What you need to know:

• Anna Russell, Polynomial: It can spout absolute nonsense with great authority.
• Anna Bohler, IAG: It's technology and technology is never perfect.
• Dean Norbiato, Kia: Cancelling the drudgery and accelerating creativity
• Liz Miller, Constellation Research: 'Sorry ladies, those highly sexualised images don't belong to you.'
• Adam Good, WPP: All businesses are going to use it.
• Scott Brinker, Chiefmartec: Orders of magnitude increases in the production of content and experience.
• Anna Koleth, Tealium: Unauthorised access or use could result in loss of productivity and greater inaccuracy.
• Rajan Kumar, The Lumery: It's not going to give you all the answers, it's impossible, it can’t.
• Kim Oosthuizen, SAP: Easy to use, explains complex issues simply, and it's free – for now.
• Lisa Collins, Akkodis: Moral and ethical obligations.
• Lucio Ribeiro, AI lecturer: Marketers will need to become "prompt engineers".
• Read all Andrew Birmingham’s coverage and analysis of CX, martech, ecom, adtech and regulation in MI3’s Tech section: https://www.mi-3.com.au/tech

Open AI's ChatGPT exploded into commercial consciousness late last year by achieving something that eluded all of its predecessors – its utility and ease of use exposed business leaders across the globe to the extraordinary promise of AI.

In the first five days after its release in November 2022, more than a million users logged onto the platform. It took Airbnb six years to meet that benchmark, Facebook one year, Spotify five months, and Instagram two and a half months.

According to marketing technology doyen and US-based editor-in-chief of chiefmartech.com Scott Brinker, “ChatGPT triggered a pivotal ah-ha moment, where almost everyone recognised that a major disruption had begun."

He describes 2022 as an inflection point for generative AI with multiple implementations shifting from the lab to production. And he cautions we need to prepare for another exponential growth curve in marketing technology innovation once the current economic hiccup washes through the system.

Follow the money

Well before its introduction to consumers, ChatGPT had already attracted the bluest of blue-chip technology investors including LinkedIn founder Reid Hoffman, and venture capitalist Peter Thiel. OpenAI, the company behind the platform, is led by Sam Altman, the former president of Y Combinator, one of the more famous technology startup accelerators which previously helped spawn platforms such as Stripe, Twitch, Dropbox and Reddit.

So what is ChatGPT precisely? Let’s go to the source. We asked ChatGPT to answer the same question three different times in three different voices.

First up, we asked it to describe itself in language a 12-year-old would understand.

• “ChatGPT is a computer program that can understand and respond to things you type to it, like a really smart robot! It can answer questions, help you with homework, or even tell jokes. It's like having a helper that you can talk to anytime you want!”

Simple enough for you?

We then posed the same question to the software but asked for the answer in language that mimicked the style of a Donald Trump speech.

• “Folks, let me tell you, ChatGPT is the best thing since sliced bread. It's a computer program that's been trained on a ton of my speeches, so it knows how to talk just like me. It's gonna revolutionise the game, believe me. It can generate new speeches, statements, tweets, you name it, all in my style. It's gonna be huge, just huge. It's gonna make America great again, and it's gonna do it FAST. So don't believe the fake news when they tell you it's not good, it's the best, the absolute best.”

Finally, we asked it to describe itself in the language of chief marketing officers.

• “ChatGPT is a state-of-the-art language generation model developed by OpenAI. It is based on the transformer architecture and is fine-tuned on a diverse set of internet text to generate human-like text. It can be used in a variety of applications such as language translation, text summarisation, content creation, customer service and more. It's a powerful tool for automating repetitive, high-volume tasks and can be integrated into various business processes to improve efficiency and customer engagement."

A better mousetrap

ChatGPT is just a tool, part of a set of technologies that can generate new content based on patterns learned from existing data. Think of it like a computer program that can create new paintings, music, or even writing in a similar style to what it has seen before, but not exactly the same. The goal is to create new, unique content that was not specifically programmed by a human. (That definition was created by ChatGTP - we asked it to use language a college student would understand.)

According to Gartner, "Venture capital firms have invested over $1.7 billion in generative AI solutions over the last three years, with AI-enabled drug discovery and AI software coding receiving the most funding."  (That doesn't include the $10bn Microsoft just announced it is tipping into ChatGPT maker Open AI.)

“Early foundation models like ChatGPT focus on the ability of generative AI to augment creative work, but by 2025, we expect more than 30 per cent — up from zero today — of new drugs and materials to be systematically discovered using generative AI techniques,” according to Brian Burke, Research VP for Technology Innovation at Gartner. “And that is just one of numerous industry use cases.”

Not the messiah, just a very naughty bot

Generative AI tools are as susceptible to the rule of garbage in/garbage out as every technology that came before them.

ChatGPT itself is a gifted grifter. It speaks authoritatively about topics with the air of an expert, says Polynomial director Anna Russell, a data science specialist. It has already passed Wharton Business School MBA exams and medical licensing tests.  If you are down for a bit of anthropomorphising, it's also a sociopath that routinely steals other people's work and passes it off as its own, or lies about sources even when its instructed not to.

Christopher Graves, President & Founder, Ogilvy Center for Behavioral Science writing on LinkedIn, for instance, warned, “Beware if you are using ChatGPT to help find scientific studies or even cite studies to support its assertions. It will not hesitate to lie and fabricate them."

He then shared an exchange with ChatGPT which he said fabricated very specific study titles and authors and then confessed: “I asked it to not lie and again to seek actual studies. It lied again.”

Nor is ChatGPT alone when it comes to the kind of behaviour that would make a compliance executive blush. Image AI tools routinely "create" new work by plundering the creativity of humans to an extent that would struggle to withstand judicial review in civil courts under long-established copyright regimes.

Into the weeds

But for every problem it creates, there is also the promise of greater efficiency, transformational change, and order of magnitude productivity and efficiency improvements.

The Australian and international executives Mi3 spoke with universally acknowledged the power of ChatGPT to not only deliver on what it promises – but also to reframe the conversation about the role of AI in business and society. Yet, they are also alive to the risks, roadblocks, pitfalls, and legal and ethical potholes we are all about to experience.

In this report we hear from B2B and B2C marketers, marketing technologists, and industry analysts including; Kia’s Dean Norbiato, Tealium’s Anna Koleth, The Lumery’s Rajan Kumar, WPP’s Adam Good, Chiefmartec’s Scott Brinker, Constellation Research’s Liz Miller, Lucio Ribeiro, IAG’s Anna Bohler, Akkodis’s Lisa Collins and SAP’s Kim Oosthuizen. We start with a data science specialist Polynomial's Anna Russell.

The data science specialist: Anna Russell on her first Impressions

Data analytics and data science specialist, Anna Russell has worked with the likes of David Jones, Mecca, Country Road, Westpac, Toyota, Blackmores, News Corp and Deloitte via her consultancy Polynomial where she is a director.

“There’s no doubt that ChatGPT has great potential value for a range of utilitarian tasks that require coherent language patterns and a degree of logic. A colleague of mine has been gleefully using it to add descriptive comments to his code, a task he abhors doing himself. ChatGPT comments up the code quickly and cheerfully, and generally quite accurately – better than most of us would do voluntarily, to be honest," per Russell.

She describes the platform as good at answering clear questions that involve gathering and presenting back a body of information to a user who has a decent grounding in the topic, but cautions that users need to apply a strong B.S. filter to the outputs the platform feeds them.

Caveatemptor.ai

“That brings me to the most concerning aspect of ChatGPT. Because its language models are so sophisticated, it can spout absolute nonsense with great authority," Russell warns. 

“It constructs an answer that appears to be conclusive but is in fact a meshing together of various data points and some logic, occasionally producing utterly daft results." 

Russell quotes an article by John Hawkins, chief technology officer at attention measurement firm Playground XYZ: “The danger with this authoritative voice to the output, is that the fundamental goal of the underlying language model is to produce output that sounds like a plausible human response, rather than conveying genuine knowledge," writes Hawkins. "In a sense it has been trained to act like a conman; to skilfully pretend to know things that it is clueless about. As impressive as it appears, there is no guarantee that the string of grammatically sound and topically relevant words it produces will hold any relation to truth.”  

It's this aspect of ChatGPT’s construction that significantly constrains the scope it can be used within, Russell says.

“For example, one would not consider using it in any clinical diagnostic setting as a means of eliciting symptoms or diagnosing illness.”

Russell argues it’s also a somewhat dangerous replacement for search: “It delivers a filtered and interpreted subset of information, that may be biased by the manner in which it was trained. This is worrying in that if used as the ‘search’ function on a topic, the user will receive a curated and biased set of content that may result in them forming biased opinions on a matter, which may be detrimental to themselves or others.”

By comparison, standard search contains far less implicit bias, “although of course makes the user work harder to extract and synthesise content".  

She acknowledges however that ChatGPT provides an important benefit beyond its own utility – it has accelerated conversations around issues such as the ethical use of AI.

“I love that ChatGPT has made all manner of non-technical people interested in understanding more about the capabilities of advanced language models and AI. It’s certainly generated a lot more ‘chat’ around AI ethics and the appropriate uses for non-human ‘entities across various sectors," says Russell, which can only be beneficial in fuelling robust discussion about the appropriate uses and constraints that will need to be applied to AI.

Curiouser and curiouser: Anna Bohler, Manager, Personalised Marketing, IAG

The consistent message from marketing leaders Mi3 spoke with about generative AI is that even the most impressive tool kits are less important than the people who use them: Generative AI won't simply transform marketing, it will change the kinds of people who are attracted to the profession.

That means brands must commit to investing in the capabilities required to leverage any AI investment.

According to Anna Bohler, Manager, Personalised Marketing at IAG, one of the most important qualities brands will need in their teams is curiosity.

"Curiosity is at the very core of teams that are going to be successful over the next 10 years versus teams that won't be. And curiosity is what we as a culture very much strive for. It’s curiosity about how we use it [tools like ChatGPT], and why is it answering things the way it is,” she says.

The brand marketer: Dean Norbiato, GM Marketing, Kia, on releasing team creativity

Kia marketing boss Dean Norbiato is keen to harness smarter machine capability to build brand and free-up thinking time across a lean team.

“How can a level of automation free up my staff to be more creative and to focus on … the brand being more salient and cutting through?”

He hopes tools like ChatGPT will accelerate the desktop research that goes into formulating briefs, or case studies, or determining things like the most prevalent influences in a certain category to enhance a brand or a certain campaign or product direction. 

“That breadth of desktop research just sucks your time having to do it. If something like that can be automated and done a lot quicker, that allows them to get the results and then start ideating themselves.”

He wants automation to do the administrative legwork freeing staff to think creatively.

"A creative idea can come from anyone in the business, and unless you have your ears open, you're going to miss it," says Norbiato.

"Bringing up all staff to be able to think creatively, and not be reduced to just doing administrative tasks would be a huge breakout for the key marketing team. We do a lot of reporting to global as well, so any level of automation we can bring would be fantastic."

The analyst: Liz Miller, VP and principal analyst, Constellation Research on why now? 

One of the biggest, earliest contributions of ChatGPT is that its helped people finally get AI says Liz Miller, VP and principal analyst, Constellation Research who previously spent more than a decade as the CMO of the CMO Council in the US.

“It’s demystifying the fear-factor around AI. People are investing hard money, patience, belief, you name it, into it and we have to start getting everyone involved to understand that there should be no fear when it comes to AI. No, it's not trying to take your job. No one's going to try to push you out of the pod bay doors, but there needs to be purpose behind what we are asking that AI to do.”

Miller says AI in and of itself isn't going to solve the world's biggest problems. “The person asking the question and how well they get to ask the question, and then synthesise all of the answers together, that's the person who is going to be the one who solves the world's biggest problems.”

There is not a lot of value in asking AI simple questions – instead the more data you give the machine, the better the guardrails and the more clearly defined the desired outputs, the better all those multiple forms of AI and multiple forms of algorithms can be layered for a recommendation.

According to Miller there is simply no way humans to truly understand the massive stores of data that true AI and autonomous computing utilise. 

Law of unintended consequences

Miller says marketers must appreciate what they are getting themselves and potentially their customers into with AI use, using imaging tool Lensa as a standout example.

Its terms and conditions make clear that anything Lensa creates remains the property of Lensa and the firm can do what it damn well pleases with it, says Miller. That's fertile ground for unintended consequences.

According to Miller: “Anything that you upload into [Lensa] is your property and you can remove it, but anything that's created by the AI belongs to the company. That means that image of you as an astronaut that you've just had to make your social media profile does not necessarily belong to you, and it can be used in any way it likes.

She references the example of where women have had highly sexualised images of themselves created by Lensa the when they asked it to make them look like an anime character.

“That doesn't belong to you ladies. I know it offended you. But guess who that belongs to? It can show up anywhere Lensa so chooses to have it show up because its terms and conditions are so broad – and horrifically so.”

The marketing technologist: Adam Good: Executive Director, Marketing Technology, WPP, on the new wave of disruption

“This is actually disrupting the whole AI landscape,” according to Adam Good: Executive Director, Marketing Technology, WPP. “That's where you really start to see that competitive nature of of the tooling,“ he says noting that Microsoft recently massively increased its ChatGPT investment, pouring another $10bn in Open AI.

According to Good, Microsoft has been considered a laggard behind Google, Apple and AWS in regards to AI. But its investment in OpenAI, which runs on Microsoft’s Azure cloud, it has the opportunity to help it catch up.

He also believes ChatGPT is a game-changer when compared to the often piecemeal approach companies have taken when developing bots.

“This is certainly a big step forward. When version four comes out with the subscription layer it's going to be a tool that a lot of businesses, indeed all businesses are going to be using." 

Like Kia's Norbiato, Good says the likes of WPP will harness generative AI to take humans off grunt work and free up creative firepower and he thinks the release of ChatGPT 4 and the launch of a subscription based professional version could vastly reduce admin.

Once organisation have the ability to easily plug in internal databases for compliance and regulatory approval, for example, they will be able to accelerate insight research and content creation. He says WPP has already experienced the benefits, using AI tools to eliminate time-consuming work for its NZ PR agencies.

Early forays

AI is also already delivering success on the creative side in North America. True Patriot Love Foundation, in partnership with Wunderman Thompson designed a campaign to educate Canadians about the importance of remembering those who served and those who continue to serve in the Canadian Armed Forces. It used AI image generator MidJourney to bring to life what it describes scenes from the battlefield, hospitals, and life in the trenches described in serving members letters home to loved ones.

A second campaign, GoFundMe with AKQA San Francisco, created 'Help Changes Everything,' taking its inspiration from a street mural but bringing scenes from the mural to life using AI generated art.

The lumascapes guy: Scott Brinker, editor-in-chief, Chiefmartec on why generative AI is really different

Marketers are already familiar with the impact of AI which, if it’s not being developed by their own IT or analytics department, is nonetheless already baked into the marketing technology apps they use – and the adtech services they source from agencies.

But generative AI is set to provide a much bigger leap in productivity says Scott Brinker, the editor-in-chief of the Chiefmartec blog who has tracked the rise of the sector from infancy to thousands of companies today.

“Generative AI was, up until very recently, somewhere between purely hypothetical and ‘oh, that's an interesting science fair project'", he says.

“With ChatGPT 3 and Dall-E, with all the different variations those large models out there, we have  crossed the tipping point where it starts producing stuff that people recognise.”

Acceleration

While the technology remains immature, it is now also commercially viable, according to Brinker.

And the discipline is accelerating. “GPT 4 is due out here any month now. Google has their engine, they're probably going to get ready to release it. We'll see the improvements and the start that exponential curve.”

Brinker is most excited about the capacity of ChatGPT and other tools like it to remove the constraints on content creation and distribution.

“You talk about personalisation, but a lot of personalisation is just like mail merge, we're substituting a few simple things, then we're done," he says.

"If you really wanted to personalise a message for a given prospect you would need humans who can do all this research, understand it, match the data, come up with a strategy. Then you need a good writer to turn it into copy. The cost effectiveness might work if you are talking about a $30m, $40m or $50m product. But for almost all other businesses, that was just impractical. “

According to Brinker generative AI will change that. It could also bring forward web3 far more quickly. 

“One of the things that is holding back the metaverse is the cost involved in crafting digital worlds. It's expensive. But now with generative AI, you just describe a set of stuff you want and it dynamically generates things for you. That’s an orders-of-magnitude of increase in the production of content and experience." 

Brinker says he can barely begin to describe the ways in which generative AI changes the world of marketing. “It’s going to really disrupt it, because everything we've done in the past has been around constraint of content.”

The content marketer: Anna Koleth, Head of Product and Content Marketing, Tealium, JAPAC on the legalities 

AI initiated disruptions in the marketplace create innovation opportunities, according to Anna Koleth, Head of Product and Content Marketing, Tealium, APAC and Japan.

“These tools present risks, and impose a learning curve and administrative overhead to some degree.”

Yet they also present opportunities to accelerate an organisation's digital transformation journey, and build innovation capability and smart companies will learn to take advantage of both.

But is will also be important to navigate the pitfalls.

According to Koleth, legal considerations are key, with issues such as copyright needing to considered from the outset before the tools are ever adopted, and effective risk mitigation strategies need to be put in place.

“Here human oversight becomes paramount to ensure the integrity of the work," she says. "I also think that tools that are already available, for example that detect plagiarism, can actually be adopted as part of a martech stack to be able to assist as a final check to ensure the integrity of a particular work.”

Koleth believes the kinds of approaches companies will take will include building AI centres of excellence, further investment in workforce reskilling, and the development of cost effective controls to guard against unauthorised access or use of these generative AI tools.

"Unauthorised access or use can result in loss of productivity of those tools, as well as inaccuracy in those outputs,"  per Koleth.

Like every technology tool before it, generative AI platforms will adhere to the law of GIGO – garbage in, garbage out.  To that end Koleth says organisations will need to define and develop a consolidated brand and product value proposition, and ensure that it is clearly understood across relevant teams.

Authenticity

It’s often said that once you can fake sincerity you’ve got it made. AI has a long way to go. 

For her part Koleth says that in the current digital landscape, authenticity and personalisation remain key.

“Trust in the brand still remains critical to commercial viability, trust based personalised engagement is going to continue to be increasingly important to commercial performance," she says. "Therefore, in the deployment of these solutions, marketing teams must first consider a strategy for their use, and ensure that it's complemented by effective human intervention to embed that authentic experience within every discrete customer engagement."

The marketing and technology partner: Rajan Kumar, CEO The Lumery on how ‘intelligent’ is still a bit dumb...

“It’s only as good as the question you ask, but that’s true for marketing teams full stop, whether you're briefing an agency on a on a creative campaign, or you're building a new strategy, or you're looking at new data and wanting to analyse it,” says Rajan Kumar, founder and CEO of martech consultancy The Lumery.

The emergence of generative AI will force marketing teams to think more strategically, he suggests.

“It's not going to give you all the answers, it's impossible, it can’t. It can get you a hell of a lot of the way there and give you the kind of framing, the bones of something, thought starters. And it can do some of the things that perhaps you wouldn't want to spend your time doing, “ says Kumar.

While these are all worthy outputs, Kumar says marketers will still need to build the overarching strategy and the context. He thinks AI tools for marketing and communications teams are best thought of as efficiency enablers. 

“That's where they play the best role. Send-time optimisation for email is the perfect example. The tool will figure out when to send all of us an email. Brilliant, because the alternative to that is a spreadsheet with a bunch of segments, and someone filling that out manually. “

The trick is to identify the opportunities where AI does this best, he says.

The innovation specialist: Dr Kim Oosthuizen, SAP on where ChatGPT fits into the AI mix

Chat GPTis just the latest and most notable of a set of generative IT tools and its fit in the spectrum on the market that can be applied to multiple media types. Dr Kim Oosthuizen, Innovation Principal, Customer Transformation Advisor at SAP unpacks where it sits in the spectrum.

First up it is important to recognise that artificial intelligence (AI) is an umbrella term that refers to any technology capable of acting intelligently, says Oosthuizen.

As such it covers a broad set of ever-evolving technologies including Machine Learning (ML), Deep Learning (DL), Computer vision and digital assistants, for example.

“Generative AI is a subset that focuses on creating and generating new content such as digital images, video, audio, text, or code," while numerous AI technologies falling into this category, per Oosthuizen, including:

• Images generates images from text (such as DALL-E, Stable Diffusion and MidJourney)
• Text creation via large language models (LLM) summarises and automates content (such as Grammarly, JasperAI, ChatGPT)
• Audio summarises, generates or convert text to audio (such as Murf, Listnr)
• Video generates or edits videos (such as Syntheys, Synthesia) and Code creation (such as Codex, GitHub). 

“All generative AI requires a human to enter a combination of words into the model to produce an image or written text," per the SAP exec.

In particular, ChatGPT fits into the large language model space space which Oosthuizen describes as the backbone for generative AI applications. “What makes ChatGPT so popular is that anyone can use the technology; it is easy to use, can explain complex issues in simple language and for now, is currently free to use.”

Other tools, such as JasperAI, and Creaitor.ai, offer paid services with a specific target market of content creators, bloggers, agencies and writers, whereas ChatGPT is for everyone. 

That open-source heritage and the (currently) free availability makes the tool very attractive to use across many domains.

She quotes user data from Statista’s Katharina Bucholts to make the point. “It is so popular that since its release in November 2022, more than a million users have logged onto the platform in the first five days. It took Airbnb six years, Facebook one year, Spotify five months and Instagram 2.5 months to reach a million users.”

Oosthuizen  told Mi3 that Generative AI  is likely to significantly impact creative work, especially art, music and writing. “Leaders need to be aware of how all AI tools will change the way of working, business models and operations. AI is not going away, and it is important to understand the best approaches to incorporate it into business.”

The B2B marketer; Lisa Collins, VP marketing and communications, Akkodis, on transparency and ethical conduct

“IP protection is one of the big challenges,” says Lisa Collins, VP marketing and communications at Akkodis.
While the impact on employment always dominates the introduction of new technology, much less consideration is given to how to protect the work of creators, she warns.

“Transparency is key and people really want to believe that they are being paid fairly," according to Collins.

“How do we, as marketers take the high ground here, and really take moral and ethical stance on this? I think that's really important.”

Legislators and ethicists are already playing catch up she says, and business can’t really afford to wait for them to make up the lag.

“We need to be on the front foot and really let our audiences know when we are using these tools and how we are using them.

Collins experienced the rise of AI during her stint as a marketer at Amazon which helped kickstart the commercial rise of the technology when it launched Sagemaker. She argues there need to be guidelines and some criteria around how brands use it. 

When companies buy new technology they go through a rigorous process involving legal compliance, but as new and often free tools emerge, companies, or at least their employees may opt to skip the inconvenience of oversight.

“What's the impact to customers? What's the impact to the business? How are we collecting data, these are all the fundamental obligations and requirements that we already have as marketers that now have to be now applied in this new context?"

According to Collins, “We recognise that this is a moral and ethical obligation and it's a business protection as well.”

The lecturer: Lucio Riberio, former consultant and university lecturer in AI on the concept of prompt engineering

ChatGPT has done more than introduce a whole generation of marketers to use-friendly AI, it has also led to the rise of a new discipline of capability – prompt engineering.

In simple terms this refers to the ability to interrogate a generative AI tool in a way that actually returns the most use results.

According to Lucio Riberio, former consultant and University lecturer in AI, the power of ChatGPT was revealed to him when he used it to try and avoid paying a traffic fine.

“I asked it to write me a fine appeal - and I have to tell you, I write relatively well," says Ribeiro. "They [ChatGPT] killed me, that was when I got their model.”*

As he has experimented with the platform, Ribeiro says he is finding more literature (admittedly some of it of dubious quality) on the concept of prompt engineering.

“I’m really struck by the concept of prompt engineering because is it the engineering of dialogues," he says.

Riberio thinks the ability to write precise and complex queries into tools like ChatGPT – prompts – will become a core capability in marketing departments as generative AI takes off. The technical skills won't be enough, he suggests, the 'prompt engineer' will also need an appreciation and understanding of the product proposition and the customer need.

Which signals another looming capability challenge.

Thanks ChatGPT.

*No word yet on whether the appeal was successful.

Post script: UPDATE: Lucio won the appeal. The money he saved could pay for the subscription of GPT Plus, at least for a few months. Talk about ROI!

Read all Andrew Birmingham’s coverage and analysis of CX, martech, ecom, adtech and regulation in MI3’s Tech section: https://www.mi-3.com.au/tech