the quality of output to go down. This result is of course not only due to the human element: with opaque update processes it is hard to know what else impacts performance and in what direction. In short, while generative AI can be an immense help in shaving off time from content and asset creation, it won’t magically solve all of our marketing challenges just yet.

THE RISE OF GENERATIVE AI IN MARKETING

Disclaimers aside, by leveraging Generative AI algorithms, marketers beyond doubt can save time and become more resourceful with the help of specialist AI tools. Most of us seem to agree with this sentiment: one research conducted by Salesforce states that 60% of marketers claim Generative AI will transform their role and 51% are already experimenting  with it. (Source: Salesforce Generative AI Research)

In our blog, we delve deeper into these new tools and opportunities and examine how you might harness their power.

UNLEASHING CREATIVITY

Generative AI taps into boundless creativity by training on extensive datasets and absorbing inspiration. This empowers marketers with a diverse array of concepts and ideas that might have remained unexplored. While it can't innately invent entirely novel solutions, it jump-starts imaginative thinking, increasing the specificity and refinement of human ideas, their evaluation and combination, ultimately merging several ideas into a fresh one. Through this collaborative fusion, numerous concepts converge, giving rise to fresh perspectives and innovative ideas.(Source: Harvard Business Review)

Generative AI is poised to shape the future of creative expression, enhancing workflows and speeding up branded asset creation. The capability of AI algorithms to dynamically generate narratives and visuals opens doors for brands to communicate their message with unprecedented personalisation. Brands can now generate content that deeply resonates with target audiences, forges connections and cultivates brand loyalty.

Generative design is an iterative design exploration approach that employs an AI-powered software programme to produce a variety of design concepts that meet a set of criteria. As one of the most promising and prone to changing fields of digital marketing, it is expected to simultaneously explore many design possibilities and to enable mass customisation through enhanced efficiency.

Designers are set to become curators, meaning they will guide AI tools by setting goals, parameters and constraints allowing these tools to generate numerous designs in a fraction of the time needed usually. This new AI-driven design will offer new frontiers for immersive virtual worlds, but while still requiring skills such as creative and social intelligence, hard to replicate by AI alone. 

One question to consider faced with the opportunity to create and publish more, is quality over quantity. Noise and distraction is already an innate feature of digital marketing and social media. Marketers will have to exercise restraint and strategic thinking in order to maintain quality over quantity and create assets and communication that is meaningful, rather than only inescapable. (Source: Artefact Group)

PERSONALISATION AT SCALE

Generative AI empowers hyper-personalization by harnessing key prompts. This facilitates unmatched personalization, ensuring that each customer feels uniquely attended to. Unlike conventional personalization methods, which leverage location or messaging, hyper-personalization employs data analytics, AI and machine learning.

Numerous AI tools contribute to this evolution, including:

• Adcreative.ai: Automates creative content generation for targeted advertising campaigns, optimising visuals and copy for distinct audience segments.‍
• LogoAI: Analyses brand characteristics to generate distinctive logos, aligning with the business's personality and identity.‍
• Chat2Build: Constructs conversational chatbots for tailored user interactions, adapting responses based on individual preferences.‍
• Notion AI: Enhances project management through AI-driven automation, optimising workflows and automating tasks.‍
• Jasper.ai: Utilises AI analytics for customer success, improving engagement and satisfaction through personalised insights.‍
• MeetCody: Automates meeting scheduling by considering participants' availability, streamlining communication.

Predictive analytics makes a powerful tandem in combination with AI. This powerful combination uses the power of historical data, statistical modelling, and machine learning to make predictions about future customer behaviour and trends. The advantages of predictive analytics in marketing are compelling. It helps identify new opportunities by extrapolating customer trends from large and complex data sets, uncovering valuable audience segments, and pinpointing optimal timing for purchases. Moreover, it enhances resource efficiency by prioritising marketing efforts based on audience segmentation, timing, and placement.

However, while predictive analytics and AI offer valuable insights, human strategic thinking and contextual understanding remain pivotal. AI provides information, but human judgement shapes decisions, ensuring a balance between technological innovation and human oversight. (Source: MarTech)

GENERATIVE AI VS. MARKETING EXPERTS

The most daunting prospect in today’s market is the mere idea of a machine replacing a human in both his essential and complex tasks. While AI offers efficiency and innovation, the human touch remains irreplaceable, especially in areas requiring empathy, creativity, and critical thinking.

When you ask an AI tool like ChatGPT to compare the performance of a digital marketing expert and Generative AI, this is the response you get:

Is this true?

AI excels in automation and handling unstructured data, as it adeptly automates repetitive tasks, streamlining time-consuming processes. For example, it automates email scheduling based on user time zones and optimises campaign recommendations by identifying responsive customer segments. Another captivating application is AI-driven sentiment analysis, which categorises extensive reviews, customer feedback, and social media comments to gauge public sentiment towards a brand or product. (source: Optimove)

Yet, human marketers possess vital qualities like judgement, strategic insight, and emotional intelligence. They provide context, interpret results, and make nuanced decisions that AI tools may struggle to replicate. Additionally, AI lacks innate human creativity, often generating solutions within predefined criteria, unlike the human ability to think beyond the confines of algorithms.

REGULATORY CONCERNS

PRIVACY

The rise of AI introduces concerns about misinformation and job displacement. To address these, regulatory frameworks are being discussed and implemented worldwide. One of them includes the European Union’s proposed Artificial Intelligence Act, which aims to establish rules and requirements for AI systems. It also includes the GPT-3 usage guidelines to ensure responsible deployment of their Generative AI models.

A framework for AI regulation proposed by the European Commission details four levels of risk:

• Minimal or no risk — including systems such as AI-enabled video games or spam filters, which can involve generative AI
• Limited risk — including use of chatbots, with users needing to be clearly informed that they are interacting with such systems from the outset
• High risk — use of generative AI in critical infrastructure such as transport; educational contexts; law enforcement; hiring and recruitment; and healthcare robotics
• Unacceptable risk — including all systems that pose a clear threat to the safety and rights of citizens, such as social scoring and voice assistants that encourage harm (source: Information Age Magazine, https://www.information-age.com/how-generative-ai-regulation-shaping-up-around-world-123503911/) 

Embracing principles such as transparency, consent, and responsible data governance is crucial. Transparency builds trust by disclosing AI usage, data sources, and potential risks and consent gives individuals control over their data for generative purposes, aligning with privacy regulations. Of the same importance is the implementation of responsible data governance practices - with proper data anonymisation techniques and access controls, sensitive user information can be protected from misuse and unauthorised access.

WHOSE WORK & IP?

Yet another ethical question is concerned with copyright. After all, AI can not generate new copy, imagery, colour palettes or music without the work of countless human beings feeding into it. These artists and creators currently receive no compensation for contributing to large language models. While governing bodies are currenlty discussing ways to protect the rights of these people, the challenge and therefore the proposed solutions are also in very early stages. 

The UK government, amongst others, is currently working with users and rights holders on a code of practice on copyright and AI. The goal of this work is to create clear guidelines, ensure legal protection of creators but also remove barriers from AI tools ability to parse and remix. They aim to make licences for data mining more available and ensure there are protections for rights holders. (source: Gov.uk)

Hollywood is already feeling the consequences of decision making that failed to take into account people’s livelihoods and artistic licence. The SAG-AFTRA strike may well go into the early months of 2024, protesting, amongst other things, against the ridiculously low remuneration proposed for scanning actors’ likeness and then using them indefinitely, without limitations. (Source: Skynews )

Furthermore, a consideration for branding and marketing professionals, is the copyright of the visuals they themselves create with AI tools. According to current regulation it is unclear who really creates these assets and therefore ownership of them is murky at best. At the very least, we must ask: can you really trademark an AI created logo? 

In the dynamic AI landscape, balancing innovation and considering legal protection is vital. This approach not only mitigates privacy and copyright risks but also encourages responsible AI usage.

HOW DOES SMALLGIANTS TEAM USE GENERATIVE AI

“I do use generative AI very frequently, recently I have used it to find quite complex equations for a mathematical problem. The major disadvantage of generative AI in my opinion is the lack of control over generated content and overdependence on AI.” 

Jahangir Chowdhury

Front End Developer at SmallGiants

“As a digital marketing consultant, Generative AI impacts my work but it needs to be very carefully used. Although it can save you time creating ideas, ad campaigns, reports, etc. humans still need to adjust everything to the target audience that's using their product/service, and also make sure there are no copyrights or privacy data issues. 

It is a tool that definitely helps us open our minds to new ideas but we should always keep the human touch.”

Malena Berchot

SEO specialist at SmallGiants

“I will collaborate closely with Generative AI to provide engaging and tailored ad content.

AI might help with identifying highly specific audience segments based on behavioural data, interests, and preferences.

AI will be used to evaluate and understand massive volumes of ad performance data, providing significant insights into audience behaviour, conversion trends, and campaign efficiency.”

Kanita Čopra

Paid Media specialist at SmallGiants

“I’ve tested Beta apps for Photoshop and new Adobe Firefly (AI solutions for design)

I’ve done Images from scratch using “Text to Image” function which generate images from a detailed text description,“Generative Fill” which removes objects or paint in new ones in images, “Extend Images” which changes the aspect ratio of an image and “Generative recolor” which generates colour variations of a vector artwork. Generative AI will accelerate and improve design results and will help create new, specific visuals.”

Ines O’Brien

Brand Design Lead at SmallGiants

"Generative AI could help me in the circumstances whereby I need additional context or a different viewpoint of a particular concept or brand, using generative AI could help plug some knowledge gaps and help me create a better understanding and solution in my role."

Steven Silva

Account Manager at SmallGiants

"The two main workflows AI help me optimise is research and (copy)writing. When using pro tools with access to the internet, finding data and patterns can be significantly faster than it would be when done manually. Of course it is important to double check sources and take everything with a grain of salt - we all know AI hallucinates and results can’t always be trusted. However when used to collate sources and run data analyses, the time saved is crucial.

When it comes to writing and copywriting, ChatGPT and Bard are a valuable tool that provide a way to bounce ideas back and force test different variations of text at speed. It is essentially Thesaurus on steroids, helping work by spitting out sentence structures and options. However, when it comes to creative and unique solutions, the human touch is so far absolutely necessary. We need to be able to spot good solutions, and edit the materials the tool offers. Take for example straplines:  there would be a lot of identical or obnoxious sounding copy online (even more then now) if we didn't put in the work of selecting and embellishing for new approaches. Not to mention of course all the work and experience that goes into understanding your audiences, motivations, feelings and all those pesky human elements."

Orsolya Toth

Head of Strategy at SmallGiants

‍

The insights provided by our team demonstrate the transformative role that AI plays in marketing. From automating mundane tasks to providing valuable data-driven insights, AI empowers marketers to make informed decisions, drive better results, and create personalised experiences for customers. As the field of AI continues to evolve, we are excited to see how it will further boost our marketing strategies and redefine the roles of our team members in the future.